ablation radioguidee des masses renales - Thèses

AIX-MARSEILLEUNIVERSITE

ECOLEDOCTORALEED251Sciencesdel’Environnement

Laboratoired’ImagerieInterventionnelleetExpérimentaleEA4264

Thèseprésentéepourobtenirlegradeuniversitairededocteur

Spécialité:EnvironnementetSanté

ABLATIONRADIOGUIDEEDES

MASSESRENALES

ParPhilippeSOUTEYRAND

Soutenuele18décembre2015devantlejury:

MrleProfesseurN.GRENIERUniversitédeBordeauxPrésident,Rapporteur

MrleProfesseurO.ROUVIEREUniversitédeLyonRapporteur

MrleProfesseurJM.BARTOLIAixMarseilleUniversitéExaminateur

MrleProfesseurC.CHAGNAUDAixMarseilleUniversitéExaminateur

MrleProfesseurG.SOULEZUniversitédeMontréalDirecteurdethèse

MrleProfesseurV.VIDALAixMarseilleUniversitéDirecteurdethèse

Remerciements

Agathe,Clément,Jules,Arthuret…

Pourmongrand-pèrePierreSouteyrandquicroyaitplusquenoustousaux

progrèsdelarecherchemédicaleenFranceetauCHU.

Les institutions qui ont financé nos travaux: la Fondation Santé Sport et

Environnementd’Aix-MarseilleUniversité,laSociétéFrançaisedeRadiologieet

la Société d’Imagerie Génito-urinaire, le Réseau de Bio-Imagerie du Québec

(RBIQ)etleFondsdeRechercheFranceCanada.

Remerciements

1

Tabledesmatières

Tabledesmatières....................................................................................................................1

Listedespublications................................................................................................................2

Tabledesannexesetdesillustrations.......................................................................................3

Résumé......................................................................................................................................4

Abstract.....................................................................................................................................5

Introduction...............................................................................................................................6

Lecancerdurein......................................................................................................................8

Laradiofréquencerénale........................................................................................................10

Lesremaniementstissulairesettomodensitométriquesinduitsparlaradiofréquence.........12

NotreexpériencecliniqueenRFA–lesscoresmorphométriques..........................................13

Notreexpérienceclinique..................................................................................................14

LesscoresmorphométriquesetlaRFA...............................................................................14

Développementd’unmodèledetumeurrénaleanimale.......................................................17

Modèlevivantdetumeurrénaleanimale..........................................................................17

Modèleinertedetumeurrénaleanimale...........................................................................25

Modélisationdynamiquedureinparuneapprochedetypemorphing.................................29

DescriptiondesmouvementsdureinenIRM........................................................................32

Trackingdurein......................................................................................................................37

Conclusion-perspectives........................................................................................................41

Bibliographie............................................................................................................................43

Publicationsenrapportaveclathèse.....................................................................................45

Conférences en rapport avec la thèse.....................................................................................103

Abréviations..........................................................................................................................103

Annexes.................................................................................................................................104

2

Tabledesannexesetdesfigures

Annexe1:classificationOMSdestumeursdurein

Annexe2:critèresdéfinissantlegradenucléairedeFührman

Annexe3:classificationTNMdescarcinomesrénaux

Annexes4/5:notionstechniquesenRFAetexempledeprotocoledechauffe

Annexes6/7:articlesdePH.Rolland(nanotubes)etV.Leonardi(morphing)

Figures1/2:lesscoresPADUAetR.E.N.A.L.

Figure3:CalculduC-index

Figure4:lesdifférentstypeshistologiquesimplantés

Figure5:tauxsanguintotaldeTacrolimusdansles24heuresquisuiventsoningestion

Figure6:exempledezoned’implantationsurlesscannersdesurveillance

Figure7:aspectmacroscopiquedesreinsexplantés

Figure8:réactioninflammatoiresurlesited’implantationdel’agrégattumoral

Figures9et10:coïlsenplatineetenacierembolisés

Figure11:embolisationdesegmentsrénauxparduCuraspon

Figure12:échantillonsdemélangedeChitosanetdechélatesdegadolinium

Figure13:pôleinférieurdureinemboliséparunmélangeChitosan-chélatedegadolinium

Figure14:enregistrementducyclerespiratoire

Figure15:exempledesegmentationdurein

Figure16:étudedesmouvementsdureindécomposésentranslationetenrotation

Figures 17 et 18: résultatsdesmesuresde translationetde rotationdes reinsdans les3

planschez10volontairessains

Figure19:exemplederecalageetd’analysedesrésultats

Figure 20: recalage2D/3Davec laCCet la fonctiond’optimisationSimplexdes imagettes

sous-échantillonnéesavecunepyramideLaplacienne

Figure21:recalage2D/3DaveclaCCetlafonctiond’optimisationSimplexd’uneimageflash

Figure22:lesrésultatsdecorrélationenmodifiantlesparamètresduplandel’imagetteet

enintégrantunerotationaurein

3

Publicationsenrapportaveclathèse

I. Souteyrand P, Chagnaud C, Lechevallier E, André M. Radiofrequency ablation of

kidneytumors.ProgUrol.2013Nov;23(14):1163-7

II. SouteyrandP,CohenF,Daniel L, LechevallierE,ChagnaudC,RollandPH,AndréM,

VidalV.Pathologicalfeaturesofradiofrequencyablation(RFA)renalscarCT-imaging

inaswinemodel.ProgUrol.2013Feb;23(2):105-12

III. Souteyrand P, André M, Lechevallier E, Giorgi R, Chagnaud C, Boissier R. Using

morphometric scores to predict RFA complications in renal tumors under 4cm.

SoumisJournalofUrology

IV. DesmotsF,SouteyrandP,MarcianoS,LechevallierE,ZinkJV,ChagnaudC,AndréM.

Morphometric scores for kidney tumours: use in current practice. Diagn Interv

Imaging.2013Jan;94(1):116-8.doi:10.1016/j.diii.2012.07.001.Epub2012Dec13

V. SouteyrandP;ChungW;DeGuiseJ;MariJL;LeonardiV;OliviéD;VidalV;SoulezG.A

MRIDescriptionofKidneyMotion.SoumisJournalofMagneticResonanceImaging

VI. ChungW,SouteyrandP,SoulezG,ChartrandG,CressonT,MariJL,DeGuiseJ.Slice-

to-volume registration for kidney guided interventions using MRI. En cours de

relectureavantsoumissionTransactionsBiomedicalEngineering

4

Résumé

La prise en charge thérapeutique des tumeurs rénales a considérablement évolué ces

dernièresannéesavec l’avènementdetraitementsmini-invasifs (commelaradiofréquence

percutanée)quioptimisentl’épargnenéphronique,amélioreleconfortdupatientavecune

efficacitéoncologiquecomparableauxtraitementschirurgicauxderéférence.Laprochaine

étapeseraitdeproposerdestraitementstranscutanés(HIFU,radiothérapiestéréotaxique…)

aussiperformantsavecunemorbi-mortalitéoptimisée.

L’objectif des travaux réalisés au Laboratoire d’Imagerie Interventionnelle et

Expérimentale du Centre Européen de Recherche en Imagerie Médicale (Aix-Marseille

Université) et au Centre de Recherche du Centre Hospitalo-Universitaire de Montréal

(UniversitédeMontréal),pourlathèseetlamobilité,étaitdedévelopperunealternativeà

la radiofréquence rénale percutanée que nous utilisons en pratique clinique à Marseille

depuisplusde10ansetquiafaitsespreuves.Cettealternativedoitpermettredetraiterdes

masses rénales avec un niveau d’efficacité et un taux de complications au minimum

identiqueà laRFA,parapplication transcutanéed’agentsphysiques sansabordpercutané

(projet KiTTKidneyTrackingTumor). Cela passe par lamise au point d’une technique de

détectionentempsréeldelamasserénale.

Nousavonspudévelopperunalgorithmederepéragefiablequidoitencoreêtreoptimisé

(rapiditédecalcul)etêtrevalidésurunmodèlequin’estpasencoredisponible.Lestravaux

d’optimisationetdevalidationdesalgorithmesdesegmentation,defonctiondeméritede

corrélationcroiséeassociéeàlafonctiond’optimisationSimplex,sontencoursdanslecadre

d’unecollaborationinternationalefranco-canadienneauLIIEetauLIO.

Mêmesinousn’avonspasencore lapossibilitédeproposerce typede traitement,nos

travaux permettent de s’en approcher pour pouvoir les proposer dans les prochaines

années.

Motsclés:rein,cancer,traitement,IRM

5

Abstract

Thetherapeuticmanagementof renal tumorshaschangedconsiderably inrecentyears

withtheadventofminimallyinvasivetherapies(suchaspercutaneousradiofrequency)that

maximizenephronsavings,improvespatientcomfortwithefficiencycomparabletosurgical

oncology treatments reference. The next step would be to propose transcutaneous

treatment (HIFU, stereotactic radiotherapy ...) as efficient with optimized morbidity and

mortality.

TheobjectiveofthisworkinthecontextoftheLIIEofCERIMED(Aix-MarseilleUniversité)

andCRCHUM(UniversitédeMontréal)wastodevelopanalternativetopercutaneousrenal

radiofrequency we use in clinical practiceMarseille for over 10 years and has proved its

worth.Thisalternativemustbecapableoftreatingrenalmasseswithalevelofeffectiveness

andcomplicationratesatleastequaltotheRFA,byapplyingtranscutaneousphysicalagents

without percutaneous approach (project KITT (Kidney Tracking Tumor)). This requires the

designoftechnicalpointdetectioninrealtimeoftherenaltumor.

Wewereabletodevelopareliableidentificationalgorithmthathasyettobeoptimized

(speed of calculation) and be validated on a model that is not yet available. Work

optimization and validation of segmentation algorithms, cross correlation merit function

associated with Simplex optimization function, are underway as part of an international

collaborationtoFrench-CanadianLIIEandLIO.

Even ifwehavenottheopportunitytoofferthistypeoftreatment,ourworkallowsto

approachinordertooffertheminthecomingyears.

Keywords:kidney,cancer,treatment,MRI

6

Ablationradioguidéedesmassesrénales-introduction

L’incidence et la mortalité des cancers du rein se stabilisent en France et dans les pays

occidentaux [1]. Avec l’évolution des techniques d’imagerie, le profil des tumeurs du rein

diagnostiquées a aussi évolué [2]. Elles sont le plus souvent de découverte fortuite alors

qu’elles sont asymptomatiques. Parce que le recours aux examens d’imagerie est plus

fréquentetqu’ilssontplusperformants,ellessontpluspetitesaudiagnosticqu’auparavant.

L’avènement de la biopsie percutanée radioguidée par l’échographie ou la fluoroscopie a

permis de confirmer le diagnostique histologique de cancer du rein sans exérèse de la

masse, avec un faible taux de complications. La biopsie a une place bien définie dans

l’algorithmedepriseenchargedesmassesrénales[3]:touteslesmassesrénalesdenature

indéterminée. Parallèlement la prise en charge des cancers du rein a évolué. La chirurgie

radicale(néphrectomie)parlombotomien’estpluslaseulealternativethérapeutique:sous

certainesconditions,onpeutproposerunenéphrectomiepartielleouunethermo-ablation

percutanée.

Laconjonctionde l’améliorationde laprécisionde l’imageriederepérage,de lapossibilité

d’avoir un diagnostique de certitude sans chirurgie d’exérèse et de la petite taille de

certaines masses a permis le développement et la validation des techniques de thermo-

ablationpercutanéedesmasses rénalescomme la radiofréquence (RFA)et lacryoablation

(CR). Ces techniques peuvent dans certaines indications se substituer au traitement

chirurgical.Ellespeuventsefaireenlaparoscopieousousguidageradiologique.

L’objectifdestravauxréalisésauLaboratoired’ImagerieInterventionnelleetExpérimentale

(LIIE) du Centre Européen de Recherche en Imagerie Médicale (CERIMED – Aix-Marseille

Université) était de développer une alternative à la radiofréquence percutanée que nous

utilisons en pratique clinique àMarseille depuis plus de 10 ans et qui a fait ses preuves.

Cettealternativedoitpermettredetraiterdesmassesrénalesavecunniveaud’efficacitéet

un taux de complications au minimum identique à la RFA, par application transcutanée

d’agentsphysiquessansabordpercutané.Mêmesinousn’avonspasencorelapossibilitéde

proposer ce type de traitement, nos travaux doivent permettre de s’en approcher pour

pouvoirlesproposerdanslesprochainesannées.

Ilss’intègrentdansleprojetKiTT(KidneyTrackingTumor)initiéparlaFondationSanté,Sport

et Environnement d’Aix-Marseille Université. L’objectif de KiTT était de développer une

méthode de suivi en temps réel d’une tumeur rénale sans préjuger d’une modalité

d’imagerienid’unefinalitédiagnostiqueouthérapeutique.Avecl’intégrationderadiologues

dansl’équipe,l’objectifaévoluéversladéveloppementd’untraitementdesmassesrénales

«non invasif»c’estàdiresansabordpercutané.D’unprojetd’ingénierie,KiTTestdevenu

unprogrammeintégrantunecomposantecliniqueàlacomposanteinformatiqueinitiale.

7

Les2premièrespartiesdecettethèsereplacent lecancerdureinet laRFArénaledans le

contexteactuel.MêmesilaRFAétaitdéjàproposécommetraitement,etqu’elleavaitfaitla

preuvedesonefficacitéclinique,nousavonsanalysé(3èmepartie)leseffetstissulairesdela

RFA sur le parenchyme rénal sur un modèle animal, et corrélé ces remaniements

histologiques,notammentlanécrose,avecl’aspectenscanner.Undesobjectifssecondaires

decetteétudeétaitd’associerlaRFAetd’autresagentspouroptimiserencoreletraitement.

L’adjonction par voie endovasculaire de nanotubes dans le rein devait permettre de

potentialiserl’actiond’agentsphysiquesexternespourréaliserdestumorectomies.

Laquatrièmepartiedecettethèsereprendlestravauxcliniquesavecl’étudeprétraitement

descancersdureinparlesscoresmorphométriques,quecesoitenbilanpréopératoiremais

aussi avant une RFA. Nous avons comparé notre expérience clinique en RFA avec les

données de la littérature: l’analyse rétrospective des suivis des patients ayant bénéficiés

d’uneRFAdansnotreserviceamontrél’efficacitéoncologiquedelaRFA.

Danslemêmetemps,nousavonsessayédedévelopperunmodèleanimaldetumeurrénale.

L’intérêt de ce modèle serait de pouvoir tester un vivo la précision des traitements

transcutanéssurlerein:HIFU,radiothérapie…Lesdifférentesétapesdeceprotocolesont

décrites dans la 5ème partie de la thèse. Comme toutes les équipes qui travaillent

actuellementsurceprojet,nousn’avonspasréussiàdéveloppercemodèle.Pourpalliercet

échec,nousavonschoisidesimulerdescibles,processusquenousdétaillons.

Les 6ème et 7ème partie traitent de la segmentation du rein et de la description des

mouvements du rein en IRM. Avec l’aide des équipes de l’Université de Montréal, nous

avonspuutiliseretoptimiserunde leursalgorithmesde segmentationquinousapermis

de:

-décrirelesmouvementsphysiologiquesdureinetsuivreentempsréelcesmouvements;

-mettreaupointunlogicieldetrackingdureinentempsréelenIRMetlevalider.

Ce logiciel en développement, l’accélération des temps de traitements et des acquisitions

IRMnous laissentenvisagerdenombreusesperspectivesdetravaux:suivreentempsréel

les mouvements du rein pour délivrer, par voie externe, un agent (HIFU, radiothérapie

stéréotaxique…)capablededétruirecettecible.Cestravauxsontencoursàl’Universitéde

Montréaldanslecadred’unecoopérationinternationaleavecAix-MarseilleUniversité.

Ce travailde thèse«ablation radioguidéedesmasses rénales»s’articuleautourduprojet

KiTT,cofinancéparlaFondationSanté,SportetEnvironnementd’Aix-MarseilleUniversité,la

SociétéFrançaisedeRadiologieet la Sociétéd’ImagerieGénito-urinaire, leRéseaudeBio-

Imagerie du Québec (RBIQ) et le Fonds de Recherche France Canada. Il a été réalisé à

MarseilleetàMontréaldanslecadred’uneannéedemobilité.

8

1. Lecancerdurein

1.1. Epidémiologie

L’incidenceducancerdureinenFranceestestiméeà10125casen2009[4].Ilreprésenteenviron3%destumeursmalignesdel’adulte.SonincidenceenFranceetauxEtats-Unis[5-7] est en augmentation depuis une trentaine d’années, en rapport avec un nombre plusimportantdedécouvertesfortuites.

Ilestdeuxfoisplusfréquentchezl’homme.L’âgemoyendudiagnosticsesitueà65ans.Lenombrededécèsestimésen2009estde3830.Cechiffreestenbaisse,enpartieliéeàunedécouverteplusprécocedecescancers.Eneffet,lasurvierelativeà5ansestglobalementde 63 % [8]. Pour un stade localisé (58 % des diagnostics), elle passe à 90 %. Le pic demortalitésesitueentre75et85ans.Cette augmentation de l’incidence est d’origine multifactorielle: amélioration desperformances de l’imagerie diagnostique (sensibilité de 91% de l’échographie dans ladétectionducancerdureinavecunespécificitéde96%[9])etaugmentationdesfacteursderisque de développer un cancer du rein corrélée au vieillissement de la population (HTA,obésité,insuffisancerénaleoutransplantationrénale[10]).Lacroissancedunombred’actesd’imagerie, notamment des scanners et des échographies de l'appareil urinaire [Rapportd'activitéCCAMdécembre2008],participeprobablementàl'augmentationdel'incidencedudiagnosticdetumeursasymptomatiques(60à70%destumeursdurein[11]).Les facteurs de risque reconnus sont la dialyse depuis plus de 3 ans (qui entraîne unedysplasie multi kystique et des carcinomes tubulopapillaires), l’obésité et le tabagisme.L’hypertensionartérielle,l’expositionàl’amianteetaucadmiumsontsuspectéesd’êtredesfacteursderisque. Ilexistedes formesfamilialeshéréditairesdont laplus fréquenteest lamaladiedeVonHippel-Lindau.

On comprend que le profil des patients et des tumeurs rénales découvertes se soientmodifiés: découverteprécocedepetitesmasses rénales asymptomatiques, diminutiondel'âgemoyendespatientsaumomentdudiagnostic(62ansenmoyenne),etaugmentationdescomorbiditésdepatientsâgés.

1.2. Diagnosticclinique,biologiqueetradiologiqueLamajorité des cancers du rein est diagnostiquée alors qu’ils sont asymptomatiques, lorsd’examensradiologiques.La lombalgie, l’hématurie, la masse abdominale, les syndromes paranéoplasiques ou lessignes secondaires aux métastases (osseuses ou pulmonaires) sont des symptômes d’unstade avancé de la maladie. L’examen clinique est peu contributif. Le bilan biologiquecomporte une créatininémie pour calculer sa Clairance, reflet de la fonction rénale, undosagedel’hémoglobine,desLDHetunecalcémiecorrigée.L’échographie(avecdoppler)estsouventlepremierexamenàévoquerlediagnosticquiest

9

secondairement confirmé par scanner et / ou IRM (avec injection IV de produit decontraste): ilsconfirment laprésenced’unemasserénale, rehaussée.Lebiland’extensionlocaleetàdistance(envahissementganglionnaire,vasculaire,desorganesdevoisinageetdureincontrolatéral,métastasespulmonaires…)peutêtreréalisédanslemêmetemps.L’avènementde laponction-biopsie rénale sousguidageéchographiqueou fluoroscopique[12]permetlediagnostichistologiqueavantl’exérèse.

1.3. Histologie,classificationTNMLaclassificationOMSde2004distingue les tumeurs rénalesbénignesetmalignes (Annexe1).85%descancersdureinsontreprésentésparlescarcinomesàcellulesrénales(CCR).Lesautres types histologiques sont les cancers papillaires type 1 et 2 (10-15%) et les cancerschromophobes (4-5%). La classification des cancers du rein se fait sur des critèreshistologiquesetgénétiquesetledegréd’agressivitéestévaluéselonlescritèresdeFührman(Annexe2).LesCCRreprésentent lagrandemajoritédescancersdureinde l’adulte.Cinqautrestypeshistologiquesetdenombreuxsous-typeshistologiquesconstituentles15%restants.Seulela prise en charge des CCR est détaillée dans cette partie, les autres types histologiquespeuventfairel’objetdeprisesenchargespécialisées.Plusieurs classifications comme la TNM (Annexe 3) et des systèmes pronostiques ounomogrammessontdécrits.NousdécrivonsplusloinlesscoresmorphométriquesPADUA,lescoreR.E.N.A.L.etleC-Indexquiontunintérêtpourprédirelerisquedecomplicationsperetpost-opératoires.

1.4. BiopsiedestumeursdureinLa biopsie rénale guidéepar l’imagerie (échographieou scanner) a uneplacebiendéfiniedanslapriseencharge[12]:

• Contextedecancerextrarénalconnu:distinctionentreuncancerdureinprimitifetunemétastase;

• Suspicion de cancer rénal non extirpable (localement avancé et/ou multimétastatique), cancer du rein métastatique quand une néphrectomie n’est pasenvisagée;

• Tumeurspourlesquellesuntraitementablatifestenvisagé;

• Patientsaveccomorbiditésnotables:déterminationdurapportbénéfice/risqued’untraitementvslasurveillanceactive;

• Tumeursrénalessurreinunique;

Lesindicationsdeprincipepourlespetitestumeursrénalessolides(<4cm)indéterminéesparl’imagerierestentdiscutéesquandunenéphrectomiepartiellepremièreestpossible.

10

Laméthodologiehabituelleest:vérifier lebilandecoagulation, lastérilitédesurineset latension artérielle, chez un patient hospitalisé ou en ambulatoire, réaliser au moins 2prélèvementsavecuneaiguillede18Gauges.

Les résultats sont excellents avec une sensibilité / spécificité > 90 %, un taux de biopsiecontributivede80%,unedéterminationexactedutypehistologiquedans80-90%etunedéterminationexactedugradedeFührmandans50-75%.

Les complicationssont peu fréquentes et exceptionnellement graves [13] : décès 0,02%,saignement0,3%.Ladouleurpost-biopsierestelaprincipalecomplication.

1.5. Traitementsactuels

Quelque soit le traitement proposé, la prise en charge estmultidisciplinaire avec décisionlorsd’uneRéuniondeConcertationPluridisciplinaire(RCP).

Le traitement de référence du cancer du rein localisé est la chirurgie par voieconventionnelleoulaparoscopique,totaleoupartielleselonlescaractéristiquesdelalésion[12].Encasdecancerdureinmétastatique,lapriseenchargepeutassocieruntraitementanti-angiogénique.

Ilyaeucesdernièresannéesundéveloppementdelachirurgieconservatrice(néphrectomiepolaire,tumorectomie)pour2raisons:

-de nécessité (préservation optimisée du parenchyme rénal): rein unique,insuffisancerénalechronique,maladiedeVonHippel-Lindau…

-deprincipe:faiblestade/situationanatomiquefavorable.

Les tumeurs exophytiques de moins de 4cm [11] sont la principale indication desnéphrectomiespartielles(néphrectomieélargiepourlestumeursintraparenchymateusesetsinusales)avecunesurvieidentiqueparrapportàlanéphrectomieélargie[14].

Les traitements mini-invasifs (cryothérapie, radiofréquence, ultra-sons, micro-ondes) sontenvisageables depuis l’avènement de la néphrectomie partielle. Ces traitements sont destechniquesd'ablation(paroppositionauxtechniquesd'exérèse)indiquéespourlestumeursdemoinsde4cm.Lanécessitéd’avoirunepreuvehistologiquepourengager letraitementoncologique [10] et l’accès à des techniques d’imagerie de repérage fiable renforcentl’intérêtdelathermo-ablation.

11

2. Laradiofréquencerénale[I]

La radiofréquence (RFA) est un de ces traitements mini-invasifs[I: Souteyrand P et al.Radiofrequency ablation of kidney tumors. Prog Urol. 2013 Nov]: elle consiste en uneablation thermique de la tumeur par excitation moléculaire par un courant de RF,responsable d’une nécrose de coagulation (Annexe 4 : notions techniques enradiofréquence).

Ondistingue2typesd’indicationspourlaRFA:

1. Lesindicationsdenécessité

masse tissulaire de moins de 40mm, non sinusale, remplissant au moins une de cesconditions:

! >70anset/oufacteursdecomorbidité,

! reinuniqueet/oufonctionrénalealtérée,

! néoplasieadjacenteet/oulocalisationsbilatérales,

! cancerhéréditaire(VHL).

2. Les indications électives: c’est à dire où l’on propose ce traitement sur les seulescaractéristiquesdelatumeur(<4cm,nonsinusale).

Les indications de nécessité ont pu être élargies aux indications électives parce que lesrésultats [15] et le taux de complications [16] étaient équivalents à ceux de la chirurgiepartielle.Laréussiteestessentiellementcorréléeàlatailledelalésioninitiale[15].

L’efficacitédutraitementestévaluéeenimagerieparl’absencederehaussementsignificatifaprèsinjectiondeproduitdecontraste[17].

L’expérience des RFA rénales de notre centre est rapportée dans la partie 4 maisl’organisationd’uneprocéduredeRFAestrelativementstandardisée:

-unebiopsiepréalabledoitconfirmerlediagnosticdecancerrénal,ladécisiondetraitementestpriseenRCP;

-lepatientestsurveillésousanesthésiegénérale,parfoissousneuro-analgésie,

-l’aiguilledeRFAestpositionnéesousguidagedel’imagerie(fluoroscopieplusfréquemmentqu’échographie),

-ladestructionthermiquedelacible(exempledeprotocoleenAnnexe5).

-la surveillance en imagerie post-RFA suit les recommandations post néphrectomiespartielles:scannerrénalou IRM(avec injectiondeproduitdecontraste)à3,6et12moispuisannuellementaprèslaRFA,pendant10ans.

12

3. Lesremaniementstissulairesettomodensitométriquesinduitsparla

radiofréquence[II]

LaRFAaprouvésonefficacitéoncologiquedansdenombreusespublicationsavecunsuccèscarcinologiqueéquivalentàlachirurgie.

Mais, sielleaprouvésonefficacitésur leplande l’imagerie (absencederehaussementetd’évolutivitédetaillede lazoned’ablation(ZA)), lesremaniementsqu’elle induitn’avaientpasencoreétédécrits.

C’est ce que nous avons fait en condition expérimentale animale [II: Souteyrand P et al.Pathological features of radiofrequency ablation renal scar CT-imaging in a swine model.

ProgUrol.2013Feb]:corréler,avecduparenchymerénalsaindeporc, l’évolutiondans le

tempsdelaZAauscanneraveclesremaniementshistologiques.

Les porcs anesthésiés ont subi des RFA aux pôles des reins puis ils ont été surveillés enscanner (sansetavec injectiondeproduitde contraste)et lesZAontétéétudiéesparunpathologistespécialiséaudécoursdelaprocédureouplusieurssemainesaprès.

Cetteétudeamisenévidencedesremaniementstissulaireshétérogènesdansleurformeetleur distribution: des tissus ischémiés, inflammatoires,mésenchymateux ou nécrosés auxlimites nettes ou flous, et une surexpression de l’apoptose. Ces remaniementsanatomopathologiques n’ont pas pu tous être corrélés formellement avec leur aspecttomodensitométrique.

Ces résultats ont remis en cause certaines notions sur la RFA rénale comme l’intérêt ducontrôle tomodensitométrique post-RFA pour prédire le succès thérapeutique, le délai dupremiercontrôledesurveillanceetlerehaussementdelaZAcommecritèresd’insuffisancedetraitement.Elleproposaitdedifférerau-delàd’unmoisaprèslaRFAlepremiercontrôletomodensitométrique,d’associersystématiquement l’étudedurehaussementde laZAà larépartitiondurehaussementetàlaprogressionvolumétrique.

Encasdediscordanceentrel’évolutiondetailleetlerehaussementdelaZA,onproposeunesurveillancerapprochéeet/ouunebiopsiede laZA.Lespatientspeuventalorsbénéficierd’uncomplémentdetraitementparRFAouparchirurgie.

Lamise en évidenced’une réactiond’apoptose induite par la RF au sein de tissus sains àdistance de la zone de traitement signifiait que la nécrose secondaire à l'élévation de latempératurelocalen'étaitpeutêtrepaslaseulevoiededestructiontissulaireetouvraitdenouvellesvoiesderecherche.

DansleprolongementdecetteétudeauLIIE,PHRollandetal[Annexe6]ontdémontréquel’associationdenanotubes(multi-walledcarbonnanotubes(MWCNT)),deMarsembol(non-adherent, lipophilicembolicagent)etdeRFAentrainait ladémarcation laplusnetteentreles cellules viables et les cellules apoptotiques. Sans les nanotubes, cette limite était plusfloue.Lànonplus,lemécanismed’actiondesondesdeRFAetdesnanotubesn’apaspuêtreidentifié mais cela ouvre la voie à des applications: embolisation sélective de massestumorales hyper vascularisées puis application externe d’agents physiques pour entraineruneapoptoseetunenécrosetissulairehypersélective.

13

4. NotreexpériencecliniqueenRFA–lesscoresmorphométriques[III,IV]

4.1. Notre expérience clinique [III: Souteyrand P et al. Using morphometric scores topredictRFAcomplicationsinrenaltumorsunder4cm.SoumisJournalofUrology]

160patientsontbénéficiéd’uneradiofréquencerénalepourcancerces10dernièresannéesdanslesservicesderadiologiedel’AssistancePubliquedesHôpitauxdeMarseille.

Tous lespatientsontvuenconsultationunurologueetunradiologue,ontbénéficiéd’unebiopsieradioguidéeavantqueladécisiondetraitementaitétévalidéeenRCP.LeDrAndréaréalisé la majorité de ces traitements, et a supervisé ceux pour lequel il n’était pasl’opérateurprincipal.Lesprocéduressousanesthésiegénéraleonttoutesétécontrôléesparscanner.

Les résultats de notre série sont développés dans l’article qui évalue les scoresmorphométriques pour la RFA,mais ils confirment ceux de la littérature que ce soit pourl’efficacitédelaRFAàcourtetàlongtermeetpourletauxdecomplications[18].Les160patientstraitéssurcettepériodeontétéincluspour180RFA:145patientsonteuuneRFApour une lésion, 10 patients ont eu plusieurs RFA pour différentes lésions, 5 patients ontbénéficiéd’une2°RFAsurlaZApourrécidiveoutraitementincomplet.Pour les 10 patients qui ont eu des RFA pour différentes lésions (synchrones oumétachrones), 7 ont été traités pour 2 lésions, 2 pour 3 lésions et un pour 5 lésions. Lescaractéristiquesdespatientstraitéssontrésuméesdansl’article.

Trente complications ont été identifiées (7.22%): 7 mineures (Clavien I–II dont 3 pourdouleurs au point de ponction et 4 hématomes) et 6 majeures (Clavien III–IV–V dont 4obstructionsurétéralesdrainéeset2décès).Undécèsestliéàuneruptured’anévrismedel’aorteabdominalelelendemaindelaRFA.Lesecondestconsécutifàunedéfaillancemulti-viscérale48heuresaprèslaRFA:lefacteurderisquequel’onaitidentifiéétaitlemaintiendutraitementanti-angiogéniquependantlaprocédure(lepatientn’avaitpasinterrompuletraitementquiestcontreindiquépendantlaRFA).

Le suivi moyen en scanner et / ou IRM était de 25.5 mois (rythme 3, 6 et 12mois puisannuellement).Lasurveillanceavaitmontréuntraitementefficacepour132/180procéduresdeRFA(73%)pour2traitementsincompletset1récidivesurlaZA(1.7%).Dansles35autrescas (19.4%), il n’y avait pas eu de récidive sur la ZA mais l’apparition de nouvelleslocalisations rénalesouàdistanceet les2décèsdans les suites immédiatesde laRFA.10patientssontexcluscarlasurveillancepostRFAétaitinférieureà6mois.

Lesqualitésparticulièresdenotresériesontl’homogénéitédepriseenchargedespatientsdepuisquenousproposons ces traitements: unpetitnombred’opérateursexpérimentés,touteslesprocéduresréaliséesselonlamêmetechniqueavecunguidagefluoroscopiqueetunesurveillancerigoureuseparscanneret/ouIRM.

Les résultats ont confirmé que la RFA est une technique efficace pour le traitement des

tumeurs rénales (T1a ≤ 4 cm): la RFA peut être proposée comme traitement électif enalternativeauxtraitementschirurgicaux,avecdesrésultatsoncologiquesidentiques.

14

4.2. LesscoresmorphométriquesetlaRFA[III,IV]

Lesscoresmorphométriques(SM)[IV:DesmotsF,SouteyrandPetal.Morphometricscores

forkidneytumours:useincurrentpractice.DiagnIntervImaging.2013Jan]quiévaluentla

complexitédestumeursrénalesàpartirduscanneroudel’IRMontprouvéqu’ilsétaientdes

indicateursprédictifsfiablesetreproductiblesdescomplicationsperetpost-opératoiresde

lanéphrectomiepartielle[19,20].

Les scores PADUA (Figure 1) et R.E.N.A.L. (Figure 2) sont calculés en tenant compte de la

tailledelatumeur,desapositionauxpôlesrénaux,desaprofondeuretdesesrapportsavec

lesystèmecollecteuretlesinus[21-23].LeC-index(Figure3)estlerésultatdurapportdela

distanceentrelescentresdelatumeuretdurein,aveclerayondelatumeur[24].

Figures1et2:scoresPADUAetR.E.N.A.L.

15

Figure3.CalculduC-indexparlerapportdeladistancedes2centres(C)diviséparlerayondelamasse.

Ces scoresontdéjàétéétudiéspour la thermo-ablation rénale,mais laplupartdu temps,

sansdistinctionentrel’abordpercutanéoulaparoscopique,nientrelaRFAetlacryoablation

(CR).SeulSchmitt[25]suggèrequelescoreR.E.N.A.L.préditlerisquedecomplicationspéri

etpost-opératoiresdeRFArénale.Lesautresauteurs[26-28]nemontrentpasderelation.

L’objectif principal de notre étude [III: Souteyrand P et al. Usingmorphometric scores to

predict RFA complications in renal tumors under 4cm. Soumis Journal of Urology] était

d’évaluersilesSMavaientunintérêtpourprédirelerisquedecomplicationdesRFArénales

radioguidées?Nousavonsconfrontéces3 scoresà la sériedepatients traitésdansnotre

centreparRFA.

L’objectif secondaire était de proposer une optimisation des scores pour les traitements

ablatifs percutanés en adaptant le critère de taille tumorale et en tenant compte

notammentdelapositionantéro-postérieuredelatumeurquiestrapportéedanslesscores

PADUAetR.E.N.A.L.maisquinemodifiepasleursvaleurs.

Lesrésultatsdes3scoresetleurcorrélationauxcomplicationsontretrouvédesmoyennes

descoresde6.5pourPADUA(5–10),6pourR.E.N.A.L.(4–10),et3.4pourC-index(0.9–12.8).

Pourlesscoresmodifiés,lesmoyennesdesscoresétaientde4.4(3-7)pourPADUA-modifié,

8.1(6-12)pourPADUA-RFA,3.5(2-6)pourRENAL-modifiéet8.6(6-13)pourRENAL-RFA.

Il n’y avait pas d’association significative entre la présence de complications et le score

R.E.N.A.L. (p=0.11), PADUA (0.18) et C-index (0.67) ni entre les scores et la gravité des

complications(p=0.27,0.32et0.89).Lesmoyennesdesscoresétaientplusélevéesencasde

complications(7mineurespourunemoyennepourR.E.N.A.L.6.4,PADUA6.7etC-index3,et

6majeures (moyennes respectives 6.8, 7.2 et 4.3).Mais les différences de cesmoyennes

selonqu’ilyaitoupasdescomplicationsn’étaientpasimportantes.Iln’yavaitpasnonplus

16

derelationstatistiqueentrelesscoresmodifiés(PADUA-modifiedp=0.27,PADUA-RFA0.39,

RENAL-modifié 0.38 et RENAL-RFA 0.32), les paramètres qui composent ces scores

indépendamment les uns des autres ou entre les catégories de groupes de risque (low,

intermediateandhigh)etlescomplicationsaigües.

LesSMainsique lescritèresdecesscoresétudiés indépendammentn’ontpasde relation

avec les complications aigües donc ils ne permettent pas de prédire le risque de

complicationspourlaRFArénaleradioguidée.

Mêmesionmodifielesscoresousionprendencomptedanslescorelapositionantérieure

/postérieureetenadaptantlecritèredetaille(0<3cm/3–4cm/<4cm)delatumeur,ces

scoresnepermettentpasdeprédirelerisquedecomplicationaigüe.

LesSMconnusetvalidéspourlachirurgiepartiellen’ontpasprouvéleurintérêtpourlaRFA

rénaleradioguidée.L’adaptationdesscoresPADUAetRENALaveclamodificationducritère

detailleetl’intégrationdelapositionantérieureoupostérieuredelatumeurnepermettent

pasnonplusdeprédirelerisquedecomplicationsdesRFArénalespercutanées.

17

5.Développementd’unmodèledetumeurrénaleanimale

5.1.Modèlevivantdetumeurrénaleanimale(travauxréalisésauLIIE)

La principale critique faite aux travaux sur la corrélation histo-radiologique et les

remaniementsinduitsparlaRFAétaitl’absencedecancerdureindenotremodèle.

Lecancerdurein leplusfréquent, leCCR,enplusdedéformer leparenchymenormal,est

hypervascularisé:cettenéo-angiogénèsepeutmodifier lecomportementet leseffetsdes

ondesderadiofréquence(maisaussidel’«ice-ball»delaCR).Lesmassesrénalesquel’on

traitesontgénéralementcorticalesetpassinusalesdoncàdistancedesvaisseauxpyéliques:

ledébitet lesvolumesdesangsontmoins importantsetons’attendàunrefroidissement

limité. L’exemple des radiofréquences hépatiques à proximité des veines sus hépatiques

illustre ce refroidissement par l’afflux de sang: en cas d’extrême proximité, on peut-être

amené à occlure temporairement, par voie vasculaire, la veine sus hépatique au contact

pourquelatempératurepuissemonter.

L’absence de cible tumorale est préjudiciable mais il n’existe pas de modèle animal de

tumeur rénale humaine que l’on puisse utiliser pour la RFA. Le seulmodèle animal est le

«VX2» [29] implantésur les reinsde lapins: les reinsde lapinsont troppetitspoursubir

une RFA sans atteinte des organes adjacents. Ce modèle n’est pas «stable», la tumeur

évoluanttropvite(envahissementlocorégional).D’ailleursMunvern’avaitexplantélesreins

qu’aumaximum15joursaprèslaRFA.

LestentativesdegreffestumoralessursourisNudeousurratsEkersn’ontjusqu’àprésent

pasdonnéde résultats significatifs. Les seulsmodèlesexistants sontdesmodèlespseudo-

tumoraux réalisésàpartirde substances inertes : agarose («Toour knowledgeno reliable

renaltumormodelexiststoevaluateprocedureefficacy»[30]),gélatineouhydrocolloïdes

[31]ouparinjectiondelysattumoraldanslereindechien[32].

Nous avons donc essayé de développer dans le cadre du laboratoire LIIE (anciennement

L2PTV) au sein de CERIMED, un modèle de tumeur rénale animale. L’objectif n’était pas

seulementdeconfirmer lesremaniementshistologiquesprécédemmentdécrits,maisaussi

de permettre d’avoir une cible pour valider le logiciel de tracking du rein développé

parallèlement. Ce modèle aurait de multiples voies d’application pour la recherche

thérapeutique:

• Expérimentation et perfectionnement des techniques mini-invasives chirurgicales

et/ouradiologiques,

• Expérimentationetsurveillancedesdifférentesthérapiesmoléculaires,

• Pharmacologie:connaissanceetmaîtrisedesposologiesminimalesefficaceseteffets

indésirables.

18

Lemodèleanimalchoisiestceluiqui ressemble leplusà l’homme(poursonanatomie, sa

position,savascularisationetsaphysiologie) lereindecochon[33]quiesthabituellement

utilisépourlachirurgieexpérimentaleurologique.

Ces travaux ont bien entendu été réalisées en accord avec les comités d’éthiques de

l’université, les soinsenaccordavec le«Guidepour l’utilisationet les soinsd’animauxde

laboratoire»[34],etavecleconsentementdespatientspourlesprélèvementstissulaires.

5.1.a.Matérielsetméthode:implantationdestumeurs

Les protocoles d’anesthésie générale, de surveillance per et post-opératoire étaient ceux

décrits dans la littérature, ils sont détaillés dans l’article sur la Corrélation

anatomopathologie–tomodensitométrie[II].

Les techniques d’implantation ont été utilisées chacune sur les 2 pôles des 2 reins de 20

cochonssoit80sitesd’implantation.

• La1ère techniqueconsistaità implanterunagrégatdecarcinomerénalhumain fraisen

sous capsulaire. L’abord était chirurgical abdominal, à ciel ouvert. Pour la 2ème série,

l’agrégataétéimplantéparvoiechirurgicaleensituationcortico-médullaire.Pourla3ème

série, l’agrégat a été implanté en situation sous capsulaire par ponction radioguidée

(scanner et échographie) Pour la dernière série, l’agrégat a aussi été implanté sous

guidageradiologiquemaisensituationcortico-médullaire.Cesméthodesd’implantation

ontétéchoisiespourleurfacilitéetleurreproductibilité.

Comptetenudel’implantationimmédiatedugreffontumoraletafindenepasmodifier

lescaractéristiquestumorales,laconservationdesgreffonstumorauxsefaisaitdansdu

sérumphysiologiqueglacéexclusifsansmilieudeculture.

• Choixdestumeurs implantées:ellesétaientprélevéessurdespiècesdenéphrectomie

totale chez des patients traités par l’équipe médicale du service d’urologie du Pr

Coulange. Les histologies des tumeurs étaient connues avant d’être implantées parce

elles avaient toutes été biopiées, et que le diagnostic avait été confirmé sur la pièce

d’exérèse. Le prélèvement (>1cm3) était organisé en collaboration avec

l’anatomopathologiste (Pr L. Daniel) en post-opératoire immédiat, avant leur

conditionnement pour l’examen pathologique: tous les types histologiques ont été

prélevés (carcinome à cellules conventionnelles, tubulo-papillaire, chromophobe et

urothélial)(Figure4).

19

Figure4:lesdifférentstypeshistologiquesdetumeurshumainesimplantées.

• Soins post-opératoires: les animaux étaient replacés en box collectif selon leurs

conditionsdeviehabituellesavecaccès libreà l’eauetà l’alimentation.Lasurveillance

cliniqueportaitsurl’étatgénéral,lareprisedel’alimentationetd’untransitsatisfaisants

et la cicatrisation. La surveillance (biologiqueeten imagerie)était réaliséedemanière

régulière.

En cas de mauvais état général, les animaux étaient sacrifiés après réalisation d’une

échographie et d’un scanner rénal avec injection de produit de contraste; les reins

implantésétaientprélevésetadressésenanatomopathologiepouranalyse.

• LesporcsontétésacrifiésparinjectionIVd’unbolusde15mgdemidazolametde25mg

dechlorpromazineavec20mldeKCl15%.Lesreinsprélevésétaientfixésdansduformol

pendant48heures.

• Lemêmepathologiste spécialiste de la pathologie rénale a examiné les reins prélevés

avec une colorationhématoxyline-éosine-safran (HES) et par un marqueur immuno-

histochimique, l'anti-CD10 (unmarqueurdesborduresenbrossedes tubescontournés

proximaux).

• Iln’yavaitpasd’étudestatistiqueprévue,seulementuneétudedescriptive.

5.1.b.Choixd’uneimmunosuppression

Compte tenu du concept de xénogreffe tumorale avec implantation de tumeur humaine,

l’introductionetlesmodalitésd’uneimmunosuppressionontétédiscutéesenconcertation

avec les équipes d’oncologie et de néphrologie et en tenant compte des résultats de la

littérature[35].

Le choix de la ciclosporine comme molécule immunosuppressive agissant sur l’ILA2 était

justifié par analogie avec lemécanisme de rejet de greffe humaine faisant intervenir une

20

cascade de cytokines associant demanière prédominante l’ILA2. Par ailleurs, le travail de

référence de modèle de greffe tumorale sur modèle canin a été réalisé avec

immunosuppression par Ciclosporine selon des posologies de 25mg/kg deux fois par jour

pendant 6 mois [36]. La posologie initiale de l’étude a été choisie par analogie avec les

posologieshumainesd’immunosuppressionentransplantationrénale(4-10mk/kg)soitune

dose de 10mg/kg deux fois par jour. Cette dose a été adaptée car les premiers dosages

montraientdestaux10foissupérieursauxobjectifs.

La surveillance de l’efficacité de l’immunosuppression a été faite selon les modalités de

surveillanced’immunosuppressiondetransplantationrénalehumaine:dosagesanguindela

ciclosporinémieà12heuresdelaprécédenteinjectionjusteavantl’injectionsuivanteavec

uneciclosporinémiecibleà100ng/ml.Comptetenudesdifficultés techniquesdesdosages

sanguins, lesdosagesn’ontpaspuêtreréalisésaussi régulièrementquerecommandé.Les

adaptationsdeposologieontétéfaitesselonlesrésultatsdesdosagesdeciclosporinémie.

Le second immunosuppresseur utilisé pour remplacer la Ciclosporine est le Tacrolimus

(fujimycine). Son principal avantage est sa facilité d’utilisation (et son coût inférieur): un

cycle avec une prise quotidienne per os pendant 5 jours suivi d’une fenêtre de 48h sans

immunosuppresseur. La surveillanceétait assuréepar lamesurede la tacroliémie totaleà

24h(cible15-20ng/ml)(Figure3).IlavaitcommeavantagesparrapportàlaCiclosporinesa

surveillanceplussimple,undosagedanslesangquasi-constant,etl’absenced’effetdélétère

sur la cicatrisation. C’est d’ailleurs l’immunosuppresseur de référence pour les patients

transplantésrénaux.

Figure5:schémaillustrantletauxsanguintotaldeTacrolimusdansles24heuresqui

suiventsoningestion.Lacourbebleuemetenévidencel’accumulationduTacrolimus

au5èmejouraprès5joursdeprisequotidienne.

21

5.1.c.Surveillanceenimageriedudéveloppementdestumeurs

Trois tomodensitométries (TDM) ont été réaliséesen apnée sous anesthésie générale

(scanner Philips Tomoscan N monobarrette ou GE Discover 750 HD) : la première en

contraste spontané avant l’implantation et immédiatement après pour vérifier l’absence

d’anomaliemorphologiquerénaleoudecomplication.LesdeuxTDMdesurveillanceontété

réalisées plusieurs semaines après l’implantation et immédiatement avant le sacrifice des

porcsetl’examenpathologiquedesreins,sansetaprèsinjection(1,5cc/kgparinjection)àla

phasecorticaleetnéphrographiquepourétudierlaforme,levolumeetlerehaussementdes

zones d’implantation et s’assurer de l’absence de complication locorégionale. Un

rehaussement significatif (>20UH) laissait présager la viabilité de la zone d’implantation,

d’éliminersanécroseetd’espérerlaprisedel’allogreffedescellulestumorales.

5.1.d.Résultats

Les 8 premiers porcs ont bénéficié de la Ciclosporine, les 12 derniers du Tacrolimus. Les

xénogreffes ont été réalisées avec des tumeurs d’histologies habituelles. Une tumeur

d’origineurothélialeaégalementété implantéepourcomparersonagressivitéparrapport

auxtumeursparenchymateuses.

Iln’yapaseudedécèsprécoceenrapportavecl’implantationnil’anesthésiemais2porcs

décédés avant la date prévue d’explantation (entre 12 et 20 semaines): ils sont décédés

sansétiologie retrouvéemais avecunealtérationbrutalede leurétat général. Leurs reins

ontaussiétéexplantéspouranalysepathologique.

LesTDMontpermisdedistinguer2typesdezonesd’implantation:

-descicatricessous la formededéfectderehaussementpar rapportauparenchymerénal

sain,sansrehaussementsignificatif(Figure6a);

-deszones±densesavantinjection,quiserehaussaientsignificativement(>20UH)(Figure6betc).

Figure6:exempledezoned’implantationsurlesscannersdesurveillance.

a. cicatrice avec infarctus – b et c. la même zone d’implantation sans et aprèsinjection,spontanémentdensemaisquiserehausse(25UH).

22

Un pathologiste avait analysé les 80 zones d’implantation tumoraleet décrit

systématiquementdesremaniementsdefibrose,ycomprislorsqu’onsuspectaituneprisede

lagreffe tumoralemacroscopiquement (syndromedemasse)ousur lesscanners (prisede

contraste).Ilexistaitdefaçoninconstantedesremaniementsinflammatoiresoudenécrose

adjacents à cette fibrose (Figure7).Aucune cellule tumoralehumaineni néo-angiogénèse

n’avaitétévisualiséesurleslames(Figure8).

Figure7:aspectmacroscopiquedesreinsexplantés.

Figure8:réactioninflammatoiresurlesited’implantationdel’agrégattumoral

5.1.e.Discussion

Unedesprincipaleslimitesdenotreétudeestladuréedesurveillancerelativementcourte

(≤6mois)comptetenudelavitessedecroissancedescancersdurein(<1cm/an)[37].Dans

la mesure où nous n’avons pas mis en évidence de rehaussement au scanner ni de

vascularisationdes zones implantées, il paraît peuprobableque les agrégats tumoraux se

soientimplantésetqu’onaitpuavoirunecroissancedetaillesurdesscannersplustardifs.

La seconde critique concerne l’effectif. Même si environ 80 implantations ont pu être

réaliséessur20porcs,nousn’avonspaspudéfinirunprotocolerigoureux:lesécueilsliésau

choix de la technique d’implantation, à l’immunosuppression, ont limité le nombre

d’implantationdansdesconditionsoptimales.

23

Nousn’avonspas réussi à implanterde tumeur rénalehumaineauporc,mais ces travaux

nousontpermisdeprogresserdansplusieursdomaines:

-l’immunosuppression,

-l’implantationdematérielexogènedanslereinparvoietranscutanée,

-le développement demodèle de tumeurs humaines chez l’animal pour d’autres organes

quelerein.

• L’immunosuppression:

Ellenousa semblé indispensable: l’analysehistologiqueamontréque toutes les tumeurs

implantées étaient lysées, avec une réaction immunitaire qui entrainait une infiltration

lymphocytaire, des phénomènes de fibrose et de nécrose. Cette inflammation non

spécifique correspondauxmécanismesphysiologiquesde l’inflammation faisant intervenir

descytokinespro-fibrosantes.Cetteréactiondefibroseestmiseenévidencequellequesoit

lanaturehistologiquedelatumeurprimitiveetl’étatd’immunitédel’animal.

AveclaCiclosporine,nousavonsrencontrédenombreusesdifficultés(2prisesquotidiennes,

adaptation de la dose pour obtenir le taux sérique cible, difficultés de cicatrisation…). Le

Tacrolimusaétéd’utilisationbeaucoupfacile,puisqu’avecuneprisequotidiennependant5

jours, nous avons obtenu des taux sanguins dans la fourchette cible pendant la semaine.

Nous n’avons plus eu de problème de cicatrisationmais cette notion est à prendre avec

précaution, puisque le changement de technique d’implantation (uniquement par voie

radiologiqueaulieud’unealternance)coïncidaitaveclechangementd’immunosuppresseur.

• L’implantationdumatérieldanslerein:

L’intérêt de l’abord chirurgical était d’implanter les plus volumineux fragments tumoraux,

alorsquepourl’injectionpercutanéedelatumeurfragmentée,nousavonsutilisélecoaxial

15Ga utilisé pour les biopsies rénales médicales percutanée. L’avantage de l’abord

percutanéétait lacicatrisationdupointd’abord,alorsqueplusieurscicatriceschirurgicales

se sont surinfectées, ce qui a nécessité des soins locaux et une antibiothérapie IV. Les

problèmesdecicatrisationontétémissurlecomptedel’immunosuppression.

Parailleurs,lorsdespremièresprocédures,nousavonsimplantédesagglomératscellulaires

en sous capsulaire et en profondeur à la jonction cortico-médullaire: le matériel en

profondeur n’a pas été visualisé ni sur les scanners intermédiaires, ni sur les lames

histologiques. Il n’y avait pas ou très peu de réaction de fibrose de ces sites. Les 3

hypothèsesquenousavonsévoquéessont:lepassagedumatérieldansl’appareilurinaire,

sadissémination(parvoiehématogène),uneréactionderejetde l’organismeplus intense

qu’enpériphériedurein?

24

5.1.f.Quelquessuccès…etdesperspectives

Ces expérimentations s’intégraient dans l’implantation sur le porc de tumeurs humaines

danslerein,maisaussilefoie,lepancréasetlaprostate.Lesrésultatsn’ontpasencoreété

publiésmais,chezcertainsanimaux,destumeurshépatiques,pancréatiquesetprostatiques

ontétéimplantéesavecsuccès.Cequiprouvequenousmaitrisonsl’immunosuppressionet

laxénogreffe.

Endébutd’annéeprochaine,unesériedenouvellesexpérimentationsvadébuter:

• les tumeurs humaines seront toujours prélevées sur les pièces chirurgicales de

néphrectomieenperopératoire.Commecespatientsaurontbénéficiéd’unebiopsie,un

échantillonnagedesdifférenteshistologies serautilisépour les implantations, etnous

utiliseronsdifférentsgrades(pourprivilégierlestumeursagressives);

• les tumeurs seront implantées en percutané sous guidage radiologique (scanner +

échographie)danslecortexrénaldeporcs;

• lesnouvellesaiguillesd’injectionontétédéveloppéespour lachirurgie réparatrice (Pr

Magalon –Thiebaud©): elles sont utilisées pour l’injection autologue de graisse (des

patients séropositifs sous thérapie ayant des lipodystrophies). Avec ces aiguilles,

l’implantationdegraisseestoptimisée,leslobulesdegraissenesontpasdétruits;

• le protocole d’immunosuppressionpar Tacrolimus sera celui utilisé pour les dernières

expérimentationsquenousmaitrisons;

• si l’étatgénéraldesanimaux lepermet, ilsserontsurveillésen imagerieparscanner±

échographieà1,3moisetavantd’êtresacrifiésà6moisetà1an;

• l’examenanatomopathologiqueseraassuréparl’équipeduPrL.Daniel.

En créant des cibles rénales, nous pourrions valider la précision de notre algorithme de

segmentation, valider la précision des HIFU ou des autres agents externes délivrés sur le

rein:ledéveloppementdecetteciblenoussembletoujoursd’actualité.

25

5.2.Modèleinertedetumeurrénaleanimale(travauxréalisésauCRCHUM)

Parcequenoussavionsqueledéveloppementd’unmodèlevivantdetumeurrénaleseraitcompliqué,nousavonsenparallèletravaillélamiseaupointd’unmodèleinertedetumeur.un modèle «vivant» de tumeur rénale aurait l’avantage de permettre d’analyser lesremaniements induits par exemple par les HIFU , un modèle inerte serait suffisant pourvaliderletrackingd’uneciblerénale.

L’objectif d’un telmodèle aurait été: une facilité de développement pour une utilisationimmédiatesansattendreledéveloppementd’unetumeur.Puisquel’objectifestdesuivreentempsréellesmouvementsdureinenIRM,ildoitêtrevisibleenIRM.Iln’yapasdemodèlesimilairedécritdanslalittérature.

Avant la revue descriptive des techniques utilisées sur des porcs anesthésiés, nous avonsdéfinilesqualitésquedevraitavoircettecible:

-positiondanslerein:ellenedoitpassedéplacerdanslereindansletemps;

-taille: lesmassesrénalesaccessiblesàuntraitementdethermo-ablationmesurentmoinsde4cm,cequiconstituela limitesupérieurequel’onsefixe.Commenotreobjectifestdetraiterdesmassesrénalesavecuneprécisionde5mm,valeurdenotre l’approximationdenotrealgorithmedesegmentation[38],lacibledoitmesureraumoins1cmdediamètre.

5.2.a.matériel

-matérield’embolisation:parvoieendovasculaire,différentsagentsontétédéposésenavaldel’artèrepréourétro-pyélique:coïlsenplatine(Figure9),enacier(Figure10),Curaspon,Chitosan…Ilsavaienttousétélarguésetnes’étaientpasmobiliséssurlescontrôles.

-injectionpercutanéed’unmarqueur:sousguidageéchographiqueetfluoroscopique,nousavons injectédans lecortexoudans lamédullairedureinunmélangedeCurasponoudeChitosanetdegadoliniumquiformaitunemassesurlecontrôleimmédiat.

-créationd’unecicatrice:desinfarctusparembolisationsélectiveontétéinduits:leproduitd’embolisationinjectéformaitune1ère ciblemais leseffetssecondairesvasculaires induitsvisiblesenIRM(Figure11).Lacicatriceétaitlacible,paslematérielembolisé(1erstests).

Figure9:coïlenplatineembolisédans lereindroitd’unporc,nonvisiblesur l’IRM(T1DIXONopp).

26

Figure10:coïlsenacier(stainlesssteel)visiblessurlesimagesd’angiographies,lesreconstructionstomodensitométriquesetlesséquencesIRM(T1DIXONoppetFLASHtrufiFreeBreathing)oùellesartefactentl’examen(«lescibles»rénalesnesontpasexploitables).

Figure 11: embolisation de segments rénaux par du Curaspon mélangé à deschélates de gadolinium. La zone embolisée est visible en hypersignal sur la lèvremédialedureingauchesurl’imageaxialeT1maisn’estplusvisibleaprès30minutessurlesséquencescoronalesFLASH.

5.2.b.concentrationdeschélatesdegadolinium

Pourlesciblesquin’avaientpasdesignaldistinctifauparenchymerénaladjacent(Curaspon,Chitosan), nous les avonsmélangéesavecdes chélatesdegadolinium.Pourdéterminer laconcentrationdegadoliniumnécessairepourquela«cible»soitvisibleenIRM,nousavons

27

testé différentes concentrations de gadolinium avec les séquences IRM définies pour leprotocole KiTT de description des mouvements du rein. Notre référence était laconcentration utilisée pour les arthro-IRM puisque l’injection associe du produitanesthésique,ducontrasteàbased’iode(pourl’injectionsousscopie)etducontrasteàbasede chélate de gadolinium. Si on ne dispose pas des produits spécifiques (ARTIREM), lemélange associe une dilution 1:200 pour obtenir une concentration de 2.5mM enmélangeant0.1ccdegadoliniumavec10ccdecontrastenonioniqueet10ccd’anesthésique(quenousavonsremplacépardusérumphysiologique).

Sur ces exemples (Figure 12), on peut voir la modification du signal des échantillons enfonction de la concentration en chélate de gadolinium. L’aspect hétérogène est du auChitosan,quiestunsupportdegélatine.

L’équipe du Pr Sophie Lerouge (Laboratoire de biomatériaux endovasculaires (LBeV),CRCHUM) a assuré le mélange du Chitosan avec les concentrations de gadoliniumdéterminées.Pourlesembolisationsdesporcs,nousavionschoisi2concentrationsà0.1et1%.

Figure12:échantillonsdemélangedeChitosanetdechélatesdegadoliniumà0.01,0.1,1et10%(etduNaCL0.9%)surunecoupeIRM(FLASH).

5.2.c.résultatsdesprocédures

Sur les contrôlesper-procédure, tous les«dispositifs»ontété implantésavec succès. Lesrésultatsn’ontpasétéconvaincants(Figure13):

-lesinfarctuscréésn’étaientpasassezvisiblessurlesIRMdecontrôle(avecnosséquences)réaliséesdanslesheuresquisuivaient.

-lescoïlsn’étaientsoitpasvisibles(troppetit),soitengendraienttropd’artefactspourqu’ilspuissentêtredélimitéspourdéfinirunecible.

-lesmatérielspositionnésenpercutanéonttousdisparus,ennelaissantquasimentaucunetrace.Ont-ilsétémétabolisés?Sont-ilspartisdanslacirculationvasculaireoudanslesvoiesurinaires?

28

Figure13:lepôleinférieurdureinaétéemboliséparunmélangedeChitosanetdechélatede gadolinium [1%] commeon le voit sur l’angiographie.Uneheure après,l’IRM(T1DIXON)étaitnormale,leterritoireembolisén’étaitpasdifférenciédurestedureinetleproduitdecontrasten’étaitplusvisualisé.

5.2.d.conclusiondumodèleanimaldetumeurrénaleinerte

Aucune des techniquesmises enœuvre n’a permis de remplir les objectifs de notre cible(ciblefixe,bienlimitée,stabledansletemps).

Soit le matériel n’est pas visible (immédiatement après son positionnement, ou aprèsquelquesminutes),soitilengendretropd’artefactpourdéterminerunecible.

Lechoixdepositionnerunballonparvoieendovasculaireaétéécarté:lesballonslargablesnesontpasaccessibles(rareté),etiln’estpaspossibledelaisserunballonenplaceavecsasondeporteuse.

Pourtesteretvaliderlesuivid’uneciblerénallorsdelarespiration,nousnepouvons,pourl’instant,quelasimulerinformatiquement.

29

6.Modélisationdynamiquedureinparuneapprochedetypemorphing

[ConférencesA-F,Annexe7]

Latechniquededescriptiondesmouvementsdureinestdéveloppéedanslapartie7mais,

nous avons exploré une autre voie avant d’utiliser la segmentation du rein à différents

momentsducyclerespiratoireavecuneapprochedetypemorphing.Elledevaitpermettre

unemodélisationdynamique.

CestravauxontétéréalisésauLIIEencollaborationavecV.Leonardide l’équipeduLSIS–

UMRCNRS7296(Aix-MarseilleUniversité)dirigéeparM.DanieletJLMari:j’aipuinitierle

projet en tant que radiologue référent (notamment pour le protocole d’acquisitions des

donnéestomodensitométriques)avantmonannéedemobilitéauCanada.

Quatreétapesontéténécessairespourcetteapproche:

-d’abord,uneméthodologiedesegmentationdu reinqui s’appuyait surunecroissancederégionsemi-automatique.

-ensuite une reconstruction 3D du rein qui reposait sur la résolution d’une équation dePoissonpourobtenirunmodèlestatiquedurein.

-puisl’implémentationd’uneméthodedemeshmorphing(transformationprogressived’unmaillageàunautre)pourpermettreuneapprochedemodélisationdynamique.

-enfin, l’adjonction de contraintes de courbures discrètes pour améliorer la modélisationdynamique.

Lesimageriesétaientréaliséesenscanner.

6.1.Extractiondescontoursdurein[A,B]

Laméthodedesegmentationsemi-automatiquequiextraitlescontoursdureins’appuiesuruneapprocheparcroissancederégioninitialiséemanuellementparunegrainesurlacouchemédiane.Lesapproximationssontéliminéesparl’analysedeshistogrammesdesrégions.Lesprincipalesdifficultéssontduesà la juxtapositiond’organesdedensitésvoisinesàcelledurein.

La principale lacune de cetteméthode est l’utilisation exclusive des niveaux de gris pourextraire lescontours.Plutôtqu’analyserdeshistogrammesdeniveauxdegris,uneanalysedetextureseraitpluspertinentepuisqu’elleseraitplusprécisepourdétecter les interfacesentrelesorganesetqu’ellepermettraitdesegmenterunetumeurduparenchymerénal.

30

6.2.Constructiond’unmodèle3Dstatiquedurein:lareconstructiondesurfacedePoisson

[A,B]

Lasurfaceàreconstruireestexpriméecommeunesolutiond’uneéquationdePoissoncequipermet de basculer du domaine de la géométrie à celui de l’analyse vectorielle où larésolution de ce type d’équation est connu. Cette technique rapide et entièrementautomatique (si «l’octree» est correctement déterminé en avant) permet de s’affranchirdeséventuelleserreursdesegmentation.

6.3.Modélisationdynamiquedureinetsuividetumeur[Annexe7,C]

L’originalité de notre méthode est qu’elle repose sur une approche entièrementgéométrique : le mesh morphing. Elle est rapide et ne nécessite que trois modèlescorrespondantsàtroisphasesrespiratoires:laphased’inspiration,laphased’expirationetunephase intermédiaire. Les résultatsobtenussontaussidenaturegéométriquepuisqu’ils’agitd’unmodèle3Danimé:lemouvementetlesdéformationspeuventêtreétudiéssousn’importe quel angle alors que certaines méthodes n’offrent que la possibilité d’unevisualisation2D.Silatumeurestexophytiqueoudéformelescontoursdurein,commeelleestaussianimée,ilestpossibledeconnaîtresapositionàtoutmoment.

Maiscette techniqueestperformantepourévaluer levolumedu reindanssonensemble,sansréelledistinctiondelatumeur.Ledeuxièmefacteurlimitantdecettesolutiontechniqueestladuréedecalcul:40secondessontnécessairespourgénérerle«méta-maillage»pournotremodèle.Cesimprécisionssontunfreinpourleciblaged’unecible.

6.4.Morphingcontraintparcourburesdiscrètes[D,E]

Danslecadredemasseexophytique,onpeutaffinerlaprécisionderepérageenétudiantles«courburesdesexcroissances»del’envelopperénale.

Cestechniquessontdécritesdans[D,E]maisonpeutregretterleurslimites:

-les tumeurs ne sont pas détectées, si elles ne sont pas exophytiques ou corticales

périphériques,siellessontdansunerégionconcave,

-larésolutiondumaillage(desimages)doitêtredequalitésuffisantepourlimiter letemps

decalcul.

31

6.5.Conclusion

Lescontraintestechniquesmisesenévidencenepermettentpasd’envisageruntrackingdureinetsurtoutd’unemasserénaledefaçonfiableparmorphing.

Finalement,leprincipedel’approchedesmouvementsdureinparmodélisationdynamiqueest la focalisation sur la déformation de la cible pour étudier les variations d’aires, devolumesoudecourburesà lasurface.Nousavonsbesoind’unsuividureinquisefocaliseplutôtsurlapositiondel’objetdansl’espaceetsesdéplacements.

Nousnoussommesdoncorientéssuruneapprochede recalaged’unvolume3Dprécisetrobuste(partie7)àpartird’acquisitionsrapides2D:cestravauxsontréalisésdanslecadred’une collaboration internationale LIIE - Aix-MarseilleUniversité / Centre de Recherche duCHU de Montréal – Laboratoire de Recherche en Imagerie et Orthopédie et École deTechnologieSupérieuredel’UniversitédeMontréal.

Pourobtenir lemeilleurcompromis résolutionspatiale– résolution temporelle,nousnoussommes orientés vers des acquisitions en IRM qui ont les avantages de pouvoir êtrerépétées (dans le cadre d’un traitement long) et de permettre un monitoring de latempératuredelazonetraitée(thermographieper-IRM).

32

7.DescriptiondesmouvementsdureinenIRM[V]

Lereinsedéplacelorsdesmouvementsrespiratoires.Cesdéplacementsontdéjàfaitl’objetdedescriptionsmaisilsn’ontjamaisétéétudiésenIRM.

L’intérêtdecetteanalyse[V:SouteyrandP;ChungW;DeGuiseJ;MariJL;LeonardiV;OliviéD; Vidal V; Soulez G. A MRI Description of Kidney Motion. Soumis Journal of MagneticResonance Imaging]étaitde corréler lesmouvementsde translationdans les3plans voiravecunerotationouuneéventuelledéformationrénale.Doit-ontenircomptedetouteslescomposantesdecesmouvements?

LesaspectstechniquessontdétaillésdansV.

7.1.Matérielsetméthodes(travauxréalisésauCRCHUM)

DesIRMontétéréaliséesenapnéeà5momentsducyclerespiratoirechezdesvolontairessainspourobtenirunvolumeen3Ddechaquereinselonleprotocole:

! 5phasesducyclerespiratoiredifférentes(Figure14)-10volontairessains

! 3D T1 VIBE sequencing DIXON (TR 4.34 / TE 1.35 (opposed phase), FOV 308*380,Matrix195*320,Slicethickness3mm)

! Skyra3TSiemensMedical–CRCHUMMontréalCanada

Figure14:enregistrementducyclerespiratoireàpartird’unesangleabdominale.Lecontrôledelarespirationestdoncprospectifetrétrospectif.

33

7.1.a.l’algorithmedesegmentationduvolumerénal

Lesvolumessontsegmentésparunalgorithmedéveloppéaulaboratoire[39]:initialementutilisépoursegmenterlefoiesurdesscannersavecinjectiondeproduitdecontraste,ilaétéadaptépourl’IRMetlerein[38].LesséquencesIRMontétédéveloppéespourpermettrelameilleuresegmentationpossibleavecladuréecommecontrainte:lesacquisitionsétantenapnée,ellesdevaientdurermoinsde10secondes.

Ils’agitd’unesegmentationsemi-automatiquequipeutêtrecorrigéeencasd’erreuretquipermetd’obtenirunmaillagesurfacique3D(techniquedéveloppéedanslapartie7.1.a[39]etlaFigure15).

Figure15:exempledesegmentationdurein-a:àpartirdelaséquenceVIBE3DT1DIXON(opp),obtentionduvolumerénalen3Dquiestcontourédans2plans(lignesrouges).-b:levolumeMESHextrapolé.

Manuellementetdans2plans, l’opérateurmarquent lacapsuledureinavec3ou4pointspuis le logiciel génère un volume par interpolation variationnelle. Ces courbes dans deuxplans servent de contraintes. Les points reliant le contour sont automatiquement liés parunesplineCatmull-Rom.Ladeuxièmeétapeestlasegmentationproprementditequiconsisteàidentifierlafrontièredel’organe.Elleestcomposéedetroisphases:-uneoptimisationdumaillageLaplaciennepermetlarelocationdespointsquinesontpasàleurplace,-un modèle d’appariement identifie les points cibles correspondant à la délimitation del’organe,-lemodèlegéométriqueestdéforméitérativementparoptimisationLaplaciennejusqu’àcequ’ilconvergeverslafrontièredel’organe.A chaque étape, l’opérateur peut corrigermanuellement les erreurs de segmentation. LemaillageMESHduvolumegénéréestexportéauformat.vkt.

34

7.1.b.étudedesmouvementsdurein(Figure16)

-translation:

àpartirdechaquevolume,lebarycentreétaitdéterminéavecl’équation:

où!représentelapositiondubarycentred’unmaillage3D;Nreprésentelenombredesommetsdanslemaillage3D;Creprésentelapositiond’unsommetdanslemaillage3D.

Latranslationdesbarycentresentre2points1et2étaitétudiéedansles3plans:T=C1–C2

-rotation:un systèmede référence (axesprincipauxdumodèle surfacique)était insérédans chaquemodèledereinsdans lesdifférentesséquencespourextraire lesrotationsparrapportauxaxesdessystèmesderéférence.L’originedeceréférentielplacéaubarycentredumodèlesurfacique et l’orientation des axes étaient positionnées en fonction des longueurscaractéristiquesàpartirdel’originedechaquesommetavecl’équation:

oùAreprésentelesaxesprincipauxdumaillage;Nlenombretotaldesommetsdumaillage;mlamassed’unsommetdumaillage;rladistancequisépareunsommetaubarycentredumaillage.

Àpartirdel’algèbrematricielle,ilétaitpossiblededéterminerlamatricederotation[R]quipermetdepasserd’unsystèmederéférenceàunautreavecl’équation:

où [AE] représente le système de référence placé dans la séquence en expirationmaximale;[R]lamatricederotationpourpasserde[A_I]à[A_E];[AI]lesystèmederéférenceplacédanslaséquenceeninspirationmaximale.

35

Figure16:décompositiondesmouvementsdureinentranslationetenrotation

7.1.c.étudedesdéformationsdurein

Lesdéformations(volumetricoverlaperrorVOE)entrel’inspirationI(deepinspirationmesh)etl’expirationE(deepexpirationmesh)maximalessontanalyséesparlaformule[40]:

⏐E∩I⏐

VOE(E,I)=1–⎯⎯⎯×100%

⏐E∪I⏐

Une VOE de 0% correspond à un chevauchement parfait entre les segmentations alorsqu’uneVOEde100%correspondàuneabsencecomplètedeleurchevauchement.

36

7.2.Résultatsetdiscussion

Ledétaildesrésultatsesttranscritdansl’articleenannexe,etdanslesfigures17et18.

Figures17et18:résultatsdesmesuresdetranslationetderotationdesreinsdansles3planschez10volontairessains

Lesmouvementsdureinsontlasynthèsedetranslationsdansles3plansdel’espace,avecunetranslationprépondérantedansleplancoronal,etd’unerotation.

Il n’a pas été réalisé d’étude statistique pour modéliser les mouvements en fonction decaractéristiquesmorphologiques ou démographiques comme le sexe, l’âge, le BMI … Destendances se dégagent: les déplacements cranio-caudaux sont plus amples chez les plusgrandspatientsalorsqu’aucunetendancenesedégageconcernantlarotation.

Commedécritdans l’article, lereindoitêtreconsidérécommeunobjetrigide,c’estàdirequinesedéformepas.Comptetenudelaprécisiondenotrealgorithmedesegmentation,nousn’avonspuenapporterlapreuvestatistiquemaislalittératurevadanscesens.

Ilestdoncnécessairepoursuivreen temps réeluneciblesurun reinaucoursd’uncyclerespiratoiredeprendreencomptelacomplexitédesesdéplacementsdansles3plansainsiqued’unepartde rotation,etpas seulementde la composantemajoritairede translationcranio-caudale.

37

8.Trackingdurein

ArticlerédigéconjointementavecWendyChung,étudianteenMasterauCRCHUM,encours

derelectureavantsoumissionàTransactionsBiomedicalEngineering[VI].

Parce qu’ils associent translations et rotations, les mouvements du rein sont complexes.

Nousavonsdéveloppéunalgorithmequirecalelevolumerénal3Dacquisenprétraitement

surdesacquisitionsen2Drapides(FLASH).

Ilestencoursdevalidationsurlesdonnéesdeporcsanesthésiésetventiléspourcontrôler

leurrespiration.

8.1.Simulations

Troisparamètresontétéétudiés:

-leplanoptimaldesacquisitions2Dpourlerecalage?

-l’algorithmepeut-ilcorrigerunmouvementdupatient(autrequelarespiration)?

-peut-on «dégrader» la qualité de nos images FLASH sans compromettre le

recalage(l’objectifétantdediminuerleurtempsd’acquisition)?

L’algorithme utilisait la fonction mérite par la corrélation croisée (CC) pour effectuer un

recalageparintensitéencombinaisonavecunalgorithmed’optimisationdeSimplex.

Lafonctionméritedecorrélationcroiséeaétécomparéeàlafonctionmérited’information

mutuelle (IM)pour s’assurerqu’elle était plusperformante: ces 2 fonctions sontutilisées

danslerecalaged’images[41-43].

8.1.a. première simulation: quel plan pour l’acquisition 2D? Un déplacement du patient

peut-ilêtrecompensé?

Danslamesureoùlerecalagesefaitparintensité,etquelesintensitésdescontourssontles

mêmesdansles3plans,lechoixduplannedevraitpasinfluencerlaqualitédurecalage.

Apartirduvolumedureinestextraitune image2Dquiest rognéepournegarderque le

rein: c’est «l’imagette» qui servira de référence. Des imagettes dans les 3 plans sont

sélectionnées.

L’étaped’initialisation consiste à positionner l’imagette où l’erreur de recalage est la plus

faibledans le volume3D. L’algorithmepar les fonctionsdemériteCCou IMdétermine la

positionoptimaledel’imagetteenlatranslatantenx,yet/ouz.

Puisl’algorithmeappliquelafonctiond’optimisationdeSimplexeninclinantl’imagettepour

affiner lacorrélation.Poursimulerunautredéplacementque la translationphysiologique,

unerotationduvolumeestappliquéepourtesterlacorrélationdurecalage.

38

8.1.b.deuxièmesimulation:peut-ongagnerdutempsd’acquisitionendégradantlesimages2D?

LesséquencesFLASHquenousutilisonspermettenten0,9sd’obtenir2plansdecoupeavecune résolution spatiale suffisante pour contourer le rein. En diminuant le tempsd’acquisition,estcequeladiminutiondelarésolutionn’estpastropimportantepouraltérerlerecalageduvolume3D?

Une pyramide Laplacienne est appliquée sur les imagettes: ce filtre retire les hautesfréquencesdel’image,lescaractéristiquesprincipalessontatténuées[44](Figure11).

8.1.c.troisièmesimulation:peut-onrecalerlevolumeàpartirdesimagesFLASH2D?

Uneséquence3DetuneséquenceFLASHdansdifférentsplansontétéréaliséespendantlamêmeapnée:lerecalageaétévalidésurcessériesenapnée(Figure12).

8.2.Résultats

Le résultat du recalage était analysé par la méthode du damier: l’image issue de lasuperpositiondesdeuximages(l’imagettederéférenceetcelledéterminéedanslevolumeparl’algorithme)étaitdiviséeenplusieurscarreauxidentiques(Figures19,20et21)[45].Lafigure22résumelesrésultatsdecorrélation.

Figure19:Exemplederecalageetd’analysedesrésultats:lesindicesdecorrélationétaientmesurésà0,99àgauchepourlafonctiondeCCetà0,66pourlafonctionIM(pourdestempsdetraitementparl’algorithmede11,39et12,07srespectivement).

39

Figure 20: recalage 2D/3D avec la CC et la fonction d’optimisation Simplex desimagettessous-échantillonnéesavecunepyramideLaplacienne.

Figure 21: recalage 2D/3D avec la CC et la fonction d’optimisation Simplex d’uneimageflashprovenantd’unvolontaireetdesonacquisition3D.Lesdivergentessontentouréesenrose.

Figure22:lesrésultatsdecorrélationsontrapportésenmodifiantlesparamètresduplande l’imagette eten intégrantune rotationau rein. Le tempsde traitementdel’imageestassociéaucoefficientdecorrélation.

40

Nosrésultatssontenaccordaveclalittérature:lafonctiondeméritedecorrélationcroiséeassociéeàl’optimisationdeSimplexestsupérieureàcelled’informationmutuelle[46]danstouteslesconfigurations:elleestperformantequelquesoitleplandel’image2D[42],avecuneimage2Ddégradée[45]ouencasdemouvementsdupatient.Ellepermetunrecalagefiableduvolumerénal3DàpartirdesimagesFLASH2Dd’unvolontairesain.

8.3.ConclusionEn combinant la fonction demérite de corrélation croisée avec la fonction d’optimisationSimplexdansMatlab,nouspouvonsrecalerprécisémentlevolumedureinàpartird’imageen2D.Laprochaineétape consiste à accélérer ce recalage (qui durequelques secondes) tout endiminuantencoreletempsd’acquisitiondesséquences2D:-le rein a été segmenté dans sa totalité. Peut-on ne segmenter qu’une partie du rein quiinclue lacible?Celapourraitnouspermettredediminuernotrechampd’explorationpourgagnerdutempsdetraitementdesdonnées.-tous les calculs ont été réalisés sur des ordinateurs de bureau: l’implémentation desalgorithmessurdesordinateursdédiésàcescalculsdoitpermettredediminuerletempsdetraitement. D’autres voies sont en cours de développement pour optimiser le suivi d’unetumeurrénaleentempsréel.Notreprocessusderecalage2D/3Destfiableàplusde90%danslaconfigurationactuelle(qualitéd’image,duréedetraitementdesdonnées…).EncombinantlesdonnéesFLASHen2D(parpaires)avecunmodèle3Ddureinobtenuenamontprécédemment,l’objectifserade calculerunmaillage4Ddu rein, avec contrôleetestimationde l’erreuretde ladérive(drift): on obtiendrait la construction d’un modèle géométrique rénal 4D par fusion decoupesIRMflash.CestravauxsontencoursavecNilsOlofsson,étudiantenthèseencotutelleETSMontréal/

Aix-MarseilleUniversité.

41

9.Conclusion-perspectives

La prise en charge thérapeutique des tumeurs rénales a considérablement évolué ces

dernièresannéesavec l’avènementdetraitementsmini-invasifs (commelaradiofréquence

percutanée)quioptimisentl’épargnenéphronique,amélioreleconfortdupatientavecune

efficacitéoncologiquecomparableauxtraitementschirurgicauxderéférence.

La prochaine étape sera de proposer des traitements transcutanés (HIFU, radiothérapie

stéréotaxique…)aussi performantsavecunemorbi-mortalitéoptimisée.Celapassepar la

miseaupointdetechniquededétectionentempsréeldelatumeurrénale.

Ces travaux nous ont permis de développer une algorithme de repérage fiable qui doit

encoreêtreoptimisé(rapiditédecalcul)etêtrevalidésurunmodèle(animal?virtuel?)qui

n’estpasencoredisponible.Lestravauxd’optimisationdesalgorithmesdesegmentation,de

fonctiondeméritedecorrélationcroiséeassociéeàlafonctiond’optimisationSimplex,sont

en cours dans le cadre d’une collaboration internationale franco-canadienne au LIIE et au

LIO.Enfonctiondel’avancéedecestravaux,jedevraimerendreauCRCHUMauprintemps

pourtesteretvaliderlesalgorithmesavecdesséquencesIRMoptimisées.

LeprojetKiTTaévoluépuisquesonobjectifactuelestdedélivrerunagentdedestruction

tissulaire sur une cible rénale avec une précision de ± 2.5mm. Dès que l’algorithme

permettradedéterminerprécisément lapositionde la tumeur (enpratiquesoncentre), il

faudra pouvoir communiquer ses coordonnées pour appliquer l’agent physique de

destruction.

LesHIFUavaientnosfaveurscommeagentdestructeur:plusieurséquipeslesutilisentpour

le traitement de fibromes utérins, de tumeurs osseuses … des cibles immobiles. Pour le

traitementdesmassesrénales, il fautdisposerd’uneméthodedetrackingdecettemasse,

déterminer les paramètres d’application des HIFU (durée, impulsions et niveaux de

puissance pour détruire la région ciblée) avant d’envisager des essais contrôlés

multicentriques.AveclenouvelHôpitalSaintLucdeMontréaletleCentredeRecherchedu

CHU deMontréal, nous disposons de plates-formes IRM de dernière génération. Dans le

cadredudéveloppementdes traitementsde radiologie interventionnelle, il aétéenvisagé

d’acquérir la technologie des HIFU. Une IRM va être installée à CERIMED en 2016: cette

plateformepeutêtreassociéeauxHIFU...

42

La radiothérapie conventionnelle n’était pas considérée comme un traitement curatif du

cancer du rein. Avec le développement de la radiothérapie stéréotaxique, on pourrait

imaginerqueladélivranced’unedosesignificativedegraysuruneciblerénalepuisseêtre

efficace.Ledépartementderecherchederadiologiedel’UniversitédeMontréaldirigéparle

Pr G. Soulez regroupe la radiologie et la radio-oncologie: une collaboration pourrait

s’envisagerpourassociertrackingdureinetradiothérapiestéréotaxique?

GrâceàlacréationduDépartementHospitalo-UniversitaireduPôled’ImagerieMédicalede

l’AP-HM, nous préparons un projet d’offre Recherche Hospitalo-Universitaire en santé

(RHU): le projet KiTT pourrait rentrer dans ce cadre «de recherche avec un potentiel de

transfert rapide vers l'industrie ou vers la société (…) de plateformes technologiques (…)

privilégiant lesnouvellesmodalitésdepriseen charge thérapeutique, les traitementsplus

efficacesetmieuxtolérés».

UnedemandedesubventionauRBIQpour leprogrammeRéseautage internationalesten

cours pour permettre de financer les échanges dans le cadre de notre collaboration

universitaireMarseille/Montréal.

Début2016,unenouvelleséried’implantationsdetumeurrénalehumainechez leporcva

débuter avec des protocoles optimisés (immunosuppression, préparation et injection de

l’agrégat):pourrat’ondisposerdecemodèledetumeuren2016ou2017pourvalidernos

algorithmesetenvisagerdestraitements«non-invasifs»desmassesrénales?Nousgardons

lapossibilitédemodéliserinformatiquementuneciblerénale.

Tous ces travaux rentrent dans le cadre du projet de développement de l’imagerie

interventionnelleenuro-radiologieetdanslesautresspécialitésauseinduPôled’Imagerie

de l’Assistance Publique des Hôpitaux de Marseille. L’embolisation de prostate, les

procédures (radiofréquence, cryoablation, micro-ondes, HIFU…) sous guidage scanner et

IRM(dontuneIRMàchampouvert,desIRMdehautchamp3ToudesIRMàlargetunnel)

doiventnouspermettrederesterenpointeenuro-radiologieinterventionnelle.

43

Bibliographie

(1) LjungbergBetal.TheEpidemiologyofRenalCellCarcinoma.Europeanurology,60(2011),615-621.(2) B.Escudier,C.Porta,M.Schmidinger,F.Algaba,J.J.Patard,V.Khoo,T.EisenandA.HorwichRenalCell

Carcinoma:ESMOClinicalPracticeGuidelines.AnnOncol(2014)25(suppl3):iii49-iii56.(3) J.-J.Patardetal.Recommandationsenonco-urologie2013duCCAFU:Cancerdurein.ProgUrol,2013,

23,RecommandationS177-S204,suppl.2(4) (Projectionsdel’incidenceetdelamortalitéparcancerenFranceen2009.Donnéesdisponiblessurle

site de l’INVs.http://www.invs.sante.fr/applications/cancers/projections2009/rapport_projections_nationales_cancer_2009.pdf

(5) ChowWH, Devesa SS,Warren JL, Fraumeni JF Jr. Rising incidence of renal cell cancer in the UnitedStates.JAMA.1999May5;281(17):1628-1631.

(6) JemalA,SiegelR,WardE,HaoY,XuJ,MurrayT,ThunMJ.Cancerstatistics,2008.CACancerJClin.2008Mar-Apr;58(2):71-96.

(7) JJRambeaud.ProgUrol.2007;17(1Suppl.1):134-135.(8) (SurvieattenduedespatientsatteintsdecancersenFrance:étatdeslieux–avril2010.Disponiblesurle

sitedel’INCa:http://www.e-cancer.fr/les-soins/4211-survie-des-patientsatteints-de-cancers-en-france-linca-dresse-un-etat-des-lieux).

(9) FicarraV,NovaraG,LafrateM,CapellaroL,BrattiE,ZattoniF,ArtibaniW.ProposalforreclassificationoftheTNMstagingsysteminpatientswithlocallyadvanced(pT3-4)renalcellcarcinomaaccordingtothecancerrelatedoutcome.Eur.Urol.2007;51:722-731.

(10) MejeanA,Correas JM,EscudierB,deFromontM,LangH, Long JA,NeuzilletY,Patard JJ,PiechaudT.[Kidneytumors].ProgUrol.2007Nov;17(6):1101-44.

(11) MéjeanA, AndréM,Doublet JD, Fendler JP, de FromontM,HélénonO, LangH,Négrier S, Patard JJ,PiéchaudT;ComitédeCancérologiede l'AssociationFrançaised'Urologie. [Kidney tumors].ProgUrol.2004Nov;14(4Suppl1):997,999-1035

(12) J.-J. Patard, H. Baumert, J.-M. Corréas, B. Escudier, H. Lang, J.-A. Long, Y. Neuzillet, P. Paparel,L.Poissonnier,N.Rioux-Leclercq,M. Soulié.RecommandationsenOnco-Urologie2010 : Cancerdu rein.ProgrèsenUrologie(2010),20Suppl.4,S319–S339

(13) AtwellTD,SpanbauerJC,McMenomyBP,StocklandAH,HesleyGK,SchleckCD,HarmsenWS,WelchTJ.TheTimingandPresentationofMajorHemorrhageAfter18,947Image-GuidedPercutaneousBiopsies.AJRAmJRoentgenol.2015Jul;205(1):190-5.doi:10.2214/AJR.14.13002

(14) AllafME,Bhayani SB,RogersC,Varkarakis I, LinkRE, Inagaki T, Jarrett TW,Kavoussi LR. Laparoscopicpartialnephrectomy:evaluationoflong-termoncologicaloutcome.JUrol.2004Sep;172(3):871-873.

(15) MéjeanA,Correas JM,EscudierB,deFromontM,LangH, Long JA,NeuzilletY,Patard JJ,PiechaudT.[Kidneytumors].ProgUrol.2007Nov;17(6):1101-1144.

(16) D.A.Gervais,R.S.Arellano,P.R.Mueller.PercutaneousRadiofrequencyAblationofRenalCellCarcinoma.EurRadiol.2005May;15(5):960-7.

(17) Inderbir Gill IS, Hsu TH, Fox RL,Matamoros A,Miller CD, Leveen RF, GruneMT, SungGT, FidlerME.Laparoscopicandpercutaneousradiofrequencyablationofthekidney:acuteandchronicporcinestudy.Urology.2000Aug1;56(2):197-200.

(18) AtwellTD,CarterRE,SchmitGD,CarrCM,BoorjianSA,CurryTB,ThompsonRH,KurupAN,WeisbrodAJ,ChowGK, LeibovichBC,CallstromMR,PattersonDE.Complications following573percutaneous renalradiofrequency and cryoablation procedures. J Vasc Interv Radiol. 2012 Jan;23(1):48-54. doi:10.1016/j.jvir.2011.09.008.Epub2011Oct28.

(19) Samplaski MK, Hernandez A, Gill IS, et al. C-index is associated with functional outcomes afterlaparoscopicpartialnephrectomy.JUrol.2010.184(6):2259-63.

(20) HewMN,BaseskiogluB,BarwariK,etal.CriticalappraisalofthePADUAclassificationandassessmentoftheR.E.N.A.L.nephrometryscoreinpatientsundergoingpartialnephrectomy.JUrol.2011.186(1):42-6.

(21) Kutikov A andUzzo RG. The R.E.N.A.L. nephrometry score: a comprehensive standardized system forquantitatingrenaltumorsize,locationanddepth.JUrol2009;182:844-53.

(22) MatthewN.Simmons,ChristinaB.Ching,MaryK. Samplaski,ChinHyongPark, Inderbir S.Gill. KidneyTumorLocationMeasurementUsing theC IndexMethod.The JournalofUrology,Vol.183,May2010,1708-1713.

(23) Christopher J. Weight, Thomas D. Atwell, Robert T. Fazzio, Simon P. Kim, McCabe Kenny.A

44

MultidisciplinaryEvaluationofInter-ReviewerAgreementoftheNephrometryScoreandthePredictionofLong-TermOutcomes.JUrol,Vol.186,October2011,1223-1228.

(24) SimmonsMN, Ching CB, SamplaskiMK, et al. Kidney tumor locationmeasurement using the C indexmethod.JUrol.2010.183(5):1708-13.

(25) Schmit GD, Thompson RH, Kurup AN et al. Usefulness of R.E.N.A.L. nephrometry scoring system forpredictingoutcomesandcomplicationsofpercutaneousablationof751renaltumors.JUrol2013;189:30–5

(26) ChangX, LiuT,ZhangF,QianC, JiC,ZhaoX, LiuG,GuoH,ThecomparisonofR.E.N.A.L.,PADUAandCentrality Index score in predicting perioperative outcomes and complications following laparoscopicradiofrequencyablationofrenaltumors,TheJournalofUrology(2015),doi:10.1016/j.juro.2015.03.095.

(27) Seideman CA, Gahan J, Weaver M, Olweny EO, Richter M, Chan D, Cadeddu JA. Renal tumournephrometryscoredoesnotcorrelatewith theriskof radiofrequencyablationcomplications.BJU Int.2013Dec;112(8):1121-4.doi:10.1111/bju.12276.

(28) Okhunov Z, Shapiro EY, Moreira DM et al. R.E.N.A.L. nephrometry score accurately predictscomplicationsfollowinglaparoscopicrenalcryoablation.JUrol2012;188:1796–800

(29) Munveretal,Urology2002;60:170(30) TaylorGD,JohnsonDB,HoggDC,CaddedduJA.Developmentofarenaltumormimicmodelforlearning

minimallyinvasivenephronsparingsurgicaltechniques.JUrol2004;172(1):382-5(31) EunD, Bahndan A, Boris R, Rogers CG. A novel technique for creating solid renal pseudotumors and

renal vein-inferior vena caval pseudothrombus in a porcine and cadaveric model. J Urol 2008;180(4):1510-4

(32) RadiofrequencyAblation:Effectofsurroundingtissuecompositiononcoagulationnecrosisinacaninetumormodel.Muneebetal.Radiology,2003

(33) Bagetti Filho HJ, Pereira-Sampaio MA, Favorito LA, Sampaio FJ. Pig kidney: Anatomical relationshipsbetweentherenalvenousarrangementandthekidneycollectingsystem.JUrol2008;179:1627A30

(34) Guideforthecareanduseof laboratoryanimals. Instituteof laboratoryanimalressources.Comissiononlifesciences.Council.NR.WashingtonD.C.1996.Nationalacademypress

(35) Granger DN, Korthuis RJ. Physiologic mechanisms of postischemic tissue injury. Annual review ofphysiology1995;57:311A32

(36) RadiofrequencyAblation:Effectofsurroundingtissuecompositiononcoagulationnecrosisinacaninetumormodel.Muneebetal.Radiology,2003

(37) CrispenPL, ViterboR, Fox EB, et al. Delayed interventionof sporadic renalmasses undergoing activesurveillance.Cancer2008;112:1051-79

(38) GotraA,ChartrandG,Massicotte-TisluckK,Morin-RoyF,Vandenbroucke-MenuF,deGuiseJA,TangA.Validation of a Semiautomated Liver Segmentation Method Using CT for Accurate Volumetry. AcadRadiol.2015Apr20.pii:S1076-6332(15)00131-2.doi:10.1016/j.acra.2015.03.010.

(39) GChartrand, T Cresson, R Chav, AGotra, A Tang, J DeGuise. Semi-automated liver CT segmentationusingLaplacianmeshes.BiomedicalImaging(ISBI),2014IEEE11thInternationalSymposiumon,641-644

(40) Heimann T, van Ginneken B, Styner MA, et al. Comparison and evaluation of methods for liversegmentationfromCTdatasets.IEEETransMedImaging2009;28(8):1251–1265

(41) Birkfellner Wolfgang, Michael Figl, Bergmann Helmar. Stochastic rank correlation for slice-to-volumeregistrationoffluoroCT/CTimaging.In:MedicalImaging.InternationalSocietyforOpticsandPhotonics,2007.p.650937-650937-5.

(42) Frühwald, Laura, Joachim Kettenbach,Michael Figl, Johann Hummel, Helmar Bergmann, etWolfgangBirkfellner. 2009. « A comparative study onmanual and automatic slice-to-volume registration of CTimages».EuropeanRadiology,vol.19,n!11,p.2647-2653.

(43) Fei, Baowei, J.L.Duerk,D.T. Boll, J.S. Lewin, etD.L.Wilson.April 2003b. « Slice-to-volume registrationanditspotentialapplicationtointerventionalMRI-guidedradio-frequencythermalablationofprostatecancer».MedicalImaging,IEEETransactionson,vol.22,n°4,p.515-525.

(44) Burt, P.J. et E.H. Adelson. Apr 1983. « The Laplacian Pyramid as a Compact Image Code ».Communications,IEEETransactionson,vol.31,n!4,p.532-540

(45) Fei, Baowei, Jeffrey L. Duerk, Daniel T. Boll, Jonathan S. Lewin, et David L.Wilson. 2003. « Slice-to-volumeregistrationanditspotentialapplicationtointerventionalMRI-guidedradio-frequencythermalablationofprostatecancer».IEEETransactionsonMedicalImaging,vol.22,n!4,p.515–525

(46) W.Birkfellner,M.Figl,J.Kettenbach,J.Hummel,P.Homolka,R.Schernthaner,T.Nau,H.Bergmann:“2D/3DSlicetoVolumeRegistrationanditsApplicationonFluoroscopicCTImages”,MedPhys34(1),

246255,(2007)

45

Publicationsenrapportaveclathèse

I. Souteyrand P, Chagnaud C, Lechevallier E, Andre M. Radiofrequency ablation ofkidneytumors.ProgUrol.2013Nov;23(14):1163-7

II. SouteyrandP,CohenF,Daniel L, LechevallierE,ChagnaudC,RollandPH,AndreM,VidalV.Pathologicalfeaturesofradiofrequencyablation(RFA)renalscarCT-imaginginaswinemodel.ProgUrol.2013Feb;23(2):105-12

III. Souteyrand P, André M, Lechevallier E, Giorgi R, Chagnaud C, Boissier R. Usingmorphometric scores to predict RFA complications in renal tumors under 4cm.SoumisJournalofUrology

IV. DesmotsF,SouteyrandP,MarcianoS,LechevallierE,ZinkJV,ChagnaudC,AndréM.Morphometric scores for kidney tumours: use in current practice. Diagn IntervImaging.2013Jan;94(1):116-8.doi:10.1016/j.diii.2012.07.001.Epub2012Dec13

V. SouteyrandP;ChungW;DeGuiseJ;MariJL;LeonardiV;OliviéD;VidalV;SoulezG.AMRIDescriptionofKidneyMotion.SoumisJournalofMagneticResonanceImaging

VI. ChungW,SouteyrandP,SoulezG,ChartrandG,CressonT,MariJL,DeGuiseJ.Slice-to-volume registration for kidney guided interventions using MRI. En cours derelectureavantsoumissionTransactionsBiomedicalEngineering

Author's personal copy

Progrès en urologie (2013) 23, 1163—1167

Disponible en ligne sur

www.sciencedirect.com

Destruction par radiofréquence destumeurs du rein. Note technique

Radiofrequency ablation of kidney tumours

P. Souteyrand a,∗, C. Chagnaud a, É. Lechevallierb,M. Andre a

a Service de radiologie, hôpital Conception, AP—HM, 147, boulevard Baille, 13385 Marseille

cedex 05, Franceb Service d’urologie, hôpital Conception, AP—HM, 147, boulevard Baille, 13385 Marseille

cedex 05, France

Recu le 1er juillet 2013 ; accepté le 3 juillet 2013

L’épidémiologie des cancers du rein s’est modifiée depuis plusieurs années avec, notam-ment, une augmentation de son incidence de 30 % en 15 ans. Deux des raisons de cetteaugmentation sont les progrès de l’imagerie (scanner et échographie) et la multiplicationdu nombre d’examens. Par ailleurs, 80 % des cancers du rein sont aujourd’hui de découvertefortuite [1]. La population vieillit, les patients à traiter aussi.

Tous ces facteurs ont nécessité le développement d’alternatives thérapeutiques cura-tives aux traitements chirurgicaux « traditionnels » (la néphrectomie élargie). Ils doiventêtre au moins aussi efficace et épargner au maximum le capital néphronique des patients.La première avancée thérapeutique a été le développement et la validation de la néphrec-tomie partielle. Plus récemment, des techniques ablatives comme la radiofréquence (RF)ou la cryoablation ont été proposés.

Le principe technique de la radiofréquence

Le principe technique de la RF est l’application d’un courant alternatif (Fig. 1) àhaute fréquence par le biais d’une électrode introduite directement dans la tumeur(Fig. 2). L’électrode est positionnée sous guidage scanner (ou échographique dans cer-taines équipes). Le courant sinusoïdal est responsable d’une agitation ionique qui, parfriction, échauffe les tissus jusqu’à dépasser les 60 ◦C : il entraîne une mort cellulaireimmédiate par destruction thermique.

∗ Auteur correspondant.Adresse e-mail : [email protected] (P. Souteyrand).

1166-7087/$ — see front matter © 2013 Elsevier Masson SAS. Publié par Elsevier Masson SAS.http://dx.doi.org/10.1016/j.purol.2013.07.003


1164 P. Souteyrand et al.

Figure 1. Mise d’une électrode est introduite directement dansla tumeur. L’application d’un courant alternatif à haute fréquenceest responsable d’une agitation ionique qui, par friction, échauffeles tissus jusqu’à dépasser les 60 ◦C.

Pourquoi la radiofréquence peut êtreconsidérée comme un traitement curatif ?

D’abord, parce que la chirurgie partielle est un traitementvalidé. Ensuite, parce que les moyens de guidage en ima-gerie sont suffisamment précis pour positionner les aiguillesde RF dans la tumeur, tout en s’assurant de ne pas léser lesorganes ou les structures vasculaires, urinaires ou digestivesvoisines. Enfin, parce que les biopsies rénales (radioguidées)permettent aux anatomopathologistes de faire le diagnosticformel de cancer du rein sur les prélèvements.

Quelles sont les indications de laradiofréquence ?

Initialement, la RF étaient réservée aux patients inopé-rables, âgés ou avec un rein unique. Les indications ont étéélargies avec des résultats équivalents à la chirurgie par-tielle [2]. Le principal facteur prédictif de l’efficacité dutraitement est la taille de la lésion à traiter : Zagoria [3] a

Figure 2. Exemple d’une aiguille de radiofréquence LeVeen (Bos-ton Scientific®) parapluie : une fois l’aiguille en place, les baleinessont déployées ce qui permet d’englober la masse à traiter.

montré qu’il n’y avait pas de récidive pour des masses demoins de 4 cm.

On distingue deux types d’indications :• les indications de nécessité, pour des masses tissulaires

de moins de 40 mm, non sinusales, remplissant au moinsune de ces conditions :◦ supérieur à 70 ans et/ou facteurs de comorbidité,◦ rein unique et/ou fonction rénale altérée,◦ néoplasie associée et/ou localisations bilatérales,◦ cancer héréditaire (Von Hippel Lindau) ;

• les indications électives, par choix du médecin ou dupatient (les masses doivent mesurées moins de 40 mm,non sinusales).

Quelles sont les contre-indications de laradiofréquence ?

Elles correspondent aux « non-indications » (masse de plusde 4 cm, situation sinusale), aux contre-indications del’anesthésie générale (même si certains les réalisent sousneuro-analgésie), à la position en décubitus ventral prolongé(certaines insuffisances respiratoires) et aux contre-indications aux gestes percutanés (TP < 60 %, TCA > 2 foisle témoin, plaquettes < 50 000). Les anti-angiogéniquesdoivent être arrêtés mais le délai ne fait pas consensus.

Quel doit-être le bilan préopératoire ?

Le traitement curatif par RF doit être validé en Réunion deconcertation pluridisciplinaire. Le radiologue rencontre lepatient en consultation pré-RF pour lui expliquer le traite-ment, les risques et la surveillance : son consentement estrecueilli.

Outre une consultation d’anesthésie, le patient béné-ficie du même bilan que pour une néphrectomie :bilan d’hémostase (TP-TCA-plaquettes) et examen cyto-bactériologique des urines pour éliminer une infectionurinaire.

Pour tous les traitements ablatifs comme la RF, lediagnostic histologique doit être confirmé par une biopsieradioguidée (généralement par scanner, parfois sous écho-graphie) avant le traitement.

Déroulement de la procédure deradiofréquence et surveillance ?

Le patient est hospitalisé en urologie en moyenne deux nuits(en préopératoire et 24 heures après la RF). Le traitementse déroule dans le service de radiologie au scanner avec uneéquipe d’anesthésiste : le patient a été au préalable préparécomme pour une néphrectomie (toilette bétadinée, à jeundepuis plus de six heures), puis il est endormi et positionnésur la table de scanner en décubitus ventral. L’examendébute par un repérage de la lésion (scanner avec injec-tion de produit de contraste). L’asepsie, la surveillance,l’installation du patient et des champs opératoires suiventles mêmes standards qu’au bloc.


Destruction par radiofréquence des tumeurs du rein 1165

Enfin, le radiologue positionne l’aiguille de RF sousguidage fluoroscopique avec un abord retropéritonéale pos-térieur. On peut écarter des organes de voisinage (côlon,foie, rate. . .) proches de la lésion en réalisant une hydro-dissection (injection de sérum physiologique entre le reinet l’organe à protéger) ou en écartant avec du monoxydede carbone injecté par une seconde aiguille. Le proto-cole de chauffe est standardisé : on réalise des « tirs » pourcouvrir toute la lésion. Chaque tir correspond à deux procé-dures de chauffe (augmentation progressive de la puissance)jusqu’à obtention du Roll-off (résistance maximale du tissuchauffé au courant de RF qui signe la destruction tissulairecomplète). En fonction de la taille de la lésion, on choi-sit une aiguille de dimension adaptée : on réalise la plupartdu temps plusieurs tirs pour épouser la forme de la lésiontout en épargnant au maximum le parenchyme rénal sain. Enmoyenne, la durée du traitement dure 1h30 avec un tempsde chauffe pour deux « tirs » de 30 minutes (quatre Roll-offde 7 minutes 30 s). En fin de procédure, la dernière acquisi-tion scannographique a pour but d’évaluer des complicationspostopératoires immédiates.

Le patient est transféré ensuite en salle de soin post-interventionnel puis en service pour être surveillé commelors de tout geste chirurgical : pouls-tension-température-douleur-saignement-diurèse et hématurie. Le point deponction ne nécessite pas de point de suture, seulementun pansement. Avant sa sortie, 24 heures après traitement,sont réalisés des contrôles de la formule sanguine (hémoglo-bine), de la créatininémie et une échographie à la recherchede complications.

Le patient doit sortir avec les dates et les ordonnancespour les consultations et les examens de contrôle.

Suivi post-RF ?

La surveillance est la même que pour les traitements chirur-gicaux avec des uroscanners et des créatininémies à trois,six, 12 mois puis annuellement. Le critère d’efficacité prin-cipal est l’absence de rehaussement de la cicatrice de RF(la zone d’ablation) après injection de produit de contrasteiodé en scanner. La surveillance clinique est assurée parl’urologue traitant du patient.

Dans certains cas, on propose de surveiller la zoned’ablation par IRM plutôt que par scanner (par exempleen cas d’altération de la fonction rénale). Si une cicatriceest douteuse (suspicion de récidive ou de traitement insuf-fisant), on réalise une biopsie sous guidage scanner et unexamen anatomopathologique de la zone d’ablation.

Quelles sont les complications de laradiofréquence ?

On distingue les complications bénignes (qui ne modi-fient pas la prise en charge) des graves. Ce sont lesmêmes que pour la néphrectomie partielle : hématomerénal, brèche des voies urinaires, douleur au point deponction, hématurie temporaire pour les complicationsbénignes ; atteinte des organes de voisinage (pneumotho-rax, perforation digestive, atteinte hépatique) voir quelques

Figure 3. Uroscanner qui met en évidence une masse du rein droitde 29 mm de la lèvre médiale du méso-rein (la masse se rehausse defacon caractéristique après injection de produit de contraste iodé).

Figure 4. Acquisition scanner de repérage — patient en décubitusventral (deux images du haut). Vérification du bon positionnementde l’aiguille de radiofréquence dans la masse (en bas à gauche).Contrôle en fin de procédure qui ne retrouve pas de complicationmais seulement du gaz dans la zone de traitement (en bas à droite).



Figure 5. Échographie de contrôle à 24 h : pas de complication locale, le rein est vascularisé en doppler.

exceptionnelles complications létales (embolie pulmonairesmassives. . .).

Intérêts de la radiofréquence rénale etcomparaison par rapport à la chirurgiepartielle

Son efficacité oncologique est de 90 à 95 % [4], ce qui estéquivalent à la chirurgie partielle. Par contre son effica-cité fonctionnelle est supérieure puisque la diminution dudébit de filtration glomérulaire (le reflet de son fonctionne-ment) est de moins de 2 % contre 30 % pour les néphrectomiespartielles. Cette variation ne dépend pas de la taille de latumeur contrairement aux néphrectomies partielles [5]. Letaux de complications bénignes et graves est similaire pourles deux techniques 3,1 et 1,25 %. La durée d’hospitalisationest par ailleurs raccourcie (deux jours au lieu de cinq enmoyenne).

Figure 6. Échographie de contrôle à 18 mois.

Un exemple

Une masse du rein droit de 29 mm est découverte de faconfortuite sur un scanner (Fig. 3) chez un homme de 81 anssans antécédent. Il a bénéficié d’une biopsie rénale scan-noguidée qui a confirmée le diagnostic de carcinome àcellules conventionnelles de grade II de Führman. Le traite-ment curatif par RF a été validé en Réunion de concertationpluridisciplinaire. En consultation pré-RF, son consentementa été recueilli. Le patient a été traité sous anesthé-sie générale et avec un guidage fluoroscopique (Fig. 4)selon les protocoles de chauffe habituels. Il a été hospita-lisé 24 heures en urologie avec un contrôle échographiqueavant la sortie (Fig. 5) pour éliminer des complications.La surveillance s’est déroulé normalement avec, sur lescanner à 18 mois (Fig. 6), une cicatrice de 20 mm quine se rehaussait pas, ce qui attestait d’un traitementefficace.

En conclusion

La RF est une alternative thérapeutique fiable qui complètel’arsenal thérapeutique proposé à l’urologue puisque lesrésultats carcinologiques sont validés. C’est un traitementpour des patients non opérables (indication de nécessité)mais aussi une solution qui peut être choisie par le patient(indication élective).

Cette technique moderne et fiable ne bénéfice d’aucunfinancement, ce qui la réserve pour l’instant à des centresréférents. Sa reconnaissance par les autorités de tutelledevrait permettre de la diffuser plus largement.

Déclaration d’intérêts

Les auteurs déclarent ne pas avoir de conflits d’intérêts enrelation avec cet article.

Références

[1] Méjean A, André M, Doublet JD, Fendler JP, de Fromont M, Hélé-non O, et al. Kidney tumors. Prog Urol 2004;14(4 Suppl. 1):997[999—1035].


Destruction par radiofréquence des tumeurs du rein 1167

[2] Mejean A, Correas JM, Escudier B, de Fromont M, Lang H, LongJA, et al. Kidney tumors. Prog Urol 2007;17(6):1101—44.

[3] Zagoria RJ, Pettus JA, Rogers M, Werle DM,Childs D, Leyendecker JR. Long-term outcomesafter percutaneous radiofrequency ablation forrenal cell carcinoma. Urology 2011;77(6):1393—7,http://dx.doi.org/10.1016/j.urology.2010.12.077 [Epub2011 Apr 13].

[4] Gervais DA, Arellano RS, Mueller PR. Percutaneous radio-frequency ablation of renal cell carcinoma. Eur Radiol2005;15(5):960—7 [Epub 2005 Mar 9].

[5] Pettus JA, Werle DM, Saunders W, Hemal A, Kader AK, ChildsD, et al. Percutaneous radiofrequency ablation does not affectglomerular filtration rate. J Endourol 2010;24(10):1687—91,http://dx.doi.org/10.1089/end.2010.0029.

Disponible en ligne sur

www.sciencedirect.com

ARTICLE ORIGINAL

Cicatrice de radiofréquence rénale : corrélationsanatomopathologiques-tomodensitométriques chezle porc. Applications pratiques pour le suivi enimagerie!

Pathological features of radiofrequency ablation (RFA) renal scar CT-imaging ina swine model

P. Souteyrand a,∗, F. Cohenb, L. Danielb,É. Lechevallier c, C. Chagnaud a, P.H. Rollandd,M. Andre a, V. Vidald

a Servide de radiologie, hôpital Conception, AP—HM, 147, boulevard Baille, 13385 Marseille

cedex 05, Franceb Service de radiologie, hôpital Timone, AP—HM, 13005 Marseille, Francec Service d’urologie, hôpital Conception, AP—HM, 147, boulevard Baille, 13385 Marseille

cedex 05, Franced Laboratoire de physiopathologie et de thérapie vasculaire (L2PTV), faculté de médecine,

université Aix Marseille, 13005 Marseille, France

Recu le 20 juin 2012 ; accepté le 1er octobre 2012

MOTS CLÉSRadiofréquence ;Scanner ;Anatomopathologie ;Traitement ablatif

RésuméObjectifs. — Étudier la viabilité tissulaire et identifier les mécanismes d’action de la radiofré-quence (RF) à partir d’une confrontation imagerie—anatomopathologie de lésions de RF rénaleun mois après traitement sur un modèle porcin.Matériels et méthodes. — Vingt-quatre lésions de RF ont été réalisées sur six porcs anesthé-siés (traitement aux deux pôles des deux reins) avec des aiguilles LeVeen 2 cm sous guidagetomodensitométrique. Quatre semaines après traitement, les cicatrices étaient contrôlées enscanner, les reins explantés étudiés en anatomopathologie et en immuno-histochimie.Résultats. — Dix-neuf zones d’ablation (ZA) ont été étudiées en anatomopathologie et en TDM(volume, morphologie, rehaussement, cavitation) quatre semaines après RF. Une ZA abcédée

! Niveau de preuve : 3.∗ Auteur correspondant.

Adresse e-mail : [email protected] (P. Souteyrand).

1166-7087/$ — see front matter © 2012 Elsevier Masson SAS. Tous droits réservés.http://dx.doi.org/10.1016/j.purol.2012.10.004


était exclue des résultats ; quatre ZA n’ont pu être analysées en raison du décès précoce d’unporc pendant l’anesthésie. L’histologie des ZA décrivait un aspect hétérogène associant deszones de nécrose, d’ischémie et une surexpression de l’apoptose dans 50 % des ZA. Une nécrosedes ZA était constante en histologie alors que leur rehaussement était variable voire significatif(> 10 UH).Conclusion. — Un mois après traitement, l’examen anatomopathologique des ZA a mis enévidence des destructions tissulaires hétérogènes par différents mécanismes sans viabilité tis-sulaire centrale. Si l’examen TDM post-RF à un mois est utile pour dépister des complicationsiatrogènes, il ne peut prédire l’efficacité du traitement car des lésions ischémiques évolutivesdes ZA persistent. Les contrôles initiaux doivent donc être réalisés au moins trois mois après laRF.© 2012 Elsevier Masson SAS. Tous droits réservés.

KEYWORDSRadiofrequency;CT-scans;Pathology;Ablation techniques

SummaryPurpose. — To analyze the changes in vicinal kidney parenchyma after percutaneous RFA.Materials et methods. — Twenty-four CT-guided RFA procedures were performed on six pigsusing 2 cm LeVeen coaxial needles. We studied volume, morphology, cavitation and enhance-ment of the ablation zones (AZ) before and after the procedure on contrast-injected CT-scans.The kidneys were removed four weeks later and studied in the path lab.Results. — All the procedures were successfully completed. Four weeks later, the CT-scans sho-wed AZ that were either clearly circumscribed or with unclear borders, heterogenous areasassociating necrosis and infarct tissue and mesenchyma showing a process of apoptosis aroundthe edges. A treatment considered as incomplete on the CT-scan (presenting as an enhan-cement) was always associated with necrosis on the histology slides, although the necroticareas behaved in various different ways on the CT-scan after injection of contrast medium: anenhancement of more than 10 HU did not mean that no necrotic tissue was present.Conclusion. — RFA causes heterogenous tissue changes, associating necrotic and ischemic zonesand an apoptotic reaction. The mechanisms of these changes and their therapeutic significanceshould be studied. CT-scans performed immediately after RFA procedure and one month laterare not predictive of the efficacy of the treatment because an enhancement of the AZ does notmean that it is not necrotic. The value of a CT-scan performed one month after the procedureis debatable, because the tissue remodeling that occurs in the kidneys is not definitive at thistime-point.© 2012 Elsevier Masson SAS. All rights reserved.

Introduction

La radiofréquence (RF) est une technique d’ablation ther-mique tumorale par excitation moléculaire par un courantde RF, responsable d’une nécrose de coagulation.

Ses résultats [1] et ses taux de complications [2] équiva-lents à la chirurgie partielle ont permis de définir sa placedans le traitement des tumeurs rénales [3]. L’efficacité dutraitement est évaluée en imagerie par l’absence de rehaus-sement significatif après injection de produit de contraste[4], sans qu’il n’y ait jamais eu de confrontation entrel’imagerie et les lésions histologiques pour valider ce cri-tère.

De nombreuses équipes pratiquent un premier contrôle à24—48 heures ou à mois [5] alors qu’il n’existe pas de consen-sus pour le délai de ces contrôles [6]. Les résultats de cettecorrélation imagerie-histologie de la zone d’ablation (ZA)devaient permettre de préciser l’utilité du contrôle tomo-densitométrique (TDM) précoce en prenant l’histologie enréférence, ainsi que la valeur du rehaussement de la ZA.

Notre étude a eu pour objectif d’analyser la ZA deRF rénale sur un modèle animal, le rein de porc, pourconfronter, pour la première fois, l’aspect TDM de la ZA auxremaniements histologiques induits.

Matériels et méthodes

Le Comité d’éthique pour l’expérimentation animale deMarseille a autorisé cette étude prospective qui s’est dérou-lée sur trois mois.

Protocole de radiofréquence et TDM

Le modèle animal expérimental choisi était le rein sain deporc : il n’existe pas de modèle animal de tumeur rénale,et le rein de porc a une anatomie et une physiologie prochedu rein humain. Le même opérateur a réalisé à chaque pôlede chaque rein de six porcs mâles (race Pietrin, Blossin SA,13-Aubagne, France) (six mois, 40 ± 3 kg) le même protocolede RF que celui pratiqué chez l’Humain pour le traitementdes tumeurs rénales : les porcs sous anesthésie générale,intubés et ventilés, étaient installés en décubitus ventraldans le scanner, les quatre électrodes de retour reliées augénérateur (RF3000® Boston Scientific), une voie veineusepériphérique assurant les apports liquidiens, de médica-ments et du produit de contraste. Le positionnement desaiguilles (Needle LeVeen Superslim® 2 cm Boston Scientific)et le déploiement des baleines dans le cortex rénal étaientcontrôlés en échographie et en TDM.

La cicatrice de RF rénale : aspects anatomopathologiques et tomodensitométriques 107

Les protocoles de chauffe ont suivi les algorithmes propo-sés par le constructeur : paliers de 10 W toutes les minutes àpartir de 20 W jusqu’à obtention d’un premier Roll-off, sui-vie d’une séquence identique jusqu’au deuxième Roll-off (endébutant à la moitié de la puissance du premier Roll-off).

Trois TDM ont été réalisées (en apnée sous anesthésiegénérale) : la première en contraste spontané avant la RF, lesdeux TDM de surveillance immédiatement après la RF (j0) età quatre semaines (j28), sans et après injection (1,5 mL/kgpar injection) au temps néphrographique [7] pour étudierla forme, le volume (obtenu en multipliant les trois plusgrands diamètres par 0,55) et le rehaussement des ZA ets’assurer de l’absence de complication locorégionale. Laphase néphrographique a été privilégiée parce qu’elle cor-respond au délai de référence pour la surveillance des RFrénales [8]. Un rehaussement de moins de 10 UH après injec-tion a été considéré comme un succès en TDM [9,10] ; lanécessité de pratiquer une nouvelle RF en cas de rehausse-ment n’était pas considérée comme un échec mais commeun traitement incomplet [11].

Les porcs ont été sacrifiés quatre semaines après laséance de RF par injection intraveineuse d’un bolus de 15 mgde midazolam et de 25 mg de chlorpromazine avec 20 mLde KCl 15 %. Les reins prélevés étaient fixés dans du formolpendant 48 heures.

Le délai de quatre semaines avait été choisi parce que,dans la littérature [8,12], les lésions histologiques sontconsidérées comme définitives à cette date. Il s’agit du délairecommandé pour le premier contrôle tomodensitométriquede l’efficacité du traitement [5]. Goldberg [13] a défini lazone de parenchyme traité comme la ZA, sans préjuger desa nature histologique.

Examen anatomopathologique

Le même pathologiste spécialiste de la pathologie rénalea examiné les reins prélevés avec une colorationhématoxyline-éosine-safran (HES) et par deux marqueursimmuno-histochimiques, l’anti-CD10 (un marqueur des bor-dures en brosse des tubes contournés proximaux) et laPurified Rabbit Anti-Active Caspase-3 (une anticaspase mar-queur de l’apoptose).

Analyse statistique

Les résultats ont été rapportés en accord avec les Stan-dards for Reporting of Diagnostic Accuracy Criteria (STARD)[14]. Les variables quantitatives ont été exprimées médiane[intervalle interquartile], les variables qualitatives enpourcentage. Après la phase descriptive, une analysecomparative a été conduite à l’aide de tests du Chi2 ou testsexacts de Fisher pour les variables qualitatives, et de testsnon paramétriques (Mann-Whitney, Rho de Spearman) pourles variables quantitatives. Pour tous les tests bilatéraux,une valeur du degré de signification p inférieure à 0,05 a étéconsidérée comme statistiquement significative. Les don-nées statistiques ont été traitées avec le logiciel SPSS V15.

Résultats

Toutes les sessions de RF se sont correctement dérouléesavec obtention des deux Roll-off. Les porcs n’ont pas eu

Figure 1. Scanner de contrôle après la radiofréquence (j0) : lescanner met en évidence comme l’examen macroscopique (Fig. 4)un défaut de rehaussement triangulaire à base périphérique corti-cale qui correspond à un infarctus rénal.

d’altération de l’état général. Un porc est décédé à lafin de la RF des complications de l’anesthésie : ses reinsont été explantés tout de suite après la RF pour êtreexaminés en anatomopathologie. Les autres reins ont étéexplantés quatre semaines après la RF. Les RF des six porcsont donc permis l’étude de 23 ZA : quatre ZA « aiguës » àj0 et 19 « chroniques » à j28 (Tableau 1). Les deux seulescomplications tardives ont été un abcès d’une ZA (excluedes résultats) et une pneumopathie d’inhalation.

TDM post-RF (j0)

La morphologie du défaut de rehaussement des ZA était tri-angulaire à base périphérique corticale dans 84 % (16/19)(Fig. 1) et rondes pour les trois autres ZA. Le rehausse-ment médian des 19 zones étudiées était de 6 UH [3—11]et 26,3 % des ZA se rehaussaient précocement (> 10 UH)de facon homogène et diffuse. Le volume médian des ZAétait de 3,6 cm3 [2,8—4,4]. Des remaniements de cavitation(bulles de densité gazeuse) induits par la RF existaient dans69,4 % des ZA (Fig. 2).

TDM quatre semaines après la radiofréquence

Nous avons retenu dix traitements complets soit 52 % desuccès, huit traitements incomplets (48 %) (Fig. 3) et unabcès de la ZA communiquant avec un abcès du psoas. Lerehaussement médian des ZA était de 9 UH [5—21], leurvolume médian à j28 de 3,2 cm3 [2,6—4,4]. Ces rehausse-ments étaient tous diffus et homogène dans les ZA.

Figure 2. Scanner en fin de procédure avec des remaniements decavitation des deux zones d’ablation.


Tableau 1 Tableau des résultats des scanners de contrôle et des examens histologiques.

Scanner Anatomopathologie

Cochon Reht j0 Reht j28 Vol j0 Vol j28 Nécrose Zonemésenchymateuse

Ischémie

827.1 7 50 2,4 2,9 1 0 1827.2 3 80 1,9 1,8 1 0 1827.3 54 39 2,5 1,7 1 0 1827.4 130 59 1,3 2,1 1 0 1828 Mort j0 1 0 1829.1 1 14 2,4 4,3 1 1 1829.2 3 8 3,2 2,9 1 1 1829.3 2 15 1,6 1,8 1 1 1829.4 1 15 5,3 3,1 1 1 1830.1 6 8 2,6 1,7 1 1 1830.2 4 2 1,9 1,7 1 1 1830.3 6 4 2,1 2,1 1 1 1830.4 X X X X 1 1 1831.1 8 Abcès 1,3 Abcès 1 0 1831.2 7 7 3 3,9 1 0 1831.3 16 9 2,5 2,5 1 0 1831.4 5 3 1,6 2,1 1 0 1832.1 0 15 2,9 0,9 1 1 1832.2 12 6 2,7 1,7 1 1 1832.3 4 4 1,9 1,2 1 1 1832.4 11 5 4,2 2,1 1 1 1Moyenne

[déviation std]14 [30,3] 19 HU [22,5] 3,73 [1,46] 3,5 [1,35]

Reht : rehaussement ; Vol : volume ; j0, j28 : jour 0, jour 28 ; X : as de mesure.827.1 : cochon 827, pôle sup. rein droit.827.2 : cochon 827, pôle inf. rein droit.827.3 : cochon 827, pôle sup. rein gauche.827.4 : cochon 827, pôle inf. rein gauche.

L’examen anatomopathologique des zonesd’ablation immédiatement après laradiofréquence

Cet examen a décrit en macroscopie une zone circulaireou triangulaire blanchâtre, correspondant à la ZA (Fig. 4).L’examen histologique des quatre ZA a mis en évidence uneaugmentation de l’éosinophilie cytoplasmique, une perte denetteté des limites cellulaires, un aspect flou des noyaux,une hémorragie interstitielle et sur le trajet de l’aiguille,une infiltration inflammatoire et de la nécrose (dans cer-tains cas un hématome sous-capsulaire ou périrénal). Il yavait aussi une coagulation intravasculaire dans les tissusremaniés. Le marquage de l’apoptose par l’anticaspase étaitpositif, principalement dans la nécrose, mais aussi en péri-phérie de la nécrose.

La répartition de ces lésions était hétérogène au sein desZA. Sur certaines lames, il y avait un enchevêtrement dezones de nécrose, d’infiltrats inflammatoires et de tissussains, alors que sur d’autres lames, les zones étaient trèsbien limitées, comme coupées au couteau (Fig. 5). Dansune ZA, des zones tissulaires mixtes ou au contraire biendifférenciées pouvaient coexister, sans que l’on ait pu sys-tématiser leur répartition.

L’examen anatomopathologique des zonesd’ablation quatre semaines après laradiofréquence

Cet examen a décrit trois zones :• des zones de nécrose (centrale ou périphérique) qui asso-

ciaient perte de l’architecture tissulaire, disparition desnoyaux, pycnose, déformations cellulaires et ruptures desmembranes. Elles correspondaient à de la nécrose de coa-gulation typique (Fig. 6) ;

• des tissus en ischémie ou en souffrance, caractériséspar une coagulation intra-vasculaire, des noyaux irrégu-liers, mais une architecture conservée, dans 100 % desZA ;

• une zone mésenchymateuse inconstante, (50 % des ZAétudiées) correspondant à du tissu fibro-conjonctif, sesituait entre la zone de nécrose et les tissus enischémie.

En immuno-histochimie, la surexpression de l’anti-caspase dans les différentes zones témoignait d’uneapoptose induite par la RF. Cette réaction était constantedans 100 % des ZA, visible en périphérie des ZA et au centreoù elle prédominait.


Discussion

Notre étude avait pour objectif principal d’étudier la via-bilité tissulaire après RF et comme objectif secondaired’identifier les mécanismes d’action de la RF par uneconfrontation histologique et tomodensitométrique (TDM)de lésions de RF rénale un mois après traitement sur unmodèle porcin.

Les objectifs et le matériel de l’étude nous imposaientdes limites de choix du modèle expérimental car ces expé-riences n’étaient pas réalisables sur des tumeurs chezdes humains puisqu’elles auraient imposé deux anesthésies(pour la RF et pour la néphrectomie), sans bénéfice cli-nique pour le patient. Les tirs de RF étaient réalisés sur duparenchyme rénal sain de porc parce qu’il n’existait pas demodèle tumoral rénal animal. Notre modèle nous imposaitdonc deux contraintes majeures : extrapoler les résultatsdu porc vers l’humain mais aussi des tissus sains aux tissustumoraux. La seconde limite expérimentale résidait dans ladurée de l’étude limitée à un mois car notre objectif était

de montrer qu’à cette période les remaniements tissulairesn’étaient pas définitivement constitués et pouvaient évo-luer. Une étude plus longue aurait pu permettre d’analyserles remaniements secondaires à la RF au-delà (à trois, six et12 mois comme les contrôles TDM de surveillance).

En tomodensitométrie, les critères de réussite reconnusreposent sur un rehaussement de la ZA inférieur à 10UHaprès injection de produit de contraste [9,10]. Ce critèrede réussite a été défini par analogie avec les études desuivi des tumeurs hépatiques traitées par RF, bien qu’aucuneétude histologique pour le rein n’a jamais été réaliséepour le prouver [15]. Clark parlait même de critère arbi-traire [10]. La persistance d’une réaction inflammatoire(ou d’un tissu de granulation déjà décrit après RF hépa-tique [16]) 4 semaines après traitement était une hypothèsesans preuve histologique. Sur ces examens d’imagerie, leslimites des remaniements tissulaires des ZA à j0 étaientsoit floues avec des plages transitionnelles de composanteshistologiques mixtes (associant tissus sains, nécrotiques etinflammatoires) soit nettes, comme « coupées au couteau ».

Figure 3. Scanners quatre semaines après la radiofréquence. Exemples de mesures de ROI : absence de rehaussement (48 puis 51 UH < 10)qui signe un traitement efficace (1) et un traitement incomplet où le rehaussement (36 puis 64 UH > 10) est significatif (2).


Figure 3. (Suite ).

On pouvait alors penser que l’hétérogénéité des rema-niements tissulaires était due à l’association de deuxmécanismes : une ischémie induisant des limites très nettes(par thrombose vasculaire), et l’effet thermique des ondesde RF responsable de limites plus floues. L’analyse tomo-densitométrique du volume de la zone de traitement parrapport au volume théorique de l’ellipsoïde (le volumeenglobé par les baleines d’une aiguille de LeVeen de 2 cmest approximativement de 3,3 cm3) mettait en évidence unezone d’hypodensité de 3,6 cm3 à j0, et de 3,2 cm3 à j28.L’existence d’un infarctus dépassant la zone de traitementthéorique expliquait le volume supérieur de l’hypodensité àj0 au volume théorique traité. La diminution de volume desZA entre j0 et j28 s’expliquait par la rétraction parenchyma-teuse et la résorption de l’œdème.

Les corrélations anatomoradiologique, ont mis en évi-dence un mois après traitement par RF, des zones d’ischémieet de nécrose qui correspondaient en TDM à des imagesd’infarctus hypodenses, triangulaires, à base périphériquecorticale et aux contours bien limités. Ces zones avaientété mises en évidence par Gill [4] qui décrivait déjà la

dévascularisation de la zone traitée en artériographie sanspreuve histologique de l’infarctus tissulaire. L’élévation dela chaleur locale pouvait induire des zones de nécrose pardestruction cellulaire et des zones d’ischémie par destruc-tion des pédicules artériels afférents.

Sur des contrôles TDM très précoce, alors que Lui [17]suggérait, la possibilité de rehaussement dans du tissunécrotique, notre étude prouvait que dans 26 % des cas ilexistait à j0 des rehaussements significatifs dans la ZA alorsque ces zones contenaient sur l’analyse histologique défini-tive des zones de nécrose non viables.

La persistance des tissus en ischémie dans 100 % desZA à j28 était surprenante car on pouvait s’attendre à uneévolution vers une nécrose ou vers une re-perfusion. La per-sistance de tissus en ischémie à un mois signifiait que lesremaniements post-RF n’étaient pas définitivement consti-tués.

Si l’absence de rehaussement pouvait être considéréecomme un critère de réussite du traitement (puisqu’il étaittoujours corrélé à de la nécrose), le rehaussement de la ZAà un mois ne pouvait être considéré comme un échec : le


Figure 4. Aspect macroscopique de la zone d’ablation immédiatement après la radiofréquence : coupes axiale et sagittale d’un rein :infarctus rénal à base périphérique corticale (flèche blanche) et hématome péri-capsulaire (flèche noire) secondaire à la ponction parl’aiguille de radiofréquence. Il s’agit d’un des reins du porc 828 explanté immédiatement après la procédure de radiofréquence.

Figure 5. Zone d’ablation étudiée en immuno-histochimie avec le marqueur de l’apoptose : les régions colorées en marron sont marquéespar l’anticaspase 3 et témoignent de phénomènes d’apoptose. On distingue sur les deux lames (zoom × 100 et × 200) : 1- tissu rénal sainassocié à une souffrance des tubes (qui correspond à l’ischémie), 2- zone mésenchymateuse, 3- nécrose tissulaire, 4- coagulation intravasculaire, 5- marquage nucléaire de l’apoptose.

Figure 6. Lames des zones d’ablation 4 semaines après la procédure (coloration HES, zoom × 50 et × 100) : 1- nécrose tissulaire, 2- zonemésenchymateuse, 3- tissu rénal sain associé à une souffrance des tubes (qui correspond à l’ischémie).


rehaussement de la ZA devait être considéré en associationavec la progression volumétrique [8] et l’hétérogénéité durehaussement.

L’élément le plus étonnant était la découverte dezones d’apoptose de distribution hétérogène puisqu’iln’existait aucune répartition topographique systématisablede l’apoptose : le marquage était d’une facon prévisiblepositif au centre mais aussi en périphérie de la zone detraitement, distinct des zones probables d’élévation de latempérature, dans des tissus sains. L’ischémie et la chaleurinduite n’étaient donc probablement pas les seuls méca-nismes de destruction cellulaire induits par la RF sauf sil’on pouvait considérer une répartition inhomogène de latempérature avec des zones d’échauffement sub-critiquesou d’autres phénomènes de dénaturation protéique parl’application locale d’ondes électromagnétiques à hautesfréquences et hautes énergies.

Conclusion

Cette étude a mis en évidence des remaniements tissulaireshétérogènes dans leur forme et leur distribution : des tissusen ischémie, inflammatoires, mésenchymateux ou nécro-sés aux limites nettes ou flous, et une surexpression del’apoptose. Ces remaniements anatomopathologiques n’ontpas pu tous être corrélés formellement avec leur aspecttomodensitométriques. Ces résultats ont remis en questiondes dogmes sur la RF rénale comme l’intérêt du contrôletomodensitométrique post-RF, le délai du premier contrôlede surveillance et le rehaussement de la ZA comme critèred’insuffisance de traitement.

Elle proposait de différer au-delà d’un mois après la RF lepremier contrôle tomodensitométrique, d’associer systéma-tiquement l’étude du rehaussement de la ZA à la répartitiondu rehaussement et à la progression volumétrique.

La mise en évidence d’une réaction d’apoptose induitepar la RF au sein de tissus sain à distance de la zone detraitement signifiait que la nécrose secondaire à l’élévationde la température locale n’était peut être pas la seule voiede destruction tissulaire et ouvrait de nouvelles voies derecherche.

Déclaration d’intérêts

Les auteurs déclarent ne pas avoir de conflits d’intérêts enrelation avec cet article.

Références

[1] Méjean A, Correas JM, Escudier B, de Fromont M, Lang H, LongJA, et al. Kidney tumors. Prog Urol 2007;17(6):1101—44.

[2] Gervais DA, Arellano RS, Mueller PR. Percutaneous radio-frequency ablation of renal cell carcinoma. Eur Radiol2005;15(5):960—7.

[3] Méjean A, André M, Doublet JD, Fendler JP, de Fromont M,Hélénon O, et al. Kidney tumors. Prog Urol 2004;14(4 Suppl.1):999—1035.

[4] Inderbir Gill IS, Hsu TH, Fox RL, Matamoros A, Miller CD,et al. Laparoscopic and percutaneous radiofrequency abla-tion of the kidney: acute and chronic porcine study. Urology2000;56(2):197—200.

[5] Goldberg SN, Gazelle GS, Compton CC, Mueller PR, TanabeKK. Treatment of intrahepatic malignancy with radiofre-quency ablation: radiologic-pathologic correlation. Cancer2000;88(11):2452—63.

[6] Venkatesan AM, Wood BJ, Gervais DA. Percutaneous ablation inthe kidney. Radiology 2011;261(2):375—91.

[7] Debra A, Gervais, Francis J, McGovern, Ronald S, Arellano W,et al. Renal cell carcinoma: clinical experience and technicalsuccess with radio-frequency ablation of 42 tumors. Radiology2003;226:417—24.

[8] Wile GE, Leyendecker JR, Krehbiel KA, Dyer RB, Zagoria RJ. CTand MR imaging after imaging-guided thermal ablation of renalneoplasms. Radiographics 2007;27(2):325—39.

[9] Bossuyt PM, Reitsma JB, Bruns DE, et al. Towards complete andaccurate reporting of studies of diagnostic accuracy: the STARDInitiative. Radiology 2003;226:24—8.

[10] Zagoria RJ, Traver MA, Werle DM, Perini M, Hayasaka S, ClarkPE. Oncologic efficacy of CT-guided percutaneous radiofre-quency ablation of renal cell carcinomas. AJR Am J Roentgenol2007;189(2):429—36.

[11] Clark TW, Millward SF, Gervais DA, Goldberg SN, Grassi CJ,Kinney TB, et al. Reporting standards for percutaneous ther-mal ablation of renal cell carcinoma. J Vasc Interv Radiol2006;17(10):1563—70.

[12] Gervais DA, McGovern FJ, Arellano RS, McDougal WS, Muel-ler PR. Renal cell carcinoma: clinical experience and technicalsuccess with radiofrequency ablation of 42 tumors. Radiology2003;226(2):417—24.

[13] de Baere T, Kuoch V, Smayra T, Dromain C, Cabrera T,Court B, et al. Radiofrequency ablation of renal cell car-cinoma: preliminary clinical experience. J Urol 2002;167(5):1961—4.

[14] Goldberg SN, Grassi CJ, Cardella JF, Charboneau JW, Dodd3rd GD, Dupuy DE, et al. Image-guided tumor ablation: stan-dardization of terminology and reporting criteria. Radiology2005;235(3):728—39.

[15] Goldberg SN, Solbiati L, Hahn PF, et al. Radio-frequencytumor ablation using a clustered electrode technique: resultsin animals and patients with liver metastases. Radiology1998;209:371—9.

[16] Rowland IJ, Rivens I, Chen L, Lebozer CH, Collins DJ, terHaar GR, et al. MRI study of hepatic tumours following highintensity focused ultrasound surgery. Br J Radiol 1997;70:144—53.

[17] Lui KW, Gervais DA, Arellano RA, Mueller PR. Radiofre-quency ablation of renal cell carcinoma. Clin Radiol 2003;58:905—13.

1

Using morphometric scores

to predict RFA complications in renal tumors under 4cm

Purpose: Morphometric scores (MS) can predict the risk of acute intraoperative and

perioperative complications for partial nephrectomies performed for renal tumors.

Radiofrequency (RFA) is an alternative to surgery for those low-grade renal tumors less than

4cm, yet few studies have assessed the value of MS for this ablative technique.

Are MS predictors for the risk of complications after RFA for kidney tumors ?

Materiels and Methods: We identified those patients who received RFA for kidney tumors in our

institution since 2005. The MS (R.E.N.A.L., PADUA and C-Index) and the modified

morphometric scores (MRFS) took into account tumor position and size adaptation criteria. These

were established retrospectively and acute and late complications were reported according to

Clavien classifications.

The relationships between MS and any complications were analyzed by the chi-squared and

Fischer tests.

Results: 160 patients underwent 180 renal RFA treatments (mean follow-up: 25.5 months [1-

106]). There was a single RFA done for 145 patients and multiple RFAs were done for 15

patients in either simultaneous or several staggered treatments. The MS medians were six [5-10]

for PADUA, six [4-10] for R.E.N.A.L., and 2.95 [0.9-12.8] for the C-index. The MRSF medians

were four [3-7] for modified_PADUA, eight [6-12] for PADUA-RFA, four [2-6] for

modified_R.E.N.A.L. and nine [6-13] for R.E.N.A.L.-RFA.

2

There were 13 complications reported (7.22%): seven minor (Clavien I-II) and six major (Clavien

III-IV-V).

There was no correlation between MS and MRSF and the occurrence of any complications

(PADUA, RENAL and C-index: p = 0.32, 0.27 and 0.89).

Conclusions: The known and validated MS for partial surgery did not permit any predictions for

risk after a renal RFA. In particular, neither the size nor the position was a discriminant for

predicting any risk of complications.

Authors:

Philippe Souteyrand MD¹, 2

, Marc André MD¹, Eric Lechevallier MD3, Roch Giorgi MD PhD

4, 5,

6, Christophe Chagnaud MD¹, Romain Boissier MD

3.

1Department of Radiology, Hôpital La Conception APHM, 147 boulevard Baille, 13005

Marseille, France.

²Aix-Marseille Université, LIIE - Laboratoire d'Imagerie Interventionnelle Expérimentale EA

4264, 13005, Marseille, France ³Service de chirurgie urologique, Hôpital La Conception,

Marseille, France. 4Aix-Marseille Université, UMR_S 912 (SESSTIM), IRD, 13385 Marseille, France.

5INSERM, UMR_S 912 (SESSTIM), 13385 Marseille, France.

6APHM, Hôpital Timone, Service Biostatistique et Technologies de l’Information et de la

Communication, 13005, Marseille, France.

3

Manuscript

Introduction

The role of renal radiofrequency (RFA) is defined [1] with results in terms of efficiency, [2] and

complications [3] and is equivalent to those obtained in partial nephrectomies done for renal

masses of 4cm or less.

Nephrometry scoring systems quantify the complexity of renal tumors and are both reliable and

reproducible predictors of pre and postoperative complications of partial nephrectomies [4, 5, 6]

by using pre-operative computed tomography or magnetic resonance imaging. The R.E.N.A.L.

and PADUA scores are validated systems that characterize renal masses based on size, depth, and

their location in relationship to the collecting system or the renal sinus, and polarity [7, 8, 9].

R.E.N.A.L. nephrometry scoring consists of: Radius (tumor size as maximal diameter),

Exophytic/endophytic properties of the tumor, Nearness of the tumor’s deepest portion to the

collecting system or sinus, Anterior (a)/posterior (p) descriptor and the Location relative to the

polar line.

PADUA scoring (Preoperative Aspects and Dimensions Used for an Anatomical classification)

consists of the tumor’s anterior or posterior face, its longitudinal and rim locations, its

relationship to the sinus and urinary collecting system, its percentage of tumor deepening into the

kidney, and its maximal diameter in centimeters.

The C-index was developed to characterize tumor centrality based on the distance ratio between

the tumor, the kidney center and the tumor radius [10].

4

These scores have already been studied for renal thermal ablation, but for the most part, with no

distinction between the percutaneous or laparoscopic access, or between RFA and cryoablation

(CR). Only Schmitt [11] suggests that the R.E.N.A.L. nephrometry scoring system predicts perio-

and post-operative complications in patients undergoing percutaneous renal ablation. Other

authors [12, 13, 14] show no relationship.

Our main objective was to evaluate if morphometric scores played a role in predicting the risk

of complications with renal RFA. We compared these three scores to the series of patients treated

in our center by scanner guided RFA.

Our secondary objective was to propose an optimization of scores for percutaneous ablative

treatments by adapting the tumor size criteria and taking into account the antero-posterior

position of the tumor, which is reported in PADUA and R.E.N.A.L. scores but does not alter

these scores values.

5

Materials and Methods

Our ethics committee gave consent for the retrospective use of the clinical, laboratory and

imaging data of these patients.

All patients treated in our center by renal RFA between January 1st 2005 and December 31st

2014 have been included in this study. The indications for RFA validated by the Tumor Board

respected the following recommendations: clinical T1a (≤4.0 cm) enhancing renal mass after

percutaneous renal mass core biopsy [1].

The MS were calculated by either a preoperative computed tomography (CT) or an MRI. The

criteria viewed as crucial before performing an RFA was the tumor’s depth and position. Because

we wanted to propose an adaptation of PADUA and R.E.N.A.L scores to percutaneous renal

ablative treatments, we independently analyzed and associated the MS and its variables: 1) tumor

position, 2) tumor depth, and 3) the relationship to the renal sinus or collecting system in order to

compare them to the risks of acute complications. The combination of these three parameters

determined the modified-PADUA and modified-R.E.N.A.L scores.

We also proposed two other modified scores: PADUA-RFA and R.E.N.A.L-RFA where two

variables have been modified:

1) size, as suggested by Gahan [15]: 1 score ≤ 3cm / 3cm <2 score ≤ 4cm / 4cm <score 3;

2) position of the posterior tumor (p) score 1 / anterior (a) score 2.

6

For the R.E.N.A.L. and PADUA systems, the scores were totaled and further categorized into

low- (4-6 for R.E.N.A.L. and 6-7 for PADUA), moderate- (7-9 / 8-9) or high-complexity (10-

12/10-13) groups for risk stratification.

The C-index system was further categorized into low complexity (C-index score≥2.5), and high

complexity scores (C-index score <2.5) according to the study of Samplaski and his colleagues

[16].

All RFA were performed under CT-control using general anesthesia with a bipolar RF generator

(RF3000 ® Boston Scientific) and needles (LeVeen Needle ® Boston Scientific). Acute

complications were identified, and classified according to the Clavien-Dindo system [17].

Patients were followed up with CT or MRI imaging with sequences before and after contrast

injection along with acquisitions and reconstruction in the three plans, at 3, 6, and 12 months and

then annually for 10 years [18].

The commonly accepted criteria for success combine size evolution, enhancement of the ablation

zone (AZ) [19], as well as the emergence of new kidney locations or. In case of doubt for a

recurrence, a CT-controlled biopsy was performed. Treatments were only considered effective if

monitoring is longer than six months; otherwise they were excluded from the results.

The examinations were reviewed by the same radiologist (senior imaging uro-nephrology

specialist) on OsiriX software V6.5.2 with results reported on an Excel spreadsheet. Statistical

7

analysis was performed using the R software (version 3.1.0) with a p value <0.05 regarded as

statistically significant.

The continuous variables were described by their mean, standard deviation, and 1st and 3rd

quartile. The categorical variables were described by their numbers and percentage. The effect of

those factors associated with binary acute complications and their efficacy was tested by the chi-

squared test, or Fisher’s exact probability test, for categorical variables, and by the non-

parametric U test of Mann-Whitney for continuous variables. The effect of factors associated

with acute complications coded in three modalities was tested by the chi-squared test or the

probability of Fisher’s exact test for categorical variables, and by the non-parametric Kruskal

Wallis for continuous variables.

8

Results

The 160 patients treated were included in 180 RFA’s overall: 145 patients underwent RFA for an

injury, 10 patients underwent RFA for several different lesions, and five patients received a 2°

RFA on the AZ for a recurrence or for an incomplete treatment.

Of the 10 patients having an RFA for different lesions (synchronous or metachronous), seven

were treated for two lesions; two for three lesions and one for five lesions.

The characteristics of those treated patients are summarized in table 1: 180 tumors were treated

on 111 men and on 49 women. The mean age was 69 (22-86) and the majority were treated for

clear renal cell carcinomas (117-65%).

There were 13 (7.22%) complications identified (table 1). Seven of these were minor (Clavien I-

II classification): three with pain at the probe site and four with a limited hematoma. There were

six major complications (Clavien III-IV-V classifications): four with ureteral obstructions

(percutaneous nephrostomy or ureteral probe). These included two deaths: one due to a ruptured

abdominal aortic aneurysm the day following the RFA and the second from secondary multiple

organ failure 48 hours after RFA. The risk factor was identified as the maintenance of the anti-

angiogenic treatment during the procedure (the patient had not interrupted the treatment which is

contraindicated during RFA).

The results of the three scores and their correlation with complications are reported in table 2.

9

The mean scores were 6 (5-10) for PADUA; 6 [4-10] for R.E.N.A.L. and 2.95 [0.9–12.8] for the

C-index. The mean MRFS were 4 [3-7] for modified-PADUA, 8 [6-12] for PADUA-RFA, 4 [2-

6] pour modified-RENAL and 9 [6-13] pour R.E.N.A.L-RFA.

There was no significant association between the presence of complications and the R.E.N.A.L.

nephrometry score (p = 0.11), PADUA classification (0.18) and the C-index score (0.67), nor any

association between the scores and severity of complications (p = 0.27, 0.32 and 0.89). Mean

scores were higher in those cases with complications. There were 7 minor complications for a

mean PADUA of 6.7; a R.E.N.A.L. of 6.4; and a C-index of 3. There were six major

complications for 7.2, 6.8 and 4.3 respectively. But the differences of these averages, depending

on whether or not there were complications, are not important.

There was also no statistical relationship between the modified scores (modified-PADUA p =

0.27, PADUA-RFA 0.39, modified-R.E.N.A.L 0.38 and R.E.N.A.L-RFA 0.32), the parameters

that make up these scores independently from each other, or between the risk groups categories

(low, intermediate and high) and any acute complications (table 3).

The mean follow-up for the CT/MRI was 25.5 months [1-106]. Monitoring showed effective

RFA treatments for 132/180 (73%) for 2 incomplete treatments and one recurrence on AZ

(1.7%). In the other 35 cases (19.4%), there was no recurrence on the AZ but there was an

emergence of new renal or remote locations and 2 deaths in the immediate aftermath of the RFA.

Ten patients were excluded because the post RFA monitoring was less than 6 months.

10

Discussion

Our series was consistent with published RFA series for the distribution of tumor type, size, and

complication rates as well as for long-term results [20].

There were 13 complications identified: 7 minor complications (Clavien I-II) and 6 major

complications (Clavien III-IV-V). There was no significant association between complications

and MS.

Chang [12] also found no association between the scores and complications; however those RFA

were performed laparoscopically. Seidman [13] also found no association between the scores and

complications, but he only analyzed RFA R.E.N.A.L scores without any distinction between a

laparoscopic or a percutaneous approach.

Surprisingly, Schmit [11] suggests that the R.E.N.A.L. nephrometry scoring system predicts

perioperative and postoperative complications in patients undergoing percutaneous R.E.N.A.L.

ablation but without any distinction for RFA and cryoablation (CR).

Okhunov [14] showed that the R.E.N.A.L. score is predictive of cryoablation complications for

small renal masses. There are no publications studying only percutaneous RF scores and

complications.

There are several arguments to explain the lack of any association between the scores and the risk

of acute complications:

11

-the 1st score factor is tumor size with the first level at 4 cm. For RFA, this factor is "1" since it

deals with masses ≤ 4 cm (T1a) and, according to the recommendations of the American

Urological Association (AUA) [1], lesions greater than 4cm are at risk for a sub-optimal

treatment.

Besides, Gahan [15] has already proposed adapting an "m-R.E.N.A.L" modified score where he

adjusts size limit factors (3 and 4 cm rather than 4 and 7 cm) with an improvement in the

performance score.

- According to the AUA [1], an RFA is contraindicated if the sinus or the collector system is

invaded: a 2nd PADUA and R.E.N.A.L factor is stable at 1.

The R.E.N.A.L. and PADUA scores characterized those renal masses that were based, in

particular, on an anterior vs posterior plane. While this factor does not change the score, it is a

fundamental criterion for recommending an RFA. The crucial criteria to consider before

performing an RFA are the tumor depth, position and size. In order to propose PADUA and

R.E.N.A.L score adaptations for renal percutaneous ablative treatments, we independently

analyzed and associated variables: 1) tumor position, 2) tumor depth, and 3) the relationship to

renal sinuses or collective systems, so as to compare the risk of acute complications. The

combination of the three parameters determined both the modified-PADUA and the modified-

R.E.N.A.L scores, which we then compared for risk complications by linking both scores.

To highlight the "location" criterion, which was reported but not recognized for both the PADUA

and R.E.N.A.L. scores, we integrated it with the Gahan score [15] (which modifies the tumor size

criterion) in order to obtain the PADUA-RFA and R.E.N.A.L-RFA scores. Regardless of the

12

independent parameters taken or the modified scores, there is no association to the risk of

complications.

This study has several limitations worth discussing. First, it is limited by the fact that data were

retrospectively analyzed and only one radiologist calculated scores. We considered that these

scores were reproducible and consistent with the literature [9]. Additionally, patients were

selected based on a tumor size not exceeding 4cm, but this size limit was imposed by the RFA

[1].

Conclusion

The PADUA, R.E.N.A.L. and C-index classifications all fail to predict acute complications

associated with RFA treatments. Notably neither tumor size nor position was a discriminant in

predicting the risk of complications. The adaptation of PADUA and R.E.N.A.L. scores along

with criteria modification of size and the integration of the anterior or posterior position of the

tumor do not predict the risk of complications of percutaneous renal RFA. The results of our

study confirmed that RFA is a safe method to treat kidney tumors (clinical T1a ≤4.0 cm). RFA

may be recommended as an elective treatment alternative to surgery and obtains identical

oncological results.

13

References

1. Campbell SC, Novick AC, Belldegrun A, et al. Guideline for management of the clinical

T1 renal mass. J Urol. 2009. 182(4): 1271-9

2. Méjean A, Correas JM, Escudier B, de Fromont M, Lang H, Long JA, Neuzillet Y, Patard

JJ, Piechaud T. [Kidney tumors]. Prog Urol. 2007 Nov;17(6): 1101-1144.

3. D.A. Gervais, R.S. Arellano, P.R. Mueller. Percutaneous Radiofrequency Ablation of

Renal Cell Carcinoma. Eur Radiol. 2005 May;15(5): 960-7.

4. Samplaski MK, Hernandez A, Gill IS, et al. C-index is associated with functional

outcomes after laparoscopic partial nephrectomy. J Urol. 2010. 184(6): 2259-63.

5. Ficarra V, Novara G, Secco S, Macchi V, Porzionato A, De Caro R, Artibani W.

Preoperative aspects and dimensions used for an anatomical (PADUA) classification of

renal tumours in patients who are candidates for nephron-sparing surgery. Eur Urol. 2009

Nov;56(5):786-93. doi: 10.1016/j.eururo.2009.07.040. Epub 2009 Aug 4.

6. Hew MN, Baseskioglu B, Barwari K, et al. Critical appraisal of the PADUA classification

and assessment of the R.E.N.A.L. nephrometry score in patients undergoing partial

nephrectomy. J Urol. 2011. 186(1): 42-6.

7. Kutikov A and Uzzo RG. The R.E.N.A.L. nephrometry score: a comprehensive

standardized system for quantitating renal tumor size, location and depth. J Urol

2009;182:844-53.

8. Matthew N. Simmons, Christina B. Ching, Mary K. Samplaski, Chin Hyong Park,

Inderbir S. Gill. Kidney Tumor Location Measurement Using the C Index Method. The

14

Journal of Urology, Vol.183, May 2010, 1708-1713.

9. Christopher J. Weight, Thomas D. Atwell, Robert T. Fazzio, Simon P. Kim, McCabe

Kenny.A Multidisciplinary Evaluation of Inter-Reviewer Agreement of the Nephrometry

Score and the Prediction of Long-Term Outcomes. J Urol ,Vol. 186, October 2011, 1223-

1228.

10. Simmons MN, Ching CB, Samplaski MK, et al. Kidney tumor location measurement

using the C index method. J Urol. 2010. 183(5): 1708-13.

11. Schmit GD, Thompson RH, Kurup AN et al. Usefulness of R.E.N.A.L. nephrometry

scoring system for predicting outcomes and complications of percutaneous ablation of

751 renal tumors. J Urol 2013; 189: 30–5

12. Chang X, Liu T, Zhang F, Qian C, Ji C, Zhao X, Liu G, Guo H, The comparison of

R.E.N.A.L., PADUA and Centrality Index score in predicting perioperative outcomes and

complications following laparoscopic radiofrequency ablation of renal tumors, The

Journal of Urology (2015), doi: 10.1016/j.juro.2015.03.095.

13. Seideman CA, Gahan J, Weaver M, Olweny EO, Richter M, Chan D, Cadeddu JA. Renal

tumour nephrometry score does not correlate with the risk of radiofrequency ablation

complications. BJU Int. 2013 Dec;112(8):1121-4. doi: 10.1111/bju.12276.

14. Okhunov Z, Shapiro EY, Moreira DM et al. R.E.N.A.L. nephrometry score accurately

predicts complications following laparoscopic renal cryoablation. J Urol 2012; 188:

1796–800

15. Gahan JC, Richter MD, Seideman CA, Trimmer C, Chan D, Weaver M, Olweny EO,

Cadeddu JA. The performance of a modified RENAL nephrometry score in predicting

renal mass radiofrequency ablation success. Urology. 2015 Jan;85(1):125-9. doi:

10.1016/j.urology.2014.08.026. Epub 2014 Oct 18.

15

16. Samplaski MK, Hernandez A, Gill IS, et al. C-index is associated with functional

outcomes after laparoscopic partial nephrectomy. J Urol. 2010. 184(6): 2259-63.

17. Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new

proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg

2004; 240: 205–13

18. Goldberg SN, Gazelle GS, Compton CC, Mueller PR, Tanabe KK. Treatment of

intrahepatic malignancy with radiofrequency ablation: radiologic-pathologic correlation.

Cancer. 2000 Jun 1;88(11): 2452-2463.

19. Inderbir Gill IS, Hsu TH, Fox RL, Matamoros A, Miller CD, Leveen RF, Grune MT,

Sung GT, Fidler ME. Laparoscopic and percutaneous radiofrequency ablation of the

kidney: acute and chronic porcine study. Urology. 2000 Aug 1;56(2): 197-200.

20. Atwell TD1, Carter RE, Schmit GD, Carr CM, Boorjian SA, Curry TB, Thompson RH,

Kurup AN, Weisbrod AJ, Chow GK, Leibovich BC, Callstrom MR, Patterson DE.

Complications following 573 percutaneous renal radiofrequency and cryoablation

procedures. J Vasc Interv Radiol. 2012 Jan;23(1):48-54. doi: 10.1016/j.jvir.2011.09.008.

Epub 2011 Oct 28.

16

Figure Legends

Table 1: Patient demographics and renal tumor characteristics.

Table 2: The 3 scores and complications.

Table 3: Statistical results:

• Modified-R.E.N.A.L corresponds to R.E.N.A.L. calculated without R or N.

• R.E.N.A.L-RFA corresponds to matches modified-R.E.N.A.L score taking into account

modified size criteria and the position a / p.

• Modified-PADUA corresponds to PADUA score calculated without sinus or size.

• PADUA-RFA corresponds to modified-PADUA score taking into account the modified

size criteria and the position a / p.

• Minor complications (Clavien I-II) and major complications (Clavien III-IV-V).

RFA 180

Number of patients 160

Gender (n,%)

Male 111 (70)

Female 49 (30)

Mean (range) :

Age (years) 69 (22-86)

Tumour size (mm) 23.8 (8-40)

Tumor side (n,%)

Left 79 (48.9)

Right 101 (61.1)

Tumor location (n,%)

Anterior 68 (35.8)

Posterior 112 (62.2)

Exophytic (n,%)

>50% 93 (51.7)

<50% 69 (38.3)

Endophytic 18 (10)

Biopsy data (n,%) 180

Clear renal cell carcinoma 117 (65)

Chromophobe carcinoma 11 (6.2)

Papillary carcinoma 36 (20)

Oncocytoma 8 (4.4)

Angiomyolipoma 2 (1.1)

Other 6 (3.3)

Complications (n, %)

minor (Clavien I-II) 7 (3.9)

I 5 (2.8)

II 2 (1.1)

major (Clavien III-IV-V) 6 (3.3)

III 3 (1.7)

IV 1 (0.5)

V 2 (1.1)

Follow up (month) 25.8

Effective RFA treatments 132 (73.3)

Treatment failures 38 (21.1)

Incomplete treatment or recurrence on ZA 3 (1.7)

New location 35 (19.4)

Excluded (follow-up < 6 months) 10 (5.6)

R.E.N.A.L. 1 2 3

R. 180

E. 92 70 18

N. 113 40 27

Anterior / posterior (no given

point, but it is specified in the

tumour description, a, p) 66 114 -

L. 80 36 64

R.E.N.A.L. Medians [SD] 6 [4-10]

modified-R.E.N.A.L. 4 [2-6]

R.E.N.A.L.-RFA 9 [6-13]

PADUA 1 2 3

Size 180 - -

Longitudinal location 78 102 -

Exophytic rate 92 70 18

Renal rim 129 51 -

Renal sinus 166 14 -

PADUA 6 [5-10]

modified-PADUA 4 [3-7]

PADUA-RFA 8 [6-12]

C-index mean 3.42 (0.9-12.8)

Complications

(Clavien)SD 1

st quartile median 3

rd quartileNo patients p

PADUA 0.32

None 1.24 5 6 7 167

I - II (minor) 1.25 6 7 7 7

III - IV - V (major) 0.98 7 7 7 6

PADUA-modified 0.27

None 1.13 3 4 5 167

I - II 1.25 4 5 5 7

III - IV - V 0.63 5 5 5 6

PADUA-RFA 0.39

None 1.42 7 8 9 167

I - II 1.11 7.5 8 9 7

III - IV - V 1.67 8.25 9 9 6

R.E.N.A.L. 0.27

None 1.72 4 6 7 167

I - II 1.81 5.5 6 8 7

III - IV - V 1.47 6 6.5 7.75 6

RENAL-modified 0.38

None 1.27 2 4 5 167

I - II 1.21 3 4 5 7

III - IV - V 0.89 3.25 4 4.75 6

RENAL-RFA 0.32

None 1.87 7 9 10 167

I - II 1.86 7.5 9 11 7

III - IV - V 1.97 8.25 9 10.5 6

C-INDEX 0.89

None 1.62 2.22 3 4.15 167

I - II 1.11 2.41 2.59 3.62 7

III - IV - V 4.23 2.46 2.9 3.41 6

No patients (%)

Complic

ations

(%)

p

PADUA 0.73

Low 6–7 141 (78) 11 (85)

Moderate 8–9 and High ≥10 39 (22) 2 (15)

R.E.N.A.L. 0.24

Low 4–6 111 (62) 7 (54)

Moderate 7–9 and High 10–12 69 (38) 6 (46)

Please cite this article in press as: Desmots F, et al. Morphometric scores for kidney tumours: Use in current practice.Diagnostic and Interventional Imaging (2012), http://dx.doi.org/10.1016/j.diii.2012.07.001

TECHNICAL NOTE / Genito-urinary imaging

Morphometric scores for kidney tumours: Use in

current practice

F. Desmots a, P. Souteyrand a, S. Marciano a,E. Lechevallierb, J.V. Zink a, C. Chagnaud a,M. André a,∗

a Radiology Department, Hôpital La Conception, AP—HM, 147, boulevard Baille,

13385 Marseille cedex 5, Franceb Urology Department, Hôpital La Conception, AP—HM, 147, boulevard Baille, 13385 Marseille

cedex 5, France

KEYWORDSPartial nephrectomy;Morphometric score;Cancer of the kidney;Complications

Tissue masses in the kidney discovered by chance are continuously increasing with thediffusion of imaging in cuts. They make up a very heterogeneous and progressive anato-mopathological contingent, the treatment of which has considerably changed over the past20 years, both in the pretreatment evaluation and in the surgical approach, where par-tial nephrectomy (PN) has become the standard due to its good results on the cancer andprogress in surgical techniques. PNs have been enriched by percutaneous removal treat-ments that complete the therapeutic range offered to patients. However, if PN makeskidney conservation possible with a long-term survival rate that is identical to that oflarger nephrectomise, they cause more common per- and pericomplications (19 to 25%).Currently, if there are not true tumoural prognosis factors, each patient befits in RCPfrom a personalised pre-therapeutic and therapeutic proposal that is locally adapted totechnical settings and medico-surgical teams.

Clinical interest

Like the Bosniak classification, morphometric scores give a ranking based on the size andlocation of the tumour using a CT-scan or MRI.

∗ Corresponding author.E-mail address: [email protected] (M. André).

2211-5684/$ — see front matter © 2012 Éditions francaises de radiologie. Published by Elsevier Masson SAS. All rights reserved.http://dx.doi.org/10.1016/j.diii.2012.07.001


2 F. Desmots et al.

These scores were suggested by Kutikov (RENAL nephrom-etry score) and Ficcara (PADUA) in 2009 and by Simmons(C-Index) in 2010 in order to standardise series and giveradiologists the tools to make it possible to classify tissuemasses and provide information that would allow urologiststo offer suitable therapeutic choices based on objec-tive morphological data [1—3]. The first series publishedusing these classification criteria show a good interobserverreproducibility [4] and provide prognosis evaluation infor-mation on per- and perioperative morbidity [5]. This cansignificantly predict (P < 0.001) the risk of complication,while the size of the lesion or its anterior or poste-rior location taken alone (univariate analysis) does notmake it possible to conclude [1]. In addition, the mor-phometric score is an interesting piece of information forpredicting the peroperative clamping time: the higher thescore (PADUA and R.E.N.A.L.), the longer the ischaemiatime, which indirectly appreciates operative difficulties(P < 0.001) [6].

Anatomical classifications

PADUA

Preoperative Aspects and Dimensions Used for an Anatomicalclassification. This score varies from 6 to 14 (Table 1) andtakes into account five lesional anatomical characteristicsin addition to the maximum diameter. A score of 8 and 9makes it possible to identify patient groups with a risk ofcomplication 14 times higher than with scores of 6 and 7.Concerning the reproducibility of the determination of thePADUA score, the interobserver correlation is approximately73% [5].

RENAL

Radius: maximum diameter of the lesion, E: exophytic orendophytic tumour, N: nearness or tumour extension indepth. This score varies from 4 to 12 (Table 2) to which thetwo suffixes a or p for anterior or posterior are attributedand the letter h if the lesion is in contact with the veinor main artery. In this score, the maximum transverse

Table 2 RENAL morphometric score.

1 point 2 points 3 points

(R)adius (maximum diameter in cm) ≤ 4 > 4 but < 7 ≥ 7

(E)xophytic (exophytic development) ≥ 50% < 50% Entirely endophytic

(N)earness (nearness of the tumour to theurine collection system in mm)

≥ 7 > 4 but < 7 ≤ 4

(A)nterieur/Posterior No given point, but it is specified in the tumour description, a, p or x

(L)ocation (tumour location relative to polarlinesa)

Entirely above theupper polar line orbelow the lowerpolar line

The lesionexceeds thepolar lines

> 50% of the lesion exceeds apolar line or crosses the axialor medio kidney line or islocated entirely between thepolar lines

Suffix ‘‘h’’ assigned to a tumour reachingthe main artery or vein

a A diagram is available on the internet on the website www.nephrometry.com.

Table 1 Preoperative aspects and dimensions used foran anatomical classification morphometric score.

Anatomical descriptionsa Score

Longitudinal tumour location (polar)Superior/inferior 1Mean 2

Exophytic> 50% 1< 50% 2

Endophytic 3

Axial tumour locationLateral 1Medial 2

Sinus of the kidneyNot infiltrated 1Infiltrated 2

Tumour size (cm)≤ 4 14.1—7 2> 7 3

a The anterior or posterior location can be indicated with theletters (‘‘a’’ or ‘‘p’’) after the score.

diameter: R (Radius) is a statistically significant predictivemarker (P = 0.02) of the risk of complication in studies witha single variable. In multivariate studies, a RENAL scoregreater than or equal to 9 is an independent and reliable fac-tor of the risk of complication. A score of 10 to 12 increasesthe risks of major complications 5.4 times [7].

C-Index

Index that calculates the ratio of the distance of the cen-tre of the tumour to the centre of the kidney: c forthe size of the tumour: r C-index = c / r [3]. This score(Fig. 1) varies from 0 with an upper limit dependent onthe size and distance of the tumour from the centre of the


Morphometric scores for kidney tumours: Use in current practice 3

Figure 1. According to Simmons [3], the distance between the two centres of the tumour and the kidney is calculated using the PythagoreanTheorem: c =

√(x2 + y2). The index is obtained by the ratio of the distances of the two centres (C) divided by the ray of the mass.

kidney. A C-index of 2.5 increases, for example, by 30% therisk of a major complication. This method appears to bemore complex to implement and requires a longer learningcurve.

Conclusion

Use of morphometric scores to evaluate kidney tumours isa more rigorous approach to their treatment, even if thesemethods remain imperfect and do not include, for example,the speed of growth and the anatomopathological Fuhrmanscore, which are currently the most commonly used pro-gnostic factors. These classifications are relatively close,and the PADUA score appears to be the easiest to imple-ment. Communication in the CT-scan and MRI reports ofa PADUA or RENAL score should make it possible to makethe prognostic evaluation of kidney tumours by radiologistsmore homogeneous. The use of this score in preoperativeevaluation of percutaneous thermal ablations remains to bedefined.

Disclosure of interest

The authors declare that they have no conflicts of interestconcerning this article.

References

[1] Ficarra V, Novara G, Secco S, Macchi V, Porzionato A, De CaroR, et al. Preoperative aspects and dimensions used for ananatomical (PADUA) classification of renal tumours in patientswho are candidates for nephron-sparing surgery. Eur Urol2009;56:786—93.

[2] Kutikov A, Uzzo RG. The R.E.N.A.L. nephrometry score: a com-prehensive standardized system for quantitating renal tumorsize, location and depth. J Urol 2009;182:844—53.

[3] Simmons MN, Ching CB, Samplaski MK, Park CH, Gill IS. Kidneytumor location measurement using the C-index method. J Urol2010;183(5):1708—13.

[4] Weight CJ, Atwell TD, Fazzio RT, Kim SP, Kenny M, Lohse CM,et al. A multidisciplinary evaluation of inter-reviewer agree-ment of the nephrometry score and the prediction of long-termoutcomes. J Urol 2011;186(4):1223—8.

[5] Hew MN, Baseskioglu B, Barwari K, Axwijk PH, Can C, Horen-blas S, et al. Critical appraisal of the PADUA classificationand assesment of the R.E.N.A.L. nephrometry score in patientsundergoing partial nephrectomy. J Urol 2011;186:42—6.

[6] Simhan J, Smaldone MC, Tsai KJ, Canter DJ, Li T, Kutikov A, et al.Objective measures of renal mass anatomic complexity predictrates of major complications following partial nephrectomy. EurUrol 2011;60(4):724—30.

[7] White MA, Haber GP, Autorino R, Khanna R, Hernandez AV,Forest S, et al. Outcomes of robotic partial nephrectomyfor renal masses with nephrometry score of ≥7. Urology2011;77(4):809—13.

FOR

PEER

REVIE

W O

NLY

A MRI Description of Kidney Motion

Journal: Journal of Magnetic Resonance Imaging

Manuscript ID Draft

Wiley - Manuscript type: Original Research

Classification:

Motion supression algorithms < Imaging technology and safety < Basic Science, Mathematical models of imaging processes < Imaging technology and safety < Basic Science, MRI Guided Therapy < Clinical Science, Lower abdomen (kidneys, retroperitoneum, colon) < Body imaging < Clinical Science

Manuscript Keywords: KIDNEY, MOVEMENT, MRI

Journal of Magnetic Resonance Imaging


FOR

PEER

REVIE

W O

NLY

Title:

A MRI Description of Kidney Motion.

Keywords:

MRI; kidney; movement.

Page 1 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Abstract:

Objective.

Ablation therapy of renal mass by focused ultrasound requires a real time estimation of

kidney motion including rotation or deformation. The goal of this project is to analyze kidney

motion with 3D MRI.

Methods.

Ten healthy volunteers were enrolled in this study. They underwent 3D T1 weighted MR

acquisitions at different phases of the respiratory cycle with volume reformation. After

segmentation of kidney contours, kidney motion in the 3 space direction was estimated on

several critical points during different respiratory phases. Kidney rotation and deformation

was also estimated.

Results.

With the MRI semi-automated segmentation algorithm, we determined the positions of three

specific points in the kidney throughout the respiratory cycle and then described the 3D

motion of the kidney. These kidney positional shifts were more important in the cranio-caudal

direction (28.9 ±12.2mm and 23.4 ±14.7 mm for the right and left kidneys); but these were

more complex than a linear excursion since kidney rotation (X 5.9° - Y 1.3° - Z 1° on the

right, and 5.2° - 2° - 3° on the left) was also observed. Its displacement was not linear

between the two extreme positions. We do not observed significant kidney deformation

during the respiratory cycle.

Conclusion.

Our study shows the need for a 4D renal tracking, to be able to target a focal lesion in the

kidney during the respiratory cycle.

Page 2 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Introduction

Kidney cancer treatments are shifting towards minimally invasive therapies [1], which prove

to be as effective as open renal surgery, but with lower morbidity, shorter hospitalizations,

reduced costs and unsurpassed patient comfort.

Beside needle radiofrequency ablation or tumorectomy guided by laparoscopy, high intensity

focused ultrasound (HIFU) has been proposed to treat small renal masses in patients with high

surgical risk [2, 3, 4]. Since the kidney is a moving target, it is important to estimate kidney

motion during the respiratory circle in order to implement appropriate motion correction

techniques.

The kidney is a richly vascularized organ that moves during respirations despite its fixations

to both the renal artery and vein and the excretory system. There is lot of variation on its

relative position with adjacent organs (liver, spleen, adrenal glands and intestines) related to

patient morphology. Under ultrasound, renal motion is visible but can only be estimated in 2D

[5]. Thus complex displacement including rotation or kidney deformation cannot be captured.

Transcutaneous ablation of a renal mass requires a very good knowledge of the kidney’s

physiologic motion during the respiration. The goal of this study is to model kidney motion

using 4D MRI acquisition combined with kidney segmentation.

Page 3 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Materials and Methods

Ten healthy volunteers were recruited in this study. The only criterion for selecting volunteers

was their ability to maintain a 10 second apnea. This study was approved by our institutional

review board approval and all participants signed a written informed consent.

MR acquisition protocol

This study was conducted on a 3T magnet (Skyra, Siemens Medical, Erlangen, Germany).

T1- weighted 3D flash acquisitions were acquired in the axial plane at 5 different phases of

the respiratory cycle (deep inspiration and expiration, intermediate inspiration/expiration, and

neutral. We used a T1 VIBE sequencing DIXON (TR 4.34 / TE 1.35 (opposed phase), FOV

308*380, Matrix 195*320, Slice thickness 3mm). The acquisition time (including both

kidneys) was less than 10 seconds, which matched the requested apnea time.

The apnea requirements were explained to the volunteers prior to testing to assure maximum

cooperation.

An elastic belt containing a load-sensitive pressure sensor was affixed to the

abdominal-low thoracic wall. A continuous signal was transmitted by the belt depending on

the pressure modeled as a curve to the MRI scan monitor. The monitoring of the respiratory

motion was recorded continuously in order to analyze retrospectively the exact phase of the

respiratory cycle for each breath-hold and 3D acquisitions (Figure 1).

Page 4 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Kidney segmentation

The data were processed with a semi-automated segmentation tool based on deformable 3D

models [6, 7 and 8] enabling a robust segmentation of kidney volume. Each kidney was

segmented with inclusion of the renal sinus and its contents. The technician initiated the

segmentation after manual positioning of several points (average 5) to outline the kidney on 2

or 3 axial slices. The software used DICOM data to obtain a volumetric axial, coronal and

sagittal 3D reconstruction of the kidneys (Figure 2). The software generated a mesh volume

via Laplacian optimization by extrapolating the contours (interfaces with signal differences)

to the other planes by iterative deformation [6]. Segmentation was corrected manually in each

plane if needed. The contact areas between the kidneys and liver/spleen or the psoas muscle

are those that generate the most mesh distortion. The volumes obtained were stored using

Visualisation Toolkit (VTK) format.

Analysis of kidney motion

2D displacement assessment

Three points were placed at the kidney barycenter (center of the renal volume), upper and

lower poles. The position of each point was defined by its coordinates x, y and z; the "X-

axis" being the lateral axis in the right-left direction, the "Y-axis" representing the antero-

posterior direction and the "Z-axis" being in the cranio-caudal direction. We measured the

linear distance between the two extreme positions of inspiration and maximum expiration,

then calculated the "curvilinear displacement" distance, which is the 3D distance traveled by a

point throughout the five respiratory stages.

Rotation assessment

To quantify the range of rotation, the principal axes were positioned in the volume

Page 5 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

reconstructed in maximum inspiration and expiration. Three vectors were traced from the

centroid of the kidney to each of the vertices of the meshes, one along the long axis of the

kidney, one along the short axis of the kidney and the latter perpendicular to the two previous

vectors. With linear algebra, we can isolate the matrix [R] from the equation [E] = [R][I],

where [T] is the rotation matrix between the reference system inserted in the kidney in

maximum expiration [E] and the one in the kidney in maximum inspiration[I].

Deformation analysis

Deformations between the different respiratory stages were computed by comparing deep

expiration and deep inspiration meshes and analyzing their differences [9].

The volumetric overlap error (VOE) is determined using the ratio of intersection and union

between two segmentations, E (deep expiration mesh) and I (deep inspiration mesh). It is

calculated as [10],

E∩I

VOE (E, I) = 1 – × 100%

E∪I

The VOE is 0% for a perfect overlap between segmentations and 100% for segmentations

without any overlap.

All the algorithms were compiled with Matlab software 2013.b (Math-Works, Natick, MA)

on an Intel Core i7-960, 3.20GHz, 12 GB RAM computer.

Data were presented as mean ± standard deviation (SD). A paired Wilcoxon (signed-rank) test

was used to compare means. P values <0.05 were taken to be statistically significant.

Statistical analyses were performed using the SSPS 21 (SPSS Inc., Chicago, IL), version

2.10.0.

Page 6 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Results

Five females and five males with an average age of 26.6 [21-34] and an average height of 173

cm [155-189] were enrolled in the study. The average weight of the participants was 66 kg

[40-90] with a BMI of 25 kg/m2

[17-26.2] (Table 2).

Acquisitions in our 10 patients were technically successful with acquisition at 5 different

respiratory phases with a continuous monitoring of respiratory motion. Overall, 20 kidneys

were included in the analysis.

Table 1 summarizes the mean (averaged of 3 directions) maximal linear and curvilinear

displacements for the three points of each kidney. The kidney centroid linear displacements

was statistically higher on the left than the right side and was respectively estimated at 31.3

±11.4 mm on the right, and 25 ±14.1mm on the left (p=0.047). Curvilinear displacements

through the five respiratory phases were estimated at 37.3 ±15.6 mm on the right and

42.4±18.8mm on the left (p>0.05). For all patients the curvilinear displacements were higher

than linear displacement (p=0.005 for all comparisons).

For both kidneys, displacement was higher in the cranio-caudal Z (28.9 ±12.2mm for the right

kidney and 23.4 ±14.7 mm for the left kidneys) than lateral X (3.7±3.3 mm for the right and

5.1 ±3.6mm for the left kidneys) and anteroposterior Y (10.4±3mm for the right and 6 ±4mm

for the left kidneys) directions (p<0.05) (Figure 3a). A trend to higher displacement in the

cranio-caudal directions was observed in the right as compared to the left kidney (p=0.058).

On the average, the rotations were estimated at X 5.9° - Y 1.3° - Z 1° on the right, and 5.2° -

2° - 3° on the left kidney (right kidney SD X 9.8°, Y 5.7 and Z 5.3° (range 31.8, 18.4 and

17.6) / left kidney SD X 9.5°, Y 3.7 and Z 5.5° (range 27.7, 12.7 and 18.7)). Rotation varies

from each subject in function of the x, y and z-axis, and Figure 3b shows that rotation can

vary considerably among subjects. Nonetheless, we can deduct, from subject six, that there is

Page 7 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

a maximum rotation of 20 degrees and an average of 3.1° rotation. While our sample is rather

small, it appears that taller subjects have bigger rotations than do shorter subjects: in our

series, e.g., volunteers under 1.65m never had such calculated displacement exceeding 40 mm,

while those over 1.85m were always higher than 40mm (for right and left kidneys). Kidney

displacement and rotation along the three axes are illustrated in Figure 3b. The average

inspiration volumes were 107cc and 114cc for right and left kidney, 108 and 116 in

expiration: the volume decrease in expiration was not significant (p=0.093 right, p=0.169

left). There was no significant variation in kidney shapes and volumes during the respiratory

cycle (Figure 4).

Page 8 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Discussion

Our study shows the complexity of kidney motion during the respiratory cycle, which

includes not only a cranio-caudal but also and antero-posterior and lateral displacement

combined with a rotation component. The last finding of this study is the absence of

deformation during the respiratory cycle. Thus we can assume the kidney is rigid enough to

avoid the use of elastic registration technique more complex to implement.

Curvilinear displacements were higher than linear displacements. These differences show that

kidney displacement is not a rectilinear displacement between the two extreme positions. The

objective was to have a better comprehension of the kidney motion before implementing

dedicated motion correction algorithm. Thus, we believe if it would be not accurate to

perform correction motion based only on two reference volumes taken in the expiratory and

inspiratory position and a basic respiratory compensation triggered by the diaphragm

excursion. This reinforces the importance of developing a real- time kidney tracking

technique.

Several authors have analyzed kidney motion due to respiration, but to our knowledge nobody

analyzes accurately the different dimensions of this complex motion, with MRI. The MRI has

the advantage to combine morphological analysis in the kidney in the three-space direction

and include volume and functional acquisitions during thermal ablation therapies [11].

Song R. et al. [12] analyzed kidney motion related to breathing in MRI. They reported a lower

kidney excursion (8.9±3.7mm) as compared to 23.4±14.7mm in our study. The amplitude

difference can be explained because theirs acquisitions were done during free breathing

whereas our acquisitions were taken in apnea in deep inspiration and expiration. Finally, in

their study Song et al, evaluate only the displacement on the cranio-caudal direction [12].

Page 9 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Although Yamashita studies were specific to the kidney centroid, he demonstrated that renal

motion was independent of age, sex, height, or body weight [13]. The absence of correlation

between renal motion and BMI was also confirmed in the study of Song et al. [12]. Abhilash

[14] observed important variations of kidney motions between subjects. However, in our

study the displacement were quite consistent between subjects as shown in figure 3a.

We observed different magnitudes of kidney motion in the three axis with greater excursion in

the cranio-caudal direction as compared to lateral and antero-posterior. This is confirmed by

the study of Tipirneni et al. [16] who only studied one point with a displacement in the three

planes of space, the most important being in the coronal plane whereas other authors have

reported similar displacement in the order of 20mm m in x, y and z axis [14, 15]. Regarding

the latter studies, Doppler ultrasound was used to evaluate kidney motion and this technique

is less reliable than MRI and do not integrate the 3D excursion of the kidney.

The analysis of kidney motion should include different portion of the kidney since we

observed in our study different magnitude of motion between our 3 points of interest in the

three space directions.

Regarding kidney deformation, previous studies suggest that the kidney is a rigid object [12,

14, 15] even if deformation had never been specifically studied by mesh comparison. There

was no significant kidney deformation during the respiratory cycle. The volume variation

observed during the respiratory cycle were within the range of software precision (4.4±1.3%)

[8].

Since there is no kidney deformation, we can conclude the displacements observed in our

study are related to a more complex movement including a rotational component that was not

previously described.

Page 10 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Regarding the comparison of displacement between the right and left kidney, we did not find

any meaningful differences: both kidneys moved independently with different amplitudes and

direction especially in the lateral axis. There was no synchronization between the kidney

motion of the right and left kidney.

As shown in our study kidney motion is not limited to the coronal plane. In order to improve

accuracy, the treatments using a 2D tracking such as extracorporeal lithotripsy must take into

account the different components of this motion. The absence of linearity in kidney

displacement and the presence of significant rotation must be considered. Since imaging-

guided treatments can last several minutes, continuous apnea cannot be achieved. Treatments

such as HIFU can last several hours and require an accurate estimation of kidney and tumor

position (margin of error <5 mm in all three axes). A correction based only on the cranio-

caudal direction or a respiratory gating is therefore not sufficient and specific algorithms to

precisely predict kidney and tumor positioned to be developed. This should include a

preoperative 4D modeling of renal motion and real time correction motion during the

respiratory cycle based on fast muliplanar 2D acquisitions that can be registered in the 4D

model.

There is several limitations in our study. First, a larger sample size would help determine if

rotation is an important factor to incorporate in the tracking method and, since our volunteers

were mostly slim, it would be beneficial to study subjects with higher body weight. Second,

we did not correlate the motion of kidney with the motion of the diaphragm because it was

not included in the volume of acquisition.

Page 11 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Conclusion

With the MRI semi-automated segmentation algorithm, we determined the positions of three

specific points in the kidney and their excursion during the respiratory cycle to analyze the 3D

motion of the kidney.

These kidney positional shifts are certainly more important in the cranio-caudal direction;

however the anterior-posterior and medio-lateral planes must not be ignored. These are

certainly more complex than a single isolated movement with a notably kidney rotational

component. Its displacement is not linear between the two extreme positions.

With regards to transcutaneous focal treatments, we cannot simply be content with respiratory

synchronization. Our study shows the need for a 4D renal targeted "tracking", which is now

under development.

Page 12 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Bibliographie

[1] Thompson RH, Atwell T, Schmit G, Lohse CM, Kurup AN, Weisbrod A, Psutka SP,

Stewart SB, Callstrom MR, Cheville JC, Boorjian SA, Leibovich BC. Comparison of partial

nephrectomy and percutaneous ablation for cT1 renal masses. Eur Urol. 2015 Feb;67(2):252-

9. doi: 10.1016/j.eururo.2014.07.021. Epub 2014 Aug 6.

[2] Hinshaw JL, Lubner MG, Ziemlewicz TJ, Lee FT Jr, Brace CL. Percutaneous tumor

ablation tools: microwave, radiofrequency, or cryoablation--what should you use and why?

Radiographics. 2014 Sep-Oct;34(5):1344-62. doi: 10.1148/rg.345140054.

[3] Luo J, Ren X, Yu T. Efficacy of extracorporeal ultrasound-guided high intensity focused

ultrasound: An evaluation based on controlled trials in China. Int J Radiat Biol. 2015

Jun;91(6):480-5. doi: 10.3109/09553002.2015.1021962. Epub 2015 Mar 28.

[4] MR-Guided Focused Ultrasound Therapy in Oncology. Med Phys. 2015 Jun;42(6):3598.

doi: 10.1118/1.4925562.

[5] Davies SC, Hill AL, Holmes RB, Halliwell M, Jackson PC. Ultrasound quantitation of

respiratory organ motion in the upper abdomen. Br J Radiol. 1994 Nov;67(803):1096-102.

[6] Chartrand, G. Cresson, T. ; Chav, R. ; Gotra, A. ; Tang, A. ; De Guise, J. SEMI-automated

liver CT segmentation using Laplacian meshes. Biomedical Imaging (ISBI), 2014 IEEE 11th

International Symposium on. April 29 2014-May 2 2014. Beijing, China.

Page 13 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

[7] Chav, R. Cresson, T. ; Chartrand, G. ; Kauffmann, C. ; Soulez, G. ; de Guise, J. A kidney

segmentation from a single prior shape in MRI. Biomedical Imaging (ISBI), 2014 IEEE 11th

International Symposium on. April 29 2014-May 2 2014. Beijing, China.

[8] Gotra A, Chartrand G, Massicotte-Tisluck K, Morin-Roy F, Vandenbroucke-Menu F, de

Guise JA, Tang A. Validation of a Semiautomated Liver Segmentation Method Using CT for

Accurate Volumetry. Acad Radiol. 2015 Apr 20. pii: S1076-6332(15)00131-2. doi:

10.1016/j.acra.2015.03.010.

[9] Extracting Euler Angles from a Rotation Matrix Mike Day, Insomniac Games /

Introduction aux transformations linéaires Algèbre linéaire I — MATH 1057 F Julien

Dompierre.

[10] Heimann T, van Ginneken B, Styner MA, et al. Comparison and evaluation of methods

for liver segmentation from CT datasets. IEEE Trans Med Imaging 2009; 28(8):1251–1265

[11] Quesson B, Laurent C, Maclair G, de Senneville BD, Mougenot C, Ries M, Carteret T,

Rullier A, Moonen CT. Real-time volumetric MRI thermometry of focused ultrasound

ablation in vivo: a feasibility study in pig liver and kidney. NMR Biomed. 2011

Feb;24(2):145-53. doi: 10.1002/nbm.1563. Epub 2010 Dec 3.

[12] R. Song, A. Tipirneni, P. Johnson, R. B. Loeffler, and C. M. Hillenbrand. Evaluation of

respiratory liver and kidney movement for MRI navigator gating. Journal of Magnetic

Resonance Imaging , vol. 33, pp. 143–148, Jan. 2011.

Page 14 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

[13] Yamashita H, Yamashita M, Futaguchi M, Takenaka R, Shibata S, Yamamoto K,

Nomoto A, Sakumi A, Kida S, Kaneko Y, Takenaka S, Shiraki T, Nakagawa K. Individually

wide range of renal motion evaluated by four-dimensional computed tomography.

Springerplus. 2014 Mar 7;3:131. doi: 10.1186/2193-1801-3-131. eCollection 2014.

[14] R. Abhilash and S. Chauhan. Respiration-induced movement correlation for synchronous

noninvasive renal cancer surgery. IEEE Trans Ultrason Ferroelectr Freq Control. 2012

Jul;59(7):1478-86.

[15] Rakkunedeth H. Abhilash and Sunita Chauhan, Respiration-Induced movement

Correlation for Synchronous Noninvasive Renal Cancer Surgery. IEEE Trans Ultrason

Ferroelectr Freq Control 2012 Jul;59(7):1478-86

[16] A. Tipirneni, R. Song, R. B. Loeffler, C. M. Hillenbrand. Evaluation of the relationship

between respiratory hepatic and renal motion using real-time MRI. Proc. Intl. Soc. Mag.

Reson. Med. 17 (2009).

[17] Peter Rogelj, Frank G. Zöllner, Stanislav Kovačič, Arvid Lundervold."Motion correction

of contrast-enhanced MRI time series of kidney". In: B. Zajc (ed.), Proceedings of the

International Electrotechnical and Computer Science Conference ERK 2007, pp.191-194.

IEEE Region 8, Slovenia section IEEE, September 2007.

Page 15 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Figure 1: respiratory cycle schematics. 1a: representation of the physiological respiratory and segmentation cycles. The first curve is a theoretical

curve.

1b: Breathing cycle during apnea. Visualizing a breathing cycle in real-time confirms the actual moment of apnea. Recording the start of the 3D sequence (red line) correlates to the respiratory cycle and enables a

retrospective gating.

125x106mm (96 x 96 DPI)

Page 16 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Figure 2: Mesh volume of right kidney enabling the coordinates of the superior, inferior and center poles. The technician initiated the segmentation after manual positioning of several points (red) to outline the

kidney on axial and coronal slices.

73x59mm (72 x 72 DPI)

Page 17 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Figure 3: Summary of kidney centroid’s linear displacement and kidney rotations. a) Red points represent the centroid displacement in the x axis, green points in the y axis and blue points in

the z axis. b) Rotation determined by the transformation of reference placed in the maximum inspiration and expiration

sequences separated in x (Rx), y (Ry) and z (Rz). 178x143mm (96 x 96 DPI)

Page 18 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Figure 3: Summary of kidney centroid’s linear displacement and kidney rotations. a) Red points represent the centroid displacement in the x axis, green points in the y axis and blue points in

the z axis.

b) Rotation determined by the transformation of reference placed in the maximum inspiration and expiration sequences separated in x (Rx), y (Ry) and z (Rz).

148x111mm (96 x 96 DPI)

Page 19 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

Figure 4: Right kidney - volunteer 2: anterior-posterior (AP) and posterior-anterior (PA) three-dimensional renderings comparing surface distance error between inspiration and expiration.

Areas in blue represent absence of error (perfect overlap) and areas in yellow represent surface distance

error (in mm). Small amounts of error are observed at the liver interface and along the inferior vena cava. AP, anterior-posterior; PA, posterior-anterior.

318x180mm (96 x 96 DPI)

Page 20 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

FOR

PEER

REVIE

W O

NLY

V

o

l

u

n

t

e

e

r

W

e

i

g

h

t

H

e

i

g

h

t

X

(lateral

)

Displa

cement

(mm)

Y

(antero

posteri

or)

Z

(cranio-

caudal

)

kg cm

Right

centroi

d

kidney

Left Right Left Right Left

1 40 153 0.3 8.2 10.1 6.9 25.8 24.2

2 72 180 1.6 1.2 7.9 3.7 21.2 15.4

3 90 185 1.9 4.1 13.8 11.9 30.6 29.8

4 82 186 4.6 4 14.4 8.2 23.2 29.4

5 70 170 7.1 2.8 12.9 10.7 25.1 24.6

6 79 185 7.4 12.4 6.6 8.3 54.5 54.2

7 45 155 1 3.6 6.4 0.8 14.4 4.5

8 56 171 1.9 6.4 11.1 6.6 33 33

9 77 189 9.7 7.8 12.7 1.8 43.3 8.2

10 49 162 1.5 0.4 8 0.7 17.9 10.4

Mean 66 �73 3.7 5.1 10.4 6.0 28.9 23.4

± SD 3.1 3.6 3.0 4.0 12.2 14.7

V

o

l

u

n

t

e

e

r

Straigh

t

displac

ement

(S)

(mm)

Curvili

gne (C)S C S C S C S C S C

TopRight

kidney

Centro

id

Botto

mTop

Left

kidney

Centro

id

Botto

m

1 24 34.9 27.7 28.0 26.5 32.7 21.9 26.2 26.4 27.5 22.2 26.4

2 20.7 24.2 22.7 23 21.6 29.1 13.5 21.9 15.9 17.4 19.3 20

3 32 42.8 33.6 37.5 36.4 41.7 37.2 47 32.3 40.2 37.5 51.7

4 29.7 61.9 27.7 47.4 30.3 64.3 36.7 56.7 30.8 51.5 31.3 57

5 24.5 29.5 29.1 33.6 28.5 38.7 29.6 38.8 26.9 28.9 35.9 39

6 58.9 74.5 55.4 71 54.9 64.2 64 73.1 56.2 57.8 52.6 61.7

7 16.6 17.5 15.7 18.5 18.8 31.3 5.2 20 5.8 19.2 4.4 22.2

8 30.8 52.9 34.9 39.8 32.1 51.1 28.8 59.9 34.2 65.8 35.9 58.5

9 49.4 57.8 46.2 53.5 46.4 59.4 13.8 82.9 11.5 75.4 13.8 72.7

10 22.9 29.7 19.7 21.2 18.8 22 11 40 10.4 40 13 43

Mean 31.0 42.6 31.3 37.3 31.4 43.5 26.2 46.7 25.0 42.4 26.6 45.2

± SD 13.3 17.6 11.4 15.6 11.2 14.6 16.4 20.3 14.1 18.8 13.8 17.2

Page 21 of 21



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

103

Conférencesenrapportaveclathèse

A. Leonardi V,Mari JL, Vidal V, DanielM. « 3D Reconstruction from CT-Scan Volume

Dataset–ApplicationtoKidneyModeling».DansSCCG2011,27thSpringconference

on ComputerGraphics, ACMPress Publisher, pp. 141-148. Viničné, Slovaquie, avril

2011.

B. LeonardiV,MariJL,VidalV,DanielM.«Reconstruction3DduVolumeRénalàpartir

d’Acquisitions Scanner Volumiques ». Dans Journées du Groupe de Travail en

ModélisationGéométrique,GTMG2011,pp.83-92.Grenoble,France,mars2011.

C. LeonardiV,Mari JL,VidalV,DanielM.«AMorphingApproachforKidneyDynamic

Modeling-From3DReconstructiontoMotionSimulation».Dans20thInternational

Conference in Central Europe on Computer Graphics, Visualization and Computer

Vision,WSCG2012,pp.179-187.Plzen,RépubliqueTchèque,juin2012.

D. LeonardiV,MariJL,SouteyrandP,FrandonJ,VidalV,DanielM.«RecalageContraint

parCartesdeCourburesDiscrètespourlaModélisationDynamiqueduRein».Dans

JournéesduGroupedeTravailenModélisationGéométrique,GTMG2013,pp.149-

158.Marseille,France,mars2013.

E. LeonardiV,DanielM,VidalV,MariJL.«Opérateursdemorphologiemathématiqueet

exploitationdescourburesdiscrètespour ladétectionde tumeursdansunmodèle

rénal ».Dans JournéesduGroupedeTravail enModélisationGéométrique,GTMG

2014,pp.28-33.Lyon,France,mars2014.

F. Leonardi V, Daniel M, Olofsson N, Souteyrand P, Vidal V, Mari JL. « Kidney 3D

DynamicModelingandTumorTrackingforTherapeuticTranscutaneousTreatment».

DansSurgetica2014.Chambéry,France,décembre2014.

Abréviations

AMU:Aix-MarseilleUniversité

CR:cryoablation

IRM:imagerieparrésonnancemagnétique

LIIE:Laboratoired’ImagerieMédicaleetInterventionnelle

RCP:RéuniondeConcertationPluridisciplinaire

RFA:ablationparradiofréquence

SM:scoresmorphométriques

TDM:scanner

ZA:zoned’ablation

104

Annexe1:classificationOMS2004destumeursrénales

Carcinomerénalfamilial

Tumeursàcellulesrénales

• BénignesAdénomepapillaireOncocytome

• MalignesCarcinomeàcellulesclairesCarcinomekystiquemultiloculaireàcellulesclairesCarcinomepapillairedureinCarcinomechromophobedureinCarcinomedestubescollecteursdeBelliniCarcinomemédullairedureinCarcinomeavectranslocationXp11CarcinomeassociéauneuroblastomeCarcinomefusiformetubuleuxetmucineuxCarcinomeinclassé

Tumeursmétanéphriques

AdénomemétanéphriqueAdénofibromemétanéphriqueTumeurstromalemétanéphrique

Tumeursmixtesépithélialesetmésenchymateuses

NéphromekystiqueTumeurmixteépithélialeetstromaleSarcomesynovial

Tumeursnéphroblastiques

RestesnéphrogéniquesNéphroblastomeNéphroblastomecystiquepartiellementdifférencié

Tumeursneuro-endocrines

TumeurcarcinoïdeCarcinomeneuroendocrineTumeurprimitiveneuroectodermiqueNeuroblastomePhéochromocytome

Autrestumeurs

TumeursmésenchymateusesTumeurshématopoïétiquesetlymphoïdesTumeursgerminalesTumeursmétastatiques

105

Annexe2:critèresdéfinissantlegradenucléairedeFuhrman

[H.Letourneux,V.Lindner,H.Lang,T.Massfelder,N.Meyer,C.Saussine,D.Jacqmin.Reproductibilitédugrade

nucléairedeFuhrman:Avantagesd’unregroupementendeuxgrades.ProgUrol2006;16:281-285]

Grades Taille du noyau Contour du Noyau

Nucléoles Cellules monstrueuses

Grade 1 10mm Régulier 0 ou

imperceptibles

0

Grade 2 15mm Discrètes

irrégularités

Visibles au

grossissement

X400

0

Grade 3 20mm Irrégularités

nettes

Visibles en

X100

0

Grade 4 20mm Irrégularités

nettes

Visibles en

X100

Cellules

monstrueuses

multilobées

106

Annexe3:traductiondelaclassificationTNM2009descarcinomesrénaux

(UICC)

T–TUMEURPRIMITIVE

• TX–Renseignementsinsuffisantspourclasserlatumeurprimitive

• T0–Pasdesignedetumeurprimitive

• T1–Tumeurintrarénale≤7cmdanssonplusgranddiamètre

T1a–tumeur≤4cm

T1b–tumeur>4cmet≤7cm

• T2–Tumeurintrarénale>7cmdanssonplusgranddiamètre,limitéeaurein

T2a–tumeur>7cmet<10cm

T2b–tumeur>10cm,limitéeaurein

• T3 – Tumeur étendue aux veines majeures ou aux tissus périrénaux mais sansenvahissement de la glande surrénale ipsilatérale ni dépassement du fascia deGérota

T3a – Tumeur macroscopiquement étendue à la veine rénale ou à sesbranches segmentaires (contenant des muscles) ou tumeur envahissant lagraissepérirénaleet/ouletissuadipeuxdusinusrénal(hilerénal)maissansdépassementdufasciadeGérota

T3b – Tumeur macroscopiquement étendue à la veine cave au-dessous dudiaphragme

T3c – Tumeur macroscopiquement étendue à la veine cave au-dessus dudiaphragmeouenvahissantlaparoidelaveinecave

• T4 – Tumeur étendue au-delà du fascia de Gérota (y compris l’extension parcontiguïtéàlaglandesurrénaleipsilatérale)

N–ADÉNOPATHIESRÉGIONALES

• NX–Renseignementsinsuffisantspourclasserl’atteintedesganglionslymphatiques

• N0–Pasd’atteintedesganglionslymphatiquesrégionaux

• N1–Atteinted’unseulganglionlymphatiquerégional

• N2–Atteintedeplusd’unganglionlymphatiquerégional

M–MÉTASTASESÀDISTANCE

• M0–Pasdemétastasesàdistance

• M1–Métastasesàdistance

107

Annexe4:notionstechniquesenradiofréquence(ablationthermiquepar

excitationmoléculaireparuncourantdeRF)

• Lecourantderadiofréquenceestuncourantsinusoïdalde400à500KHz.Lesrégions

traverséesparcecourantsubissentuneagitationionique,quiinduitparfrictionentre

lesparticulesunéchauffementtissulaire.Lebutestd'exposerlescellulestumoralesà

une température supérieure à 60° C qui provoque de façon quasi immédiate une

dénaturation cellulaire irréversible. Par contre, il n'est pas souhaitable d'atteindre

une température supérieure à 100° C qui, en provoquant une ébullition et une

carbonisationdestissus,augmenteleurrésistanceélectriqueetaltèrelespossibilités

dediffusiondu courant de radiofréquence, diminuant ainsi la taille de la lésionRF

induite.

• Les différents systèmes adaptent le temps de traitement et l'intensité de courant

délivré, soit en monitorant la température, soit en monitorant les variations de

résistancetissulaireentrel'électrodeetlesplaquesdedispersion(parexempleavec

leRF3000deBostonScientific).

• Laprocéduresedéroulesousanesthésiegénérale,lepatientendécubitusventralsur

la table de scanner (certaines équipes ne choisissent pas comme nous la voie

percutanéemaisplutôtlavoielaparoscopiqueoulalaparotomie).

• Toute la procédure est réalisée sous repérage de l’imagerie: on débute par un

uroscannerquipermetun repérage (masse, vaisseaux, voiesexcrétrices, structures

devoisinage),puisl’aiguilledeRFestpositionnée(extrémitéaucentredelatumeur)

etlesbaleinessontdéployéessouscontrôlefluoroscopique.

• Lechoixdelatailledel’aiguilledépenddelatailledelatumeur(diamètredel’aiguille

égalàceluidelatumeur,voirsupérieurde5mmpourcoaguler5mmdeparenchyme

sainquicorrespondaux«margesdesécurité»).

• Le repérage échographique utilisé par certaines équipes, et pour d’autres organes

(foie), nous semble moins adapté pour le rein: les phénomènes de cavitation

(dégagementdegaz)artefactentlerepérage.

• Dans certaines situations, on peut prévenir l’atteinte d’organe de voisinage en les

écartantdelazonetraitée(parinjectiondesérumphysiologique).

• L’intensité et la durée des paliers de RF suivent les recommandations du

constructeur.

Lapuissancedebaseestde20W,qu’onaugmenteparpalierde10Wtouteslesminutes

jusqu’à ce que l’augmentation brutale ne coupe le générateur («Roll-off»). La RF est

reprise30saprèsl’arrêt,à50%delapuissanceappliquéeauroll-offpuisaugmentéeavec

lesmêmespaliersjusqu’àobtenirun2èmeroll-off.

108

Annexe5:protocoledechauffepréconiséparleconstructeurpourune

aiguilledeLeVeende2cm.

Déployer à 2cm l’électrode LEVEEN™.

PROCEDURE FOIE ET REIN

ALGORHYTME pour électrode LEVEEN de 2cm uniquementREF 26-206, REF 26-206, REF 26-226, REF 26-227

Lancer le traitement : Régler la puissance à

30W. Augmenter de 10W par minute jusqu’à

60W maximum

Vérifier deploiement

complet

ROLL OFF ???Arrêter traitement.

Aiguille

complètement déployée ???

avant 5 minutes

Non

Phase 1 terminée : arrêter.Attendre 30 secondes. Redémarrer au même

endroit à 70% de la puissance du ROLL OFF

Continuer le traitement jusqu’au

ROLL OFF

FIN DU TRAITEMENT

Redémarrer à 50% de la puissance du

ROLL OFF

Oui Si pas de ROLL OFF après 5 minutes, augmenter la puissance de 10W par minutes jusqu’à 60W

maximum et attendre le ROLL OFF

Entre 5 et 15 minutes

Si pas de ROLL OFF après 5 minutes,

augmenter la puissance de 10W par

minutes jusqu’à 60W maximum

Non pas de ROLL OFFArrêt automatique après 15 minutes

Redémarrer immédiatement

à 60W

A Nanoengineered Embolic Agent for Precise Radiofrequency Ablation

PIERRE HENRI ROLLAND,1 JOEL L. BERRY,3 GUILLAUME LOUIS,1,2 LIONEL VELLY,1 VINCENT VIDAL,1,2

PAULINE BRIGE,1 VINUTA MAYAKONDA,3 and DAVID L. CARROLL4

1Experimental Interventional Imaging Laboratory, European Center for Medical Imaging Research, School of Medicine,Aix-Marseille University, Marseilles, France; 2Department of Radiology and Medical Imaging, La Timone Hospital, Marseilles,France; 3Department of Biomedical Engineering, University of Alabama at Birmingham, Birmingham, AL 35294, USA; and

4Center for Nanotechnology and Molecular Materials, Department of Physics, Wake Forest University, Winston-Salem,NC 27109, USA

(Received 13 November 2013; accepted 15 January 2014)

Associate Editor Agata A. Exner oversaw the review of this article.

Abstract—The purpose of the work is to investigate whetherthe electromagnetic properties of multi-walled carbon nano-tubes (MWCNT) in the presence of radiofrequency (RF)energy is (1) safe, and (2) improves the precision of thetherapeutic efficiency of the RF-ablation (RFA) procedure.An in vitro phantom was created for evaluating temperaturenear RF treated nanotubes. For the in vivo study, threebaboons and six pigs were submitted for RFA procedure insuperior/inferior kidney poles embolized with a non-adher-ent, lipophilic embolic agent (marsembol) with or withoutMWCNT. Tissue damage in the surrounding kill zone wasassayed through caspase-3 activation. The in vitro resultsshowed marked heat increase only in the region of thenanotubes. In vivo, necrosis/ischemic damage resulted fromRFA therapy alone, RFA plus marsembol only. In marsem-bol + MWCNT condition, dramatic disruption of cellmembranes and sub-cellular organelles was found whereasthe nuclear membranes and basal cell membranes remainedlargely intact. The marsembol vaporized under RFA andtissue fluid filled the space. This caused the MWCNT tocluster within the new aqueous environment. RFA plusmarsembol + MWCNT created a well-defined demarcationbetween healthy and apoptotic cells as evidenced by amarked reduction of caspase-3 expression. By contrast, therewas a much less defined ablation zone in the absence ofMWCNT. In conclusion, the combination of RFA plusmarsembol + MWCNT embolization delineated the killzone in vitro and in vivo. We demonstrate that MWCNTsremain in the ablation region thus minimizing their migrationto the systemic circulation.

Keywords—Radiofrequency, Ablation, Carbon nanotubes,

Embolic agent.

INTRODUCTION

Ablative therapies are required for small-sized can-

cers in order to overcome the deleterious effects of sys-

temic drugs.5,41However, ablative therapies kill healthy

tissue as well as diseased tissue. Radiofrequency abla-

tion (RFA) guided by computerized tomography (CT)

or ultrasound-guided percutaneous radio-frequency

probe ablation (RFA) is important in the primary

management of a multitude of cancers including renal

cell carcinomas (RCC). This treatment provides low

morbidity andminimal clinical impact on postoperative

renal function.1,4,7,8,14 In spite of this clinical success, it is

difficult to confine the to only cancerous tissue volumes.

Further, image-guided measurements are inadequate to

delineate the kill zone because vaporized tissue cannot

be visualizedwith thesemodalities.1,14,23Selective tumor

embolization prior to RFA is a viable means to define

the tumor margin.1,12

It is known that hyperthermic ablation of tissue is

enhanced in the presence of multi-walled carbon nano-

tubes (MWCNT). Previous results from our group

demonstrated complete tumor ablation in mice treated

with the combination of infrared radiation and

MWCNT.5,9,40,41 These nanoparticles combine the un-

ique physical properties of receiving antennas for elec-

tromagnetic energy and of transducers of energy into

marked heat generation adjacent MWCNT. For in-

stance, the average temperature of theMWCNT-loaded

subcutaneous tumor increased to 74 !C compared with

46 !C in control tumors exposed to NIR laser alone.5,9

Nanoparticle-based therapies currently face major

technical and safety challenges. One of the most impor-

tant considerations is the fate of these particles after they

have achieved their therapeutic action.9,15,25,27,36,38

Address correspondence to Joel L. Berry, Department of Bio-

medical Engineering, University of Alabama at Birmingham, Bir-

mingham, AL 35294, USA. Electronic mail: [email protected]

Annals of Biomedical Engineering (" 2014)

DOI: 10.1007/s10439-014-0977-9

" 2014 Biomedical Engineering Society


Nanotube length, diameter, impurities, functionaliza-

tion, and method of administration result in different

pathologies.20 In physiological environments, MWCNT

spontaneously form coarse agglomerates diminishing

their ability to transduce electromagnetic energy to heat.

Moreover, dispersion of MWCNT by injection into a

tumor mass requires keeping these nanoparticles from

exiting into healthy tissue.

Our laboratory has developed an embolic agent

(marsembol) that is a colloidal mixture of radiopaque

lipids and hydrophobic proteins designed to achieve effi-

cient dispersion of MWNCTs.42 The present study relies

upon in vitro and in vivo methods to test the hypothesis

that the killing zonemay be precisely delineated by highly

selective tumor embolisation with an embolic agent

doped with MWCNT that capture and focus the radio-

frequency (RF) energy within the embolized tissue while

minimally affecting the surrounding tissues. The in vitro

study revealed a marked increase in temperature in the

presence of RF that was confined to the region of the

nanotubes. This temperature increase required less time

and lower energies than without nanotubes. We further

report a similar result in the in vivo study that these

MWCNTs are not lost to the systemic circulation.

MATERIALS AND METHODS

Preparation of Embolic Agent: Marsembol and

Multi-Walled Carbon Nanotubes (MWNTs)

A gelatin-resorcinol mixture (37.5 g Rousselot 250

bloom PS acid process gelatin, 2.5 g resorcinol and

CaCl2 (1.25 g) in 48.75 mL water, gently stirred 1 h,

40 !C) was mixed with iodized lipiodol (Gerbet, Roissy,

France) (0.5:2, vol:vol). The lipiodol-loaded cross-

linked gelatin was stabilized by glutaraldehyde and

formaldehyde (0.9 and 15.5% in water) (2.5:0.1, vol:vol,

40 !C, 1 h); 100 mM glycine-chlorhydrate (5:1, vol:vol)

were added to neutralize unreacted aldehydes.42

Multi-walled carbon nanotubes (MWNTs) were pro-

ducedbychemical vapordeposition (CVD) ina two-stage

quartz furnace (hydrogen carrier gas). Ferrocine (2.7% in

toluene) was injected into the preheater (5 mL/h) and

furnace temperature ranged from 600 to 900 !C (1 h).

The resultingMWNTwere 50 nm in diameter and cut by

sonication to about 900 nm in length, and assessed for

purity by TEM (Phillips400).5,9,40,41 The aqueous

MWCNT suspension (3 mgMWNT vortexed/sonicated

five times in 0.1 mL water) was mixed with marsembol

(2.5 mL, stopcock-linked 2-syringes).9,15,25,27,36,38

In Vitro Model

Tissue phantoms were created from 8 g/100 mL

Knox gelatin and 4 g/100 mL sugar-free psyllium

hydrophilic mucilloid fiber added to boiling water and

thoroughly mixed until dissolved. Formalin solution

was added to the mixture and thoroughly mixed again,

poured into centrifuge tubes and cooled until firm.

Two cylindrical tissue models were created by

pouring 10 mL of the gelatin mixture into a centrifuge

tube and placed in the refrigerator. A small well was

created at the surface of the gelatin as a reservoir for

the embolic agent. For the first model, the well was

filled with 200 lL of marsembol. For the second

model, the well was filled with 200 lL of marsembol

and a MWCNT suspension. A clinical RF generator

(Radionics, model RFG-3C lesion generator) was used

as the RF energy source to heat the phantoms. RF

measurements were taken at two points equidistant

from the thermocouple. The first temperature probe

was placed within the well containing marsembol, and

the second temperature probe was placed directly into

the gelatin, outside the hollow oval region (Fig. 1). The

temperature of both regions was recorded simulta-

neously. Measurements were taken at four different

RF powers of 0.5, 1, 1.5, and 2 W and at 20-s intervals

for 120 s. The gelatin and marsembol were allowed to

return to room temperature before proceeding to the

next power.

Animal Model

The Provence Ethics Committee for Experimental

Research (#14, Marseilles, France) approved our

experimental protocol. The rationale for using swine

and baboon as experimental models is as follows: the

swine has a thin cortex but low-flow radial arteries

whereas the baboon has high-flow arcuate and radial

arteries but a very thick cortex compared to

humans.13,16,32 These animals served models for low-

flow and high-flow embolization regions that repre-

sent microcirculatory features in a renal carcinoma.

Three female baboon (Papio Anubis, 5–7 years old,

15–17 kg; Station de primatologie CNRS, 13790

Rousset, France) and six pigs (8 months old,

65 ± 3 kg; Blossin SA, 13-Aubagne, France) were

housed for a two-week acclimatizing period. Anes-

thesia was tiletamine 2.5 mg/kg, zolazepam 2.5 mg/kg

and atropine 50 lg (IM), and propofol 0.6 mg/kg/

10 mL (swine) or 2 mg/kg/10 mL (baboon) (IV) and

maintained with gaseous sevoflurane (1.2%) and

sufentanil (2.5 lg/kg/h). Euthanasia was 15 mg

midazolam + 25 mg chlorpromazine in 20 mL KCl

(15%), bolus IV injection.35

Radiological Procedures

Digital subtraction angiographic Stenoscop and

Fluorostar AEC Systems (GE-MS, Minneapolis,

ROLLAND et al.


USA) and Helix-CT-Scanner TomoscanM (Philips,

The Netherlands) were used.18,34 Aorto-renal angio-

grams (Hexabrix Guerbet-Inc., 25 mL, 12 mL/s

(swines), 10 mL, 10 mL/s (baboons) were obtained

through a 5F, 65 cm angiographic UFcatheter (Cordis,

Miami, FL, USA) from percutaneous access (Seldin-

ger, right-SFA, 10 cm-long standard 5F-vascular

sheath (Radiofocus Terumo Corporation, Tokyo, Ja-

pan) over 0.035 inch guide-wire (Radiofocus Terumo

Corporation, Tokyo, Japan). After an interlobular

artery (left-or-right kidney) was catheterized, selec-

tively, the embolic agent was delivered via a Rapid

Transit Microcatheter (Cordis, Miami, FL, USA) to

complete zonal embolization (1.7 ± 0.2 mL). Animals

were given CT-scans before/after the procedure with-

out/with injection (1.5 mL/kg, 90 mL 4 mL/s).

RF Treatment in Animals

One pair of return electrodes per lower limb was

connected to the generator (RF3000# Boston Scien-

tific). LeVeen Superslim# needles (2 cm, Boston Sci-

entific Corp.) percutaneously were advanced to the

desired renal cortex sites, and deployed under imaging

control. The RF protocol necessitated 10 W incre-

ments of RF power every minute from 20 W until the

first roll-off and from half the first roll-off wattage

until the second roll-off occurred.

In the swine (N = 6), the RFA treatment procedure

was applied to the inferior pole of the left and right

kidneys (previously embolized or not) as follows:

Control (RFA only) vs. RFA plus marsembol (N = 2),

control vs. RFA plus marsembol + MWCNT (N = 2)

and RFA plus marsembol vs. RFA plus marsem-

bol + MWCNT (N = 2). The same six-sites protocol

was applied to three baboons, but in both the superior

and inferior lobes in each kidney of each animals.

Pathological Examinations: Statistics

A standardized 4-h time was required for early

caspase-3 activation from the initial RFA procedure to

euthanasia. The kidneys were surgically removed and

perfused with buffered 10% formalin, then sectioned

through the center of the embolization zone, and re-

fixed 2 days later.45 Under radiological guidance and

macroscopic examination, serial sections were cut to

obtain 5 mm thick sagittal slices were obtained. These

slices encompassed the embolized zone and the

apparently unscathed adjacent kidney. Secondary sli-

ces 6–7 mm in thickness were cut at the center,

periphery, and outside of the ablation zone at 25, 50,

and 100% distance relative to the ablation zone.18,34,35

Fragments for transmission electron microscopy

(TEM) examination were re-fixed in glutaraldehyde

(2.5%, v/v) in 0.1 M cacodylate (pH 7.4) buffer for 1 h,

FIGURE 1. In vitro phantom study. Results of application of RF to the marsembol + MWCNT reservoir within the gelatin phantomshow a marked increase in temperature within the well. There is a steep gradient of temperature equidistant from the RF probe butoutside the region of the nanotubes.

A Nanoengineered Embolic Agent for Precise RFA


post-fixed in OsO4 2.4% prior progressive alcohol

dehydration. Semi-thin (1 lm) and ultra-thin (70 nm)

sections in epoxyresin Epon812 were cut with diamond

knifes (Leica UltracutE) and stained with uranyl ace-

tate (5% in water) and lead citrate (M in NaOH, 1 M).

TEM investigations were carried out on a Jeol JEM

1400 microscope (80 kV) and image recordings with

MegaView3 system (Olympus).

Pathologic and IHC exams were performed on 4 lm

serial thin sections in paraffin (56 !C Histowax,

Goteborg, Sweden) stained with hematoxylin–eosin–

safranin (HES) and Masson’s trichrome. The Histo-

stain# Kit System (streptavidin–alkaline phosphatase,

DAB chromogen) was used to detect active caspase-3

activity (Zymed Laboratories Inc., San Francisco, CA,

USA). Morphometric analysis was performed using an

automatic video analyzer Nikon Elipse H600L.18,34,35

Data treatment and statistical analyses (Kruskal–

Wallis non-parametric anova and Mann–Whitney U

test) were performed by using Systat12 (SPSS Inc.,

Chicago, IL). A non-parametric kernel density esti-

mator was preferred to rule out a functional form

on the distribution function curves. Results were

expressed as mean ± SD.18

RESULTS

In Vitro

Figure 1 shows the change in temperature between

gelatin and marsembol + MWCNTs using an RF

output of 0.5 W through 2 W up to 120 s. The tem-

perature increased with increasing power in all exper-

imental conditions, which can be accounted for by the

difference in RF susceptibilities of gelatin and mar-

sembol + MWCNTs. The difference in temperature

between the two temperature probes is most evident

at 2 W. Outside the region of the marsembol +

MWCNTs, it required six times longer to achieve the

same temperature of 52 !C.

In Vivo

Embolization and RF Treatments

The interventional protocol was successfully com-

pleted in each animal. Renal embolization was

achieved with marsembol, in the presence or absence of

MWCNTs (Fig. 2). Between the three experimental

groups, there was no statistical significance in energy

FIGURE 2. Imaging the RFA procedure. Digital angiograms of the renal artery at the inferior pole of the right kidney before (topleft) and after (top middle) embolization with MWCNT-loaded marsembol and after RF needle insertion (top right) in a porcinesubject. Renal CT-scans of the implanted RF needle (bottom, left and middle; porcine) and after embolization and RFA procedure(bottom left, baboon).

ROLLAND et al.


input or roll-off time/wattage (Table 1). However,

there was a marked tendency for the roll-off values to

be lower in the primates as compared to the swines

(p< 0.10). marsembol + MWCNTs tended to reduce

the RFA treatment time and intensity.

Multi-Walled Carbon Nanotubes and RF Treatments

Multi-walled carbon nanotubes (MWNTs) cluster

into dimensions ranging from 0.1 to 3.0 mm. In order

to disperse the clusters, the first-step was sonication

reducing the cluster size to a 20–50 lm network of

overlapping nanotubes. These networks were intro-

duced into the non-adhesive, lipophilic, and amor-

phous marsembol such that individualized MWCNTs

were dispersed and engulfed in the embolic agent

(Fig. 3). No differences were observed in delivering

marsembol with or without MWCNT through identi-

cal micro-catheters.

The transduction RF energy to heat by the presence

of MWCNTs suspended in marsembol was demon-

strated in separate experiments (Fig. 3). The marsem-

bol + MWCNT mixture and marsembol alone (50:50,

v:v) was delivered into a swine kidney and submitted to

a single-step RF therapy protocol. The procedure

caused marsembol + MWCNT to partially decom-

pose (vaporize) under the RF energy but not in mar-

sembol alone underscoring the efficiency of energy

transduction in the presence of MWCNTs.

Tissue Damage in the Ablation Zone: MWCNT Fate

Gross observation of RFA-induced tissue damage

was similar in the kidneys of the swine and the

baboons (Fig. 4; Table 1). However, there was a ten-

dency for an increased volume of the RFA kill zone in

the baboons, which is presumably to be accounted for

by the higher thickness of the baboon renal cortex, as

compared with the porcine renal cortex. Ablation

using RFA induces the usual cellular and subcellular

damage (necrotic shrinking of cellular components and

ischemic separation of epithelial cell layers from the

renal tubules) whether an embolic agent is present or

not.

Histological observation of ablated marsem-

bol + MWCNT revealed that cell damage showed

dramatic disruption and complete disorganization of

cell membranes, and sub-cellular organelles (Fig. 5).

Identical findings in the porcine and baboon kidneys

supported this observation. Specifically in the mar-

sembol + MWCNT zones, the tubule cell contents

were condensed, and the cell membrane was disag-

gregated and dispersed whereas the nuclear cell mem-

branes and basal membranes remained largely intact.

Using TEM observation of RFA-treated baboon

and swine nephron structures, it was revealed that the

marsembol + MWCNT which once resided within the

very distal arterioles was found to have largely

vaporized (Fig. 6). Some marsembol + MWCNT is-

lets remained. These marsembol + MWCNT residues

were characterized by a homogeneous gray color core

(the usual appearance of marsembol under TEM) and

a border of clustered structures having the size and

appearance of the original non-dispersed structure of

MWCNTs (Fig. 6). These marsembol + MWCNT

residues were detected in cases where the external and

nuclear membranes of renal tubules were preserved

only. Clusters of MWCNTs formed once the mar-

sembol embolic agent vaporized under RF treatment.

The marsembol vaporized under RFA and tissue fluid

filled the space. This caused the MWCNT to cluster

within the new aqueous environment. This clustering

behavior is typical of nanotubes changing from a

lipophilic environment to an aqueous environment.

Tissue Damage Outside the Ablation Zone

The extent to which the RFA procedure damaged

the surrounding tissue was investigated through

observation of the activation of apoptotic cell death.

Activation of caspase-3 was used as an early marker

(within 4 h) of apoptosis (Fig. 7). In all experimental

conditions, the expression of caspase-3 decreased dra-

matically in a volume of tissue extending outward to

twice the diameter of the ablation zone. Figure 7 fur-

ther illustrates that the level of RFA-induced apoptosis

varied depending on whether MWCNTs were present.

TABLE 1. Energy input and roll-off time/wattage in RFA procedures.

Baboon Swine

No embolization Roll-off 1: 5.5 ± 1.6 min/50 W Roll-off 1: 3.5 ± 1.4 min/50 W

Roll-off 2: 4.1 ± 1.6 min/40 W Roll-off 2: 2.5 ± 0.9 min/40 W

Marsembol Roll-off 1: 4.5 ± 1.5 min/50 W Roll-off 1: 3.4 ± 1.1 min/50 W


MWCNT ± marsembol Roll-off 1: 4.1 ± 1.4 min/50 W Roll-off 1: 2.1 ± 1.0 min/50 W


Kill zone volume (% kidney volume) 35.15 ± 11.28 27.28 ± 9.14



All of the cells were killed inside the RFA scar

regardless of the treatment, thus no activation of cas-

pase-3 occurred. In the control and marsembol alone

treatment groups, almost all of the cells exhibited

apoptosis outside the ablation zone. The region of

apoptosis extended outward approximately 25% be-

yond the diameter of the ablation zone. No cell death

occurred beyond this region. In contrast, the presence

of marsembol + MWCNT drastically reduced the

expression of caspase-3 in all regions surrounding the

ablation zone. Results indicate that, in the presence of

MWCNT, tissue outside the ablation was spared from

apoptotic death induction.

DISCUSSION

In Vitro

Our in vitro results demonstrated that for a volume

of 5 mL of gelatin (analogous to living tissue) sur-

rounding 2.5 mL of marsembol + MWCNT solution,

a power output of 2 W, and a distance of 0.3 cm from

the thermocouple, cell necrosis would occur at

approximately 15 s from the application of RF. Con-

versely, cell necrosis would likely occur at about 75 s in

the presence of marsembol alone. Hence, marsem-

bol + MWCNT reduces the time to cell necrosis by

80%. Since the temperature was measured at two

points equidistant from the thermocouple, (the first

point in the gelatin region, and the second point in the

MWCNT region) it can be inferred that MWCNTs

concentrate RF energy, and produce a more focused

FIGURE 3. Dispersion of MWCNTs in marsembol and decomposition of MWCNT-marsembol under RFA procedure. Scanningelectron microscopic (SEM) appearance of the newly formed aggregates of MWCNT (left), followed by transmission electronmicroscopy (TEM) appearance of the MWCNT networks dispersed in aqueous medium after moderate sonication in water (middleleft), and homogeneous dispersion of MWCNT in marsembol (middle right). Under RFA protocol applied to a mixture of marsembolalone and MWCNT-loaded marsembol, marsembol vacuolized and decomposed in the presence of MWCNT, only (left).

FIGURE 4. The dissected kidneys after RF ablation. In thebaboon (top, left), the spheroid scars of the ablation zonecontained the unscathed (middle, left), and carbonized mar-sembol + MWCNT-filled vessels (middle, right). In the swine(top, right), the spheroid scarsof the ablation zone (surroundedby white dashed lines) within the thin renal cortex (Rc) aredistinct from the unscathed surrounding tissue. The 2-syrin-ges-one stopcock devices for mixing MWCNT and marsembol.

ROLLAND et al.


temperature elevation. The distance between the two

equidistant temperature probes was 0.42 cm. A

potential source of error may have occurred during the

preparation of marsembol with MWCNTs, where the

inhomogeneous dispersion of MWCNTs may have led

to MWCNTs clumping together in certain regions,

which may have caused inaccuracies in RF measure-

ments.

FIGURE 5. Microscopic features of marsembol-embolization in renal tissue. Photonic appearance of kidney tissue embolized withMWCNT-loaded marsembol before (left, the arrows indicate the distal vascular embolization, up to the distal afferent artery, notethe amorphous nature of marsembol) and after the RFA procedure (middle; note the vacuolized marsembol in arteries and thepersistent nuclear membrane in destroyed tubular cells). TEM image (right) showing tubular cell damage in the presence ofMWCNT in marsembol showing non-biologic, homogenous dot-like residual masses of MWCNT loaded marsembol.

FIGURE 6. TEM images showing the post RFA clusterization of MWCNT. The residual MWCNT-loaded masses still containednanoparticles, but, after marsembol decomposition, the clustering of MWCNT is illustrated by the surroundings of the residues byconcentric gatherings of MWCNT and MWCNT clusters.



In Vivo

The present in vivo study reports an enhanced

ablation procedure combining RF, an embolic agent,

and nanoparticles. We show that RF-induced cell

death can be confined to the zone of embolization if

the embolic agent is doped with MWNTs. We also

show that these nanotubes remain largely confined to

the embolic region within the scarified tissue. This re-

sult has important implications for the preservation of

healthy tissue in the RFA procedure. Moreover, con-

finement of these nanoparticles to the ablation zone

may have positive implications for their use in clinical

practice.

The safety problem inherent in the clinical use of

MWCNTs is of major importance.6,30,43 These

MWCNTs are susceptible to release in the systemic

circulation and deposition in organs that cannot clear

or detoxify them and they have a striking resistance to

destruction and chemical alterations.

These MWCNTs naturally form stable macro-sized

clusters that preferentially deposit in lungs and kid-

neys.44 The likelihood of these clusters reaching these

end organs is minimized by the fact that ablated

tumors develop a scar that shrinks and retracts from

the normal parenchyma.21,23,29,31 It is logical to infer

that that this unique behavior of the ablated tumor

impedes passage of these nanotubes clusters into the

systemic circulation. Moreover, these nanotubes clus-

ters do not fulfill the characteristics of pathogenic fi-

bers whose toxicity can be predicted based on length,

biopersistence, and aspect ratio.26,28

Finally, the method of delineation of the kill zone is

enabling by deployment of the radiopaque marsembol

prior to RFA application. This is a major innovation

in RFA since real-time US and CT-scan imaging are

not possible during the ablation procedure. None of

the interventional radiologists in this study noticed any

problem in delivering the MWCNT-doped embolic

agent.

The basic requirement for using MWCNTs in RFA

clinical applications is that they formation of a deliv-

erable, stable, dispersed suspension. This is required

for the optimal transduction of electromagnetic energy

to heat.10,43,44 Dispersion of MWCNTs for biological

studies involves sonication to disaggregate the nano-

tubes clusters.36 The second step is to add dispersion

stabilizers11,43,44; such are phospholipids and surfac-

tant associated proteins.11,43 In our hands, MWCNTs

form highly stable dispersions in marsembol made of

aqueous collagen (an hydrophobic protein) with caged

oily iodinated lipiodol. The dispersed hydrophobic

MWCNTs are confined in the lipophilic embolic agent,

surrounded by the physiologic aqueous medium which

functions as an anti-solvent for the nanotubes dis-

persed in an hydrophobic colloidal medium. Upon

RFA-induced marsembol decomposition, the anti-sol-

vent precipitation of MWCNT occurred.6,22,33 The

eventual fate of all nanotubes is unknown within the

parameters of this study. However, we were able to

confirm that nanotubes and nanotube aggregates were

present within remnants of embolic agent after abla-

tion. This observation suggests that the embolic agent

restricts the movement of nanotubes (particularly by

circulatory pathways) only to the region of emboliza-

tion.

RF-ablation (RFA) induces a cascade of detrimen-

tal events. One important consequence is upregulation

of proangiogenic factors providing apoptosis resis-

tance.19 Perfusion-mediated vascular cooling limits the

efficiency of RFA by the so-called ‘‘heat sink effect’’

and embolized RFA regions are significantly larger

and more spherical, with a trend of less lesion vari-

ability compared with non-embolized RFA

regions.22,24,39

The experimental hypothesis for the use of

MWCNTs was these nanoparticles have unique elec-

trical properties which capture the RF waves to gen-

erate heating in the nano-object to affect temperature

rises in the surrounding cells.1–3,17,37 However, it re-

mains that the drastic restriction of the kill zone to the

area of embolization deserves further investigation. We

cannot exclude that there are ancillary killing mecha-

nisms contributing to the MWCNT therapeutic effects.

It has been shown that incubation of cells in the

presence of MWCNTs caused a dose- and time-

dependent increase in the formation of free radicals,

FIGURE 7. Apoptotic cell damage in the tissue surroundingthe ablation zone. Activation of caspase-3 along the apoptoticpathways is expressed as the percentage of the positivelystained cell nuclei. The extent of activation is plotted againstthe normalized distance from the ablation zone. Activation ofcaspase-3 was determined by immuno-histological chemistryin serial pathological slides of all kidneys.

ROLLAND et al.


the accumulation of peroxidative products, the loss of

cell viability, and antioxidant depletion.15,38 The

cytotoxicity caused by MWCNTs was generated

through oxidative stress and excessive generation of

oxidative species leading to the collapse of anti-oxidant

mechanisms and increased cell death.15,38 These data

demonstrate the marked cytotoxicity of carbonaceous

nanoparticulate materials is related to reactive oxygen

species (ROS) generation and the subsequent cellular

oxidative stress.25

CONCLUSION

The present study was a necessary pre-clinical study

to validate the safe and efficient improvement of

MWCNT-loaded embolization in the RFA therapy.

These results provide confidence to pursue optimization

of the RFA killing zone such that cell death is confined

only to themargins of the tumor. It is acknowledged that

this investigation was performed in healthy porcine and

primate kidneys, and it is unknown whether the results

reflect the human neoplastic conditions.

REFERENCES

1Ahmed, M., C. L. Brace, F. T. Lee, and S. N. Goldberg.Principles of and advances in percutaneous ablation.Radiology 258:351–369, 2011.2Ahmed, M., Z. J. Liu, K. S. Afzal, D. Weeks, S. M. Lobo,J. B. Kruskal, R. E. Lenkinski, and S. N. Goldberg.Radiofrequency ablation: effect of surrounding tissuecomposition on coagulation necrosis in a canine tumormodel. Radiology 230:761–767, 2004.3Ahmed, M., Z. J. Liu, A. N. Lukyanov, S. Signoretti, C.Horkan, W. L. Monsky, V. P. Torchilin, and S. N. Gold-berg. Combination radiofrequency ablation with intratu-moral liposomal doxorubicin: effect on drug accumulationand coagulation in multiple tissues and tumor types inanimals. Radiology 235:469–477, 2005.4Altunrende, F., R. Autorino, S. Hillyer, B. Yang, H.Laydner, M. A. White, R. Khanna, W. Isac, G. Spana, R.J. Stein, G. P. Haber, C. M. O’Malley, E. M. Remer, and J.H. Kaouk. Image guided percutaneous probe ablation forrenal tumors in 65 solitary kidneys: functional and onco-logical outcomes. J. Urol. 186:35–41, 2011.5Burke, A., X. F. Ding, R. Singh, R. A. Kraft, N. Levi-Polyachenko, M. N. Rylander, C. Szot, C. Buchanan, J.Whitney, J. Fisher, H. C. Hatcher, R. D’Agostino, N. D.Kock, P. M. Ajayan, D. L. Carroll, S. Akman, F. M. Torti,and S. V. Torti. Long-term survival following a singletreatment of kidney tumors with multiwalled carbonnanotubes and near-infrared radiation. Proc. Natl. Acad.Sci. U.S.A. 106:12897–12902, 2009.6Cheng, X., J. Zhong, J. Meng, M. Yang, F. Jia, Z. Xu, H.Kong, and H. Xu. Characterization of multiwalled carbonnanotubes dispersing in water and association with bio-logical effects. J. Nanomater. 2011:1, 2011.

7Cooperberg, M. R., C. J. Kane, K. Mallin, and P. R.Carroll. National trends in treatment of stage I renal cellcarcinoma. J. Urol. 181:319, 2009.8Cornelis, F., X. Buy, M. Andre, R. Oyen, J. Bouffard-Vercelli, A. Blandino, J. Auriol, J. M. Correas, A. Pluvi-nage, S. Freeman, S. B. Solomon, and N. Grenier. De novorenal tumors arising in kidney transplants: midterm out-come after percutaneous thermal ablation. Radiology260:900–907, 2011.9Ding, X. F., R. Singh, A. Burke, H. Hatcher, J. Olson, R.A. Kraft, M. Schmid, D. Carroll, J. D. Bourland, S. Ak-man, F. M. Torti, and S. V. Torti. Development of iron-containing multiwalled carbon nanotubes for MR-guidedlaser-induced thermotherapy. Nanomedicine 6:1341–1352,2011.

10Fraczek-Szczypta, A., E. Menaszek, and S. Blazewicz.Some observations on carbon nanotubes susceptibility tocell phagocytosis. J. Nanomater. 2011:473516, 2011.

11Gasser, M., B. Rothen-Rutishauser, H. F. Krug, P. Gehr,M. Nelle, B. Yan, and P. Wick. The adsorption of bio-molecules to multi-walled carbon nanotubes is influencedby both pulmonary surfactant lipids and surface chemistry.J. Nanobiotechnol. 8:31, 2010.

12Gebauer, B., M. Werk, E. Lopez-Haenninen, R. Felix, andP. Althaus. Radiofrequency ablation in combination withembolization in metachronous recurrent renal cancer insolitary kidney after contralateral tumor nephrectomy.Cardiovasc. Intervent. Radiol. 30:644–649, 2007.

13Gill, I. S., T. H. S. Hsu, R. L. Fox, A. Matamoros, C. D.Miller, R. F. LeVeen, M. T. Grune, G. T. Sung, and M. E.Fidler. Laparoscopic and percutaneous radiofrequencyablation of the kidney: acute and chronic porcine study.Urology 56:197–200, 2000.

14Kang, D. C., D. A. Palmer, M. Zarei, P. Shah, C. Folsom,R. J. Beyth, T. L. Stoffs, M. M. Neuberger, and P. Dahm.A systematic review of the quality of evidence of ablativetherapy for small renal masses. J. Urol. 187:44–47, 2012.

15Krug, H. F., and P. Wick. Nanotoxicology: an interdisci-plinary challenge. Angew. Chem. Int. Ed. Engl. 50:1260–1278, 2011.

16Lavoie, P., P. Metellus, L. Velly, V. Vidal, P. H. Rolland,M. Mekaouche, G. Dubreuil, and O. Levrier. Functionalcerebral venous outflow in swine and baboon: feasibility ofan intracranial venous hypertension model. J. Invest. Surg.21:323–329, 2008.

17Lee, S., and J. W. Peng. Effect of plasma treatment onelectrical conductivity and Raman spectra of carbonnanotubes. J. Phys. Chem. Solids 72:1101–1103, 2011.

18Levrier, O., C. Mekkaoui, P. H. Rolland, K. Murphy, P.Cabrol, G. Moulin, J. M. Bartoli, and C. Raybaud. Effi-cacy and low vascular toxicity of embolization with radicalversus anionic polymerization of n-butyl-2-cyanoacrylate(NBCA)—an experimental study in the swine. J. Neurora-diol. 30:95–102, 2003.

19Lewandowski, R. J., J. F. Geschwind, E. Liapi, and R.Salem. Transcatheter intraarterial therapies: rationale andoverview. Radiology 259:641–657, 2011.

20Madani, S. Y., A. Mandel, and A. M. Seifalian. A concisereview of carbon nanotube’s toxicology. Nano Rev. 4:1–14,2013.

21Margulis, V., E. D. Matsumoto, G. Lindberg, L. Tunc, G.Taylor, A. I. Sagalowsky, and J. A. Cadeddu. Acute his-tologic effects on temperature-based radiofrequency abla-tion on renal tumor pathologic interpretation. Urology64:660–663, 2004.



22Matlaga, B. R., R. J. Zagoria, R. D. Woodruff, F. M.Torti, and M. C. Hall. Phase II trial of radio frequencyablation of renal cancer: evaluation of the kill zone. J. Urol.168:2401–2405, 2002.

23Matsumoto, E. D., L. Watumull, D. B. Johnson, K. Ogan,G. D. Taylor, S. Josephs, and J. A. Cadeddu. The radio-graphic evolution of radio frequency ablated renal tumors.J. Urol. 172:45–48, 2004.

24Michaels, M. J., H. K. Rhee, A. P. Mourtzinos, I. C.Summerhayes, M. L. Silverman, and J. A. Libertino.Incomplete renal tumor destruction using radio frequencyinterstitial ablation. J. Urol. 168:2406–2409, 2002.

25Murr, L. E., K. M. Garza, K. F. Soto, A. Carrasco, T. G.Powell, D. A. Ramirez, P. A. Guerrero, D. A. Lopez, and J.Venzor. Cytotoxicity assessment of some carbon nanotubesand related carbon nanoparticle aggregates and the impli-cations for anthropogenic carbon nanotube aggregates inthe environment. Int. J. Environ. Res. Public Health 2:31–42, 2005.

26Osmond-McLeod, M. J., C. A. Poland, F. Murphy, L.Waddington, H. Morris, S. C. Hawkins, S. Clark, R. Ait-ken, M. J. McCall, and K. Donaldson. Durability and in-flammogenic impact of carbon nanotubes compared withasbestos fibres. Part. Fibre Toxicol. 8:15, 2011.

27Pacurari, M., V. Castranova, and V. Vallyathan. Single-and multi-wall carbon nanotubes versus asbestos: are thecarbon nanotubes a new health risk to humans? J. Toxicol.Environ. Health A 73:378–395, 2010.

28Pacurari, M., Y. Qian, D. Porter, M. Wolfarth, Y. Wan, D.Luo, M. Ding, V. Castranova, and N. Guo. Multi-walledcarbon nanotube-induced gene expression in the mouselung: association with lung pathology. Toxicol. Appl.Pharmacol. 255:18–31, 2011.

29Polascik, T. J. Ablation of renal tumors in a rabbit modelwith interstitial saline-augmented radiofrequency en-ergy—reply by the authors. Urology 54:382–383, 1999.

30Ravichandran, P., S. Baluchamy, B. Sadanandan, R.Gopikrishnan, S. Biradar, V. Ramesh, J. C. Hall, and G. T.Ramesh. Multiwalled carbon nanotubes activate NF-kap-pa B and AP-1 signaling pathways to induce apoptosis inrat lung epithelial cells. Apoptosis 15:1507–1516, 2010.

31Rendon, R. A., M. R. Gertner, M. D. Sherar, M. R. Asch,J. R. Kachura, J. Sweet, and M. A. S. Jewett. Developmentof a radiofrequency based thermal therapy technique in anin vivo porcine model for the treatment of small renalmasses. J. Urol. 166:292–298, 2001.

32Rendon, R. A., J. R. Kachura, J. M. Sweet, M. R. Gertner,M. D. Sherar, M. Robinette, J. Tsihlias, J. Trachtenberg,H. Sampson, and M. A. S. Jewett. The uncertainty of radiofrequency treatment of renal cell carcinoma: findings atimmediate and delayed nephrectomy. J. Urol. 167:1587–1592, 2002.

33Rodemann, H. P., A. Binder, A. Burger, N. Guven, H.Loffler, and M. Bamberg. The underlying cellular mecha-nism of fibrosis. Kidney Int. Suppl. 54:S32–S36, 1996.

34Rolland, P. H., A. B. Charifi, C. Verrier, H. Bodard, A.Friggi, P. Piquet, G. Moulin, and J. M. Bartoli. Hemody-namics and wall mechanics after stent placement in swine

iliac arteries: comparative results from six stent designs.Radiology 213:229–246, 1999.

35Rolland, P. H., V. Vidal, C. Mekkaoui, M. F. Bertrand, O.Levrier, and J. M. Bartoli. Embolization-driven occlusionof the abdominal aortic aneurysmal sac as the basis ofprevention of endoleaks in a new swine model. Eur. J. Vasc.Endovasc. Surg. 31:28–35, 2006.

36Savolainen, K., L. Pylkkanen, H. Norppa, G. Falck, H.Lindberg, T. Tuomi, M. Vippola, H. Alenius, K. Hameri, J.Koivisto, D. Brouwer, D. Mark, D. Bard, M. Berges, E.Jankowska, M. Posniak, P. Farmer, R. Singh, F. Krom-bach, P. Bihari, G. Kasper, and M. Seipenbusch. Nano-technologies, engineered nanomaterials and occupationalhealth and safety—a review. Saf. Sci. 48:957–963, 2010.

37Shuba, M. V., G. Y. Slepyan, S. A. Maksimenko, and G.W. Hanson. Radiofrequency field absorption by carbonnanotubes embedded in a conductive host. J. Appl. Phys.108:114302, 2010.

38Sohaebuddin, S. K., P. T. Thevenot, D. Baker, J. W. Eaton,and L. Tang.Nanomaterial cytotoxicity is composition, size,and cell type dependent. Part. Fibre Toxicol. 7:22, 2010.

39Sommer, C., N. Kortes, S. Zelzer, F. Arnegger, U. Stampfl,N. Bellemann, T. Gehrig, F. Nickel, H. Kenngott, C.Mogler, T. Longerich, H. Meinzer, G. Richter, H. Kauc-zor, and B. Radeleff. Renal artery embolization combinedwith radiofrequency ablation in a porcine kidney model:effect of small and narrowly calibrated microparticles asembolization material on coagulation diameter, volume,and shape. Cardiovasc. Intervent. Radiol. 34:156–165, 2011.

40Terrones, M., P. M. Ajayan, F. Banhart, X. Blase, D. L.Carroll, J. C. Charlier, R. Czerw, B. Foley, N. Grobert, R.Kamalakaran, P. Kohler-Redlich, M. Ruhle, T. Seeger,and H. Terrones. N-doping and coalescence of carbonnanotubes: synthesis and electronic properties. Appl. Phys.A 74:355–361, 2002.

41Torti, S. V., F. Byrne, O. Whelan, N. Levi, B. Ucer, M.Schmid, F. M. Torti, S. Akman, J. Liu, P. M. Ajayan, O.Nalamasu, and D. L. Carroll. Thermal ablation therapeu-tics based on CN(x) multi-walled nanotubes. Int. J. Nan-omed. 2:707–714, 2007.

42Vidal, V., P. H. Rolland, L. Daniel, G. Moulin, J. M.Bartoli, and O. Levrier. Effectiveness of endovascularembolization with a collagen-based embolic agent (mar-sembol) in an animal model. J. Vasc. Interv. Radiol.21:1419–1423, 2010.

43Wang, L., V. Castranova, A. Mishra, B. Chen, R. R.Mercer, D. Schwegler-Berry, and Y. Rojanasakul. Disper-sion of single-walled carbon nanotubes by a natural lungsurfactant for pulmonary in vitro and in vivo toxicitystudies. Part. Fibre Toxicol. 7:31, 2010.

44Wirnitzer, U., B. Herbold, M. Voetz, and J. Ragot. Studieson the in vitro genotoxicity of baytubes (R), agglomeratesof engineered multi-walled carbon-nanotubes (MWCNT).Toxicol. Lett. 186:160–165, 2009.

45Yang, C., V. Kaushal, S. V. Shah, and G. P. Kaushal.Autophagy is associated with apoptosis in cisplatin injuryto renal tubular epithelial cells. Am. J. Physiol. RenalPhysiol. 294:F777–F787, 2008.

ROLLAND et al.


The Visual Computer manuscript No.(will be inserted by the editor)

Multiple reconstruction and dynamic modeling of 3D digital objectsusing a morphing approach

Application to kidney animation and tumor tracking

Valentin Leonardi · Vincent Vidal · Marc Daniel · Jean-Luc Mari

Received: date / Accepted: date

Abstract Organ segmentation and motion simulation of or-

gans can be useful for many clinical purposes such as or-

gan study, diagnostic aid, therapy planning or even tumor

destruction. In this paper we present a full workflow start-

ing from a CT-Scan resulting in kidney motion simulation

and tumor tracking. Our method is divided into three ma-

jor steps: kidney segmentation, surface reconstruction and

animation. The segmentation is based on a semi-automatic

region-growing approach that is refined to improve its re-

sults. The reconstruction is performed using the Poisson sur-

face reconstruction and gives a manifold 3-dimensional mod-

el of the kidney. Finally, the animation is accomplished by

using an automatic mesh morphing among the models pre-

viously obtained. Thus, the results are purely geometric be-

cause they are 3-dimensional (3D) animated models. More-

over, our method requires only a basic user interaction and

is fast enough to be used in a medical environment, which

satisfies our constraints. Finally, this method can be eas-

ily adapted to Magnetic Resonance Imaging (MRI) acquisi-

tion because only the segmentation part would require minor

modifications.

Keywords Geometric modeling · surface reconstruction ·

dynamic modeling · mesh morphing

Valentin Leonardi, Marc Daniel, Jean-Luc Mari

LSIS, UMR CNRS 7296

Aix-Marseille Universite

Vincent Vidal

LIIE, EA 4264

CERIMED, Aix-Marseille Universite

E-mail: [email protected]




1 Introduction

Tumors can be treated by low-invasive approaches. The goal

is to minimize the interactions between the surrounding en-

vironment and the patient to limit the consequences of sur-

gery (incision treatment, convalescence) and their possible

complications (nosocomial infections). Kidney tumors can

be treated by radiofrequency [50]. Radiofrequency is a low-

invasive, non-surgical percutaneous heat treatment. The prin-

ciple is to locate the tumor through a CT-Scan, and to insert

a radiofrequency electrode in its center. An electric current

is then delivered, to destroy the tumor. However, there is a

chance of cancerous cell displacement when removing the

electrode.

The KiTT project (Kidney Tumor Tracking, of which

we are a part) is fully involved in this low-invasive proto-

col. Its goal is to create a totally non-invasive new approach

by transmitting radiofrequency waves using transcutaneous

method until the tumor is eradicated. The main difficulty

is keeping the wave beam continuously focused on the tu-

mor while the kidney is deformed and moves because of

the respiratory cycle. A kidney (and tumor) tracking model

is therefore necessary. Before the organ-tracking stage, we

must obtain a solid 3D model of the organ.

Thus, we present an entire workflow that aims to track

a kidney tumor and to simulate the deformation of the or-

gan from three sets of medical acquisitions. Each of these

sets is obtained for a precise breathing phase: one acquisi-

tion for the exhale phase, another acquisition for the inhale

phase and the third acquisition for the middle phase of the

respiratory cycle. For the remainder of this paper, this phase

will be referred to as the mid-cycle phase. The method pre-

sented here is divided into three major steps: the first step is

the kidney segmentation for every slice of the three acquisi-

2 Valentin Leonardi et al.

tions. Three point clouds are issued in this first step. The sec-

ond step is the reconstruction to produce three manifold 3D

models that represent the same kidney for the three differ-

ent breathing phases. We call these models M1 (the kidney

model for the inhale phase), M2 (the kidney model for the

mid-cycle phase) and M3 (the kidney model for the exhale

phase). Finally, the last step is a soft transition among the

three models to simulate the organ movements and defor-

mation. This transition is accomplished by mesh morphing

[32] between M1 and M2 and between M2 and M3.

Section 2 of this article addresses previous work on seg-

mentation and organ tracking. Section 3 introduces the whole

process of kidney tracking from three medical acquisitions

in which each step is detailed. In section 4, the results are

presented and their performances are described. Finally in

section 5, we discuss the limits of our method and perspec-

tives on how to overcome them.

The present paper is an augmented and enhanced version

of our previous study, which is described in [38].

2 Related work

2.1 Segmentation

Clustering methods aim to assign pixels to homogeneous

subsets (or clusters). The main difficulty is the cluster def-

inition (what are the conditions necessary for a pixel to be-

long to a cluster) and the number of clusters. These param-

eters can be set manually [1], [51], [30] or automatically

[37], [39]. There are several approaches for the segmen-

tation itself: a Bayes classifier [37], [39], the Expectation-

Maximization algorithm [66], Maximization A Posteriori [55],

or K-Means (or Fuzzy K-Means) [1], [51].

Strictly speaking, Markov Random Fields are not a

segmentation approach. They are used with other segmen-

tation methods to improve their results. The principle is to

model the spatial interactions between a pixel and its neigh-

borhood. Markov Random Fields are generally used with

clustering methods [29], [25], [72].

Artificial Neural Networks are not often used for seg-

mentation. Usually, they are employed as clustering meth-

ods [36], [42]. The advantage of this approach is its learn-

ing capacity through performed tasks. The more a network

is used for segmentation, the better the results. Moreover,

it is also possible to parallelize methods by using an artifi-

cial neural network and, therefore, to accelerate the compu-

tations [64].

Deformable models are widely used for segmentation.

They constitute positioning a curve next to the object to seg-

ment and then deforming the curve to make it fit the contour.

Similar to clustering methods, the main difficulty here is the

initialization. Two major approaches come from the litera-

ture: the first approach is to segment the object coarsely to

detect its contour and to fit the model on it [6], [7], [12], [21].

The second approach requires a set of deformable models

that have already been used to segment the desired object.

The model initialization is then obtained by the calculation

of a mean deformable model [10], [18], [22].

Finally, region growing approaches are often used in

medical imaging, the principle of which is to place a point

(seed) inside the object to segment (e.g., organ, tumor, bone).

This seed defines the first pixel in the region. The region

then becomes iteratively larger by adding to it the surround-

ing pixels according to a given homogeneity criterion. The

first difficulty of this approach is to define a robust seed.

There are several approaches to defining the seed automat-

ically. In [41], a seed is considered robust if the difference

between the highest and lowest gray values within its neigh-

borhood does not exceed a given threshold. Wu et al. [68]

define a Region of Interest (ROI) around the organ to seg-

ment. Then, they apply a function to every pixel in the ROI

based on the spatial and feature space distances between the

pixel, its neighborhood and the ROI contour. The seed will

be the pixel for which the function is a minimum. Finally,

Rusko et al. [59] first binarize the image and then define the

seed by successive erosion of the largest connected compo-

nent.

2.2 Organ reconstruction

Graphs can be used for organ reconstruction in order to rep-

resent interactions between two segments (the contour being

defined by several segments) close to each other but in dif-

ferent plans. The triangulation, i.e. the final surface, can ei-

ther be obtained directly by a certain path with lowest cost

[19] or indirectly by relying on the graph during the recon-

struction step in order to consider spatial interactions [27].

Slice to slice triangulation is a method widely used for

organ reconstruction. The main difficulty here is to connect

coherently two slices with different contours (abrupt shape

change, or when one contour has to be connected to several

ones). Two types of approach exist to handle this problem.

The first one is to create one or more intermediate contours

between two given slices [17,33]. The second one relies on

adding or removing points on the contours [67,9,11].

Multiple reconstruction and dynamic modeling of 3D digital objects using a morphing approach 3

Implicit surfaces are also widely used in this domain.

Implicit surfaces generated by skeleton are the most found.

It is possible to generate an implicit surface from 2D [63,

52] or 3D skeletons [4]. It is also possible to extract the sur-

face from a decomposition of the 2D contours into trape-

zoids [20].

Other approaches also exist for organ reconstruction

from 2D planar contours. Xu and Lu [69] choose to repre-

sent the surface mathematical formulation with partial dif-

ferential equations. The final surface is obtained be solv-

ing the equations through the finite difference method. Cotin

et al. [13] use simplex meshes [16] for liver reconstruction.

Simplex meshes can be viewed as a 3D active contour exten-

sion, with the particularity of a higher point density in the

high curvature zones. Mari [44] uses a multi-layered model

which considers the global shape aspect of the organ as well

as local specificities of the surface. This way, the cardiac

muscle is efficiently reconstructed.

Many reconstruction methods from a point cloud can

be applied for organ reconstruction from 2D contours. In-

deed, we just have to consider the whole set of contours in-

stead of one independently from the others. Note that the

Marching Cubes algorithm [43] is a reference in medical

volume visualization. It is still used although improvements

are sometimes necessary [2,24,65]. For a full review on 3D

reconstruction methods, the reader can refer to [40].

2.3 Tracking

Organ tracking methods are based either on mathematical

models that represent the respiratory cycle as a periodic func-

tion or on heuristics that predict future movements from the

observation and analysis of previous movements.

The most intuitive way to track an organ is to place a

marker that is highly detectable by a classic medical imag-

ing acquisition near this organ [46], [47], [49], [61], [62].

This formalism is also used in all-in-one robotic radiosurgery

systems such as the Cyberknife [45]. This type of method

requires a surgical intervention which is not suitable in our

case.

The following approaches assume that the kidney has

been segmented and reconstructed previously for two or more

phases of the respiratory cycle. Usually, only two models are

needed, but three [60] or even more [57] are sometimes nec-

essary. These extra acquisitions can be used to improve the

precision of the organ deformation. In other cases, it is not an

extra acquisition of the organ that is needed but other types

of data that are essential to the method. Hostettler et al. [26]

use the diaphragm movement to reflect it on the abdominal

organs. In [60], air, tissue and lungs must be segmented for

three acquisitions in order to obtain the organ tracking.

Deformation fields are used to understand the motion of

an organ. These fields compute the deformations that must

be applied to a given source model Ms to deform it into a

given target model Mt . The deformation field can be com-

puted using several methods such as Maximum Likelihood

/ Expectation-Maximization [56], least squares [60] or ap-

proaches based on Normalized Mutual Information [58]. De-

formations can also be applied on a mesh through a de-

formable superquadratic to obtain the movement of an organ

[8]. Additionally, deformation fields can be found through

registration methods which can be used for organ tracking.

Nicolau et al. [47] use two acquisitions: on the first acquisi-

tion, markers are used to obtain the position of the organ of

interest. Then, a second acquisition is obtained without these

markers. By analyzing the differences in the position of the

spine for both acquisitions, the registration is performed us-

ing the minimization of the Extended Projective Point Cri-

terion. In [58] two operations are performed to compute the

registration: affine transformation is used for global move-

ments, while Free-Form Deformation is used for local move-

ments. Two registration algorithms based on the optical flow

are implemented and accelerated using GPU programming

in [48] to perform image-guided radiotherapy.

Finally, as well as for the segmentation, probabilistic

approaches can also be used for organ tracking. The most

common way is to build up a database in order to study the

desired organ motion for the entire respiratory cycle through

surrogate markers. In both [5] and [54] this information is

reflected on a model of the liver by capturing the breathing

signal. A specific transformation is applied to the 3D volume

according to the respiratory phase which leads to the organ

tracking.

3 Dynamic modeling process

3.1 Overview

An overview of our entire workflow can be seen in Figure 1.

This figure shows how to obtain the kidney motion simula-

tion from 3 sets of CT-Scans or MRI via the 3 major steps

detailed in the following subsections. More details of this

overview can also be found in Algorithm 1 as pseudo-code.


INPUT : Three CT-Scans acquisitions for the inhale (I1),

mid-cycle (I2) and exhale phases (I3).

For I1, I2, I3 do

2D Point : Px,y

Px,y = Coordinates of a manually defined seed for the middle

slice of Ii

Region : Ri

Ri = RegionGrowing(Px,y)

[Region-growing approach is initialized by the seed whose

location is given by Px,y]

[Pseudo-code for this function can be found in section 3.2.1]

Region : R0i

R0i = SegmentationRefinement(Ri)

[Refinement of the region Ri]

[Pseudo-code for this function can be found in section 3.2.2]

3D Model : mi

mi = PoissonSurfaceReconstruction(R0i)

[3D reconstruction of R0i. At this stage, R0

i is a point cloud]

[Pseudo-code for this function can be found in section 3.3]

end For

Metamesh : M1,2,M2,3

M1,2 = MeshMorphing(m1,m2)

M2,3 = MeshMorphing(m2,m3)

[Compute the mesh morphing between mi and m j]

[Pseudo-code for this function can be found in section 3.4]

Breathing phase : BP

If (Inhale Phase BP AND BP Mid-cycle phase) then

Display M1,2 for BP

else

Display M2,3 for BP

end If

OUTPUT : Dynamic model of a kidney for any breathing

phase

Algorithm 1: Pseudo-code for our entire workflow from

CT-Scan to dynamic kidney model

3.2 Kidney segmentation

3.2.1 First segmentation

Among the existing approaches for segmentation, we want

to use an approach that does not require any learning dataset.

Building such a set would be time- and memory-consuming.

The execution time must be compatible with a medical real-

time (or semi real-time) environment and should not exceed

one minute. The initialization must also be compatible with

a real-time constraint. A method that can be initialized au-

Fig. 1: Overview of our entire workflow: three sets of images

that result from medical imaging acquisition for the inhale,

mid-cycle and exhale phases (first line). The kidney and the

tumor are segmented for every image of these three acqui-

sitions (second line). The Poisson surface reconstruction is

then applied to the point cloud extracted from the segmen-

tation of each of the three different phases. We call the re-

sulting models M1, M2 and M3 (third line). Mesh morphing

is computed between M1 and M2 and between M2 and M3.

The results are two metameshes that allow a smooth tran-

sition from M1 to M2 and from M2 to M3 (fourth line). By

alternating the two metameshes, a full and smooth transi-

tion from M1 to M3 is possible, which results in the kidney

motion visualization (fifth line)

tomatically or manually (in this case, it should require only

a few interactions) is necessary. Thus, the approach that is

chosen is the region-growing approach. This approach does

not require a learning dataset, and the runtime is fast be-


cause such methods are usually based on recursive func-

tions. Finally, the initialization which comprises defining a

point, can be performed automatically or manually (the time

needed for defining only one point is acceptable).

In our case, we use a region-growing approach that is

manually initialized; the user must define the seed with a

mouse. Only one click is necessary to segment all of the

kidney regions that are present in each image of the set. Let

I1, I2, I3, ..., IN be the images to segment (we assume that the

kidney is present in every one of them) and let I1 and IN be

both ends of the kidney. The seed is manually defined for

image IN2

and then automatically propagated to the adjacent

images (IN2 −1

et IN2 +1

) (see Figure 2.a). The propagation is

accomplished by considering the weighted barycenter of the

kidney contour on the previous image as the seed for the

current image (Figure 2.b & 2.c). The segmentation is per-

formed first for images IN2

to IN and then for images IN2

to I1.

Fig. 2: (a, left) Initial seed propagation. – (b, above right)

Kidney contour for image Ik. – (c, below right) A new seed

for image Ik+1 is calculated from the kidney contour in im-

age Ik

For each slice of the acquisition, the growing region me-

thod is the same and is detailed below. The growing is lim-

ited to an ROI which size is 30% of the image width and

is centered on the seed. This size is defined by several ex-

perimentations and is adequate to contain the whole kidney.

First, to homogenize the gray values and to reduce the pres-

ence of noise (inherent in medical images), we apply a Gaus-

sian blur of size 1. Then we evaluate the mean gray value

in a 5-pixel radius around the seed (we do not include the

extreme values). Let meanseed be this value. For each new

image Ix to segment, we know the mean gray value in every

kidney region in the previous images segmented so far. Let

meankid be this value. For each pixel p of gray value Gvp in-

side the ROI and next to the region border, we calculate the

distance between p and the seed. Using this distance we can

set a threshold thres (see Figure 3). Next, p is added to the

region if |Gvp−meanseed |< thres and |Gvp−meankid |< 30.

Thus, the region grows iteratively until no pixel can be added

anymore.

Despite the Gaussian blur, noise is still present in the

image, which leads to non-satisfactory results (Figure 4.a).

A succession of mathematical morphology operations al-

lows us to overcome this problem. The purpose of this post-

treatment is to fill holes and to eliminate the parts where

the region overflows into a second organ next to the kidney,

for which the grayscale is slightly different. To do this, we

first perform a morphological closing of size 2 followed by

an opening of size 4. Both sizes appear to be the most effi-

cient after several experimentations. Note that it is essential

to first perform the closing because, at this stage, the kidney

binary volume is composed of several small connected com-

ponents; an opening would suppress them. We can see that

the holes are filled correctly (Figure 4.b), but that the most

important overflowings are still present (Figure 4.c).

Fig. 3: The distance between the current pixel and the seed

defines the threshold. The larger the distance, the stricter the

condition that a pixel should meet to be considered a part of

the region


Fig. 4: (a, left) Region growing without post-treatment. Contours that should not be considered appear because of noise. –

(b, middle) Post-treatment makes wrong contours disappear. – (c, right) Post-treatment does not allow us to suppress areas

that have important overflowing

Function RegionGrowing(2D Point P) :

For all neighbours of P, p do

If (p meets a homogeneity criterion) then

Add p to the final region

RegionGrowing(p)

end Ifend For

End

Algorithm 2: Pseudo-code of the region growing ap-

proach

3.2.2 Segmentation refinement

To improve the results and to avoid adjacent organs being

considered part of the kidney, the cumulative histogram H

for the kidney regions is computed. This histogram repre-

sents the number of appearances of each gray value in the

segmented kidney region in all of the images I1, ..., IN . H

shows a peak because the kidney has the same grayscale

range [min;max] in all of the images (Figure 5.a). The re-

finement is performed as follows. The same ROI as in 3.2.1

is defined, and we apply a Gaussian blur of size 1 to homog-

enize the gray values. Let appmax be the maximal number

of appearances of all gray values in H, and let H(Gv) be

the function that returns the number of appearances in H

for a given gray value Gv. For each pixel p inside the ROI,

the distance dist between p and the center of the ROI sets a

threshold thres according to the following formula:

thres =dist − sizeROI

4

100(1)

This expression of thres is another way to define a con-

dition which becomes stricter as the considered pixel is far

from the center of the ROI. p is considered part of the kidney

if H(Gvp) ≥ appmax(0.85+ thres). Literally, this approach

defines an acceptable threshold for the number of appear-

ances, based on optimal observed results for the set of im-

ages we processed. Gray values for which the number of

appearances in H is below this threshold are no longer con-

sidered part of the kidney. In this way, the grayscale range

[min;max] is reduced to the new range [mina f f ;maxa f f ] (Fig-

ure 5.b), which allow the less frequent gray values to be

eliminated. The less frequent gray value zone represents the

part where the region overflows into an adjacent organ (Fig-

ure 5.c & 5.d). Finally, for the same reason as in 3.2.1, we

perform a post-treatment based on mathematical morphol-

ogy operators. First, we apply a closing followed by small

particle elimination. Another closing is then applied to con-

nect the remaining connected components. Finally, a hole-

filling algorithm fills the final kidney.

Function SegmentationRefinement(Region R) :

For All pixels p of R do

If (The gray value of p is present enough in R) then

Keep p in the final region

end Ifend For

End

Algorithm 3: Pseudo-code for the segmentation refine-

ment method

3.2.3 Comparison

Some of the previous work in abdominal organ segmenta-

tion are closely related to ours: they are all recent region

growing approaches dedicated to detect borders in CT-Scan

or MRI images. Thus, a comparision between these methods

and our is relevant and presented in Table 1. It is based on


Fig. 5: (a, above left). Cumulative histogram. The number

of appearances of each gray value in the kidney region is

shown. The grayscale is located between [min;max] – (b be-

low left). A threshold is set to reduce the kidney grayscale

to [mina f f ;maxa f f ] which will define the refined kidney re-

gions. – (c, above right). Segmentation showing inaccurate

results. – (d, below right). Refinement of the kidney region

shown in 5.c

the following criteria since they make up all the constraints

our medical environment imposes:

• Overall results: this is the most important aspect of the

considered methods since we make a judgement on the

segmentation quality and the precision with which bor-

ders are extracted.

• 3D: this criterion informs whether the method deals with

the entire volume of the organ or just one slice.

• Automatic: tells if the method is manual, semi-automatic

or entirely automatic.

• Execution time: when possible, informs how long it takes

to process the data and perform the segmentation.

• Type of images: tells the kind of images used as input

data (normal, contrast-enhanced, precise phase, ...).

3.3 Point cloud reconstruction

Because the output of the segmentation step is a point cloud,

it is now necessary to reconstruct it to have a triangular and

manifold model. The Poisson surface reconstruction [31] is

a recent algorithm that aims to mesh a point cloud of a model

M that has been oriented beforehand, i.e., the normal at each

point is known. There are several reasons why we use this

method: the main reason is that it is possible to compute an

approximating surface, i.e., the final surface does not fit ev-

ery point in the point cloud, but instead represents the mean

shape of the cloud. All of the methods described in section

2.2, except [63], assume that every point located in the pla-

nar slice is part of the contour, which is not the case of the

method we choose. Thus, it is very robust to noise which is

primordial in our case since medical images are noisy. Fur-

thermore, some segmentation errors that might remain are

ignored. This method also decreases the number of vertices

on the final surface (but not too radically, the resulting shape

is coherent and correct), which is helpful because the more

vertices there are, the slower the morphing methods. An-

other reason that we use this method is its fast computing

time and its automatism, which is primordial for our medi-

cal environment constraint.

The principle of the Poisson surface reconstruction is to

define an indicator function χ , that is specific to model M,

and that equals 1 for every point inside the point cloud and

0 outside. The final reconstruction is deduced from the ex-

traction of an appropriate isosurface (Figure 6). There is an

existing relation between the oriented point cloud and its

function χ: the gradient of χ is a vector field that is 0 almost

everywhere except near the surface, where it is equal to the

inward surface normal. Thus, the oriented point cloud can be

seen as samples of the gradient of χ . Computing χ amounts

to finding a function χ whose gradient best approximates a

vector field−!V that is defined by the normals at each point;

in other words, χ is a solution of−!∇ χ =

−!V . The applica-

tion of the divergence operator transforms this problem into

a classical Poisson equation: ∆ χ ⌘ ∇ ·−!∇ χ = ∇ ·

−!V .

Fig. 6: Overview of the Poisson surface reconstruction

The normal at each point of the point cloud is known,

which implies that−!V (=

−!∇ χ) is known at these points.

Nevertheless, the normals must be known for every point

p in R3. The main idea is then to find an expression for the

vector field−!V , the gradient of χ , and to deduce χ as a solu-

tion of the Poisson equation ∆ χ = ∇ ·−!V . Once χ is known,

we can extract the isosurface, which gives the final recon-

struction.


Method Overall results 3D Automatic Execution time Type of images

Wu et al. [68]

Organs are well

segmented although

some of their

dishomogeneities

can lead to minor

errors

No mention of 3D

volumeFully automatic

No mention of

execution timeRegular images

Rusko et al. [59]

In cases of large

lesion, the organ is

under-segmented

although these

errors can be

eliminated using a

probabilistic model

No mention of 3D.

However a volume

segmentation is

presented in the

results section

Fully automatic

Average time of 33s

using Intel Core2

Duo 2.2 GHz

processor with 2

GB RAM

Mutli- and

single-phase

contrast-enhanced

images

Lin et al. [41]

The results section

show accurate

kidney

segmentation, but

the method has only

been tested on

healthy organs

No mention of 3D

volumeFully automatic

No mention of

execution time

No mention about

the images is made,

but the results are

presented only for

contrast-enhanced

images.

Our method

A post-treatment is

used to eliminate

small segmentation

errors, although the

important ones are

only reduced

Segmentation of the

whole kidney

volume

Semi-automatic

(only one mouse

click is necessary)

Around 60s using

Intel Core i7

processor and 4 Go

of RAM

Regular images

Table 1: Comparison between our method and other region growing approaches for abdominal organ segmentation.

Because we have an oriented point cloud that is often

noisy, we consider the gradient of the smoothed function χ ,

which result from χ convolved with a smoothing filter. Let

M be the model to reconstruct, let δM be its surface, and

let χM be its indicator function. Let−!N δM(p) be the inward

normal at point p (p 2 δM), where F(q) is a smoothing fil-

ter and Fp(q) = F(q− p) is its translation to point p. The

gradient of the smoothed indicator function is defined as:

∇(χM ⇤F)(q) =Z

δMFp(q)

−!N δM(p)d p (2)

Obviously, we cannot use this formula because we do

not know δM yet, and therefore, we cannot evaluate its in-

tegral. However, it is possible to approximate the result by

partitioning δM into distinct patches Ps according to the

initial point cloud. More precisely, an octree is defined in

such way that every point of the original point cloud falls

into a leaf, where each leaf is considered a patch Ps. Let

S be the point cloud that is composed of a set of points s,

for which the positions are s.p and the normals are s.−!N . To

each point s is associated the leaf that it falls into. The union

of the leaves approximates δM. The integral of a patch Ps

is thus approximated because the coordinates of point s.pscaled to the area of Ps:

∇(χM ⇤F)(q) = ∑s2S

R

PsFp(q)

−!N δM d p

⇡ ∑s2S |Ps|Fs.p(q)s.−!N ⌘

−!V (q)

(3)

Now that the vector field−!V has been defined, we want

to know χ such that its gradient is close to−!V , i.e., a solu-

tion to the Poisson equation ∆ χ = ∇ ·−!V . Resolving such an

equation is a well-known problem (especially in physics),

and several methods exist for this purpose, but we will not

describe them here. To obtain the desired surface δM of the

initial point cloud, it is first necessary to set an isovalue to

extract the corresponding isosurface. We choose an isovalue

whose isosurface is the closest to the initial points. This ac-

tion is performed by evaluating χ at these positions and us-

ing the mean value:

δM = {q 2 R3kχ(q) = γ} with γ =

1

kSk ∑s2S

χ(s.p) (4)

where kSk is the number of points in the initial point

cloud S. Finally, the isosurface of the indicator function is

extracted using a method similar to the adaptation of March-

ing Cubes for an octree representation.


Fig. 7: Final result of a kidney point cloud (left) and its re-

construction using the Poisson surface reconstruction (right)

Function PoissonSurfaceReconstruction(Region R) :

Orient points of R

Discretize space with an octree

Compute a vector field that approximates the gradient of the

function f of the final surface

Deduce f from its gradient, (i.e., solve a Poisson equation)

Calculate the isosurface using f

End

Algorithm 4: Pseudo-code for the Poisson surface recon-

struction method

3.4 Dynamic modeling through mesh morphing

The originality of our method is that the tracking part is

based on a fully geometric approach, mesh morphing. Mesh

morphing is a method that is used to progressively trans-

form a source model Ms into a target model Mt by comput-

ing a smooth transition between the two models. Thus, here

propose a new approach that has two goals: the first goal is

the motion and deformation visualization of an organ (the

kidney in this case) under the influence of natural breathing.

The second goal results directly from the first goal and is the

tracking of part of this organ: its tumor. The advantage of

using mesh morphing is that our method is fast and requires

only three models, which correspond to the three breathing

phases the inhale phase, the mid-cycle phase and the exhale

phase. Moreover, the results obtained are fully geometric;

the output is an animated 3D model. In this way, the general

motion and all of its deformations can be studied at the same

time, where some methods offer only the possibility of a 2D

visualization. Finally, because the tumor is also animated, it

is possible to know its position at any time.

The most usual method for performing a mesh morphing

is to find a common vertex/edge/face network for both mo-

dels to compute a metamesh Mm that contains the topology

of Ms and Mt . This approach was first used by Kent et al.

in [32], in which both models are divided into small parts

(also called patches). Every part is then mapped onto the

unit disk. Most mesh morphing methods are based on disk or

other regular polygon mappings ([3], [23], [28], [34], [35]).

Unfortunately, these approaches always require either user

interaction (which can be very time consuming for some

methods) or a vertex correspondence between Ms and Mt

prior to the mesh morphing, which does not satisfy our con-

straints.

There are several ways to fully automate a mesh morph-

ing method. The most straightforward way is to map models

onto a sphere, which was first introduced in [32]. Indeed,

there is no need to divide the models anymore because they

are homeomorphic to a sphere. On the other hand, they must

be star-shaped, although the method described in [3] allows

us to solve minor covering problems. Another approach is

to use a constraint field, as described in [70].

The morphing stage must have very basic user-interac-

tions. The two models to morph are close to each other be-

cause they both come from the same kidney. Thus, the mesh

morphing method uses an automatic mesh cutting into two

patches, a unit disk mapping and a metamesh creation. We

cannot map onto a sphere because kidney models are not

star-shaped. All of these steps are described in detail be-

low. For the remainder of this paper, we will use the follow-

ing symbols: M represents a given model, Ms is the source

model and Mt is the target model. C is the connectivity be-

tween the vertices, edges and faces of M. V = {v1,v2, ...,vn}

is the position in R3 of the vertices. The edges are repre-

sented as a pair of vertices {i, j}, and the faces are repre-

sented as a triplet of vertices {i, j,k}. Finally, N(i) is the

set of vertices that are adjacent to vertex {i}, i.e., N(i) ={{ j}|{i, j} 2C}.

Mesh cutting: obtaining the tearing path

The first stage of the mesh comprises computing its princi-

pal axis. This computation can be performed by considering

only the vertices and using Principal Component Analysis

(PCA). Moreover, the PCA gives the 3 principal vectors of

the mesh; the first two and the barycenter of the mesh de-

fine the principal plane. Thus, the next stage is computing

the intersections between the edges of M and the principal

plane, defining what we call the intersected edges. In the

same way, the vertices {i, j} of an intersected edge are called

intersected vertices. This set of intersected edges is the first

stage of the final tearing path (see Figure 8).

The tearing path must be a unique loop of edges in C, i.e.

{{i1, i2},{i2, i3}, ...,{in−1, in}, {in, i1}|{ik, im}2C 8(k,m)2

[1;n]}; this set of edges is a subset of C and is called c. Thus,


Fig. 8: Intersection between the kidney model and its prin-

cipal plane (in blue). The resulting tearing path is displayed

in red

two successive intersected edges must share a vertex. The

purpose of the first post-process of the intersected edges is

to remove dead-end edges from c. Such an edge has one of

its vertices not shared with any other intersected edge, i.e.,

{{i, j}|8l 2 N( j){ j, l} /2 c}. To detect such edges, we first

compute the partial adjacency list of each vertex in c. This

list is the set of adjacent vertices { j} in c to a vertex {i}, i.e.,

{{ j}|{i, j} 2 c}. A dead-end edge is then simply detected

when at least one of its vertices has only one neighbor, i.e.,

its partial adjacency list length is 1 (see Figure 9 - b). The

second post-process consists of removing local loops; the

tearing path must be a unique succession of edges, and each

vertex must be shared by two and only two edges. Because

of the partial adjacency list, vertices from which the tearing

path separates are easily detected: such vertices have at least

3 neighbors. Thus, local loops are removed as follows. Start-

ing from a 2-adjacency vertex we arbitrarily choose one of

its neighbors, repeating this process until a 3-adjacency ver-

tex is reached. During this step, each vertex is skimmed only

once to ensure that it appears at most once in the final tear-

ing path. An arbitrary neighbor of the current 3-adjacency

vertex is still chosen, but every other edge that contains the

current vertex is suppressed from c. Because such a process

creates new dead-end edges, every edge of each 2-adjacency

neighbor is recursively suppressed until the neighbor is a 3-

adjacency vertex (see Figure 9 - c,d,e). Because the current

3-adjacency vertex becomes a 2-adjacency vertex, the whole

process is repeated until we return to the first vertex.

Mapping mesh onto the unit disk

Once the tearing path has been computed, vertices are tagged

Fig. 9: Example of the post-process of a tearing path. Al-

though this example cannot exist in a real situation, it

presents all of the cases that are needed to understand how

the full post-process works. From top to bottom: (a) Origi-

nal tearing path - (b) 1-adjacency vertex detection (diamond)

and dead-end edge suppression - (c) 3 (or more)-adjacency

vertex detection (square). Starting from the pointed vertex,

an arbitrary neighbor is chosen. - (d) For a 3-adjacency ver-

tex, we still choose an arbitrary neighbor, but every other

edge is suppressed. - (e) To avoid the appearance of new

dead-end edges when edges are suppressed, recursive sup-

pression of every edge from a 2-adjacency neighbor is per-

formed. - (f) The final tearing path that is obtained after the

post-process

in three different ways. We call them tag 0, 1 and 2. Tagging

the mesh allows us to define the two parts of it that will be

mapped later. Vertices that define the tearing path are tagged

as 0. A unique arbitrary neighbor of a vertex that is tagged

as 0 is tagged as 1. We recursively tag all of its neighbors,

which causes that connected part of the mesh to be tagged

as 1. The other part of the mesh is tagged as 2. Both meshes

are then rotated in such a way that their principal planes are


aligned with the xz-plane. In this way, it is possible to deter-

mine whether the parts that are tagged the same in the two

models have the same y orientation. If not, then tags 1 and 2

of one model are swapped. This step is essential because the

part of Ms that is tagged as 1 (resp. 2) will be morphed into

the part of Mt that is tagged as 1 (resp. 2) (see Figure 10).

Fig. 10: Example of two models for which the same tag has

a different y orientation. The vertices in red are tagged as

0, the vertices in cyan are tagged as 1 and the vertices in

magenta are tagged as 2

Now that every vertex is tagged, they can be mapped

onto the unit disk. Although any type of mapping is appli-

cable, we choose discrete harmonic mapping [53] because it

preserves the topology of the faces of both models as much

as possible. The most straightforward step of this mapping

is for the intersected vertices. These vertices are fixed on the

unit circle in such a way that the arc length between each

pair of successive vertices is proportional to the original

length of the edge in the mesh. For vertices that are tagged

as 1 or 2, discrete harmonic mapping (as well as other map-

ping) amounts to solving a linear system, which is described

as follows. Two distinct mappings are performed, one for

each tag. Let Vi be the vertices to map, with vertex indices i,

0 i < n tagged as 1 (resp. 2) and vertex indices i, n i < N

tagged as 0. Then, the linear system to solve is the following:

(I −Λ)

0

B

B

B

B

B

@

v1

v2

v3

...

vn−1

1

C

C

C

C

C

A

=

0

B

B

B

B

B

@

∑N−1i=n λ0,ivi

∑N−1i=n λ1,ivi

∑N−1i=n λ2,ivi

...

∑N−1i=n λn−1,ivi

1

C

C

C

C

C

A

(5)

where Λ = {λi, j} and λi, j is a coefficient that depends

on the mapping used. Here, for discrete harmonic mappings,

we have:

λi, j =

(

cotαi, j+cotβ i, j

∑ j2N(i)(cotαi, j+cotβi, j)if {i, j} 2C

0 if {i, j} /2C(6)

with αi, j =∠(i,k0, j) and βi, j =∠(i,k1, j). Edge {i, j} is

adjacent to two and only two faces because M is a triangular

mesh. Here, k0 and k1 are the two vertices that define these

faces. We call M0sN

M0tN

the mapping of Ms and Mt , for tag

N. Similarly we call {i0} a mapped vertex. Although such

a notation should not exist because only the position of the

vertices (vi) changed during the mapping, this notation will

make further expressions more straightforward.

Metamesh creation and animation: computing intersections

and barycentric coordinates

The next step is to overlay M0sN

M0tN

for both tags to compute

the metamesh. The first stage is to detect the intersections

between mapped edges. When two edges {i0, j0} 2C for M0s

and {k0, l0} 2 C for M0t cross, a new vertex is created. Two

valid definitions of this intersection point are v0i +α−!v0iv

0j and

v0k + β−−!v0kv0l . Coefficients α and β are saved along with the

new vertex. These coefficients will be necessary for interme-

diate models because they are sufficient to compute the co-

ordinates of the vertex, even when vi,v j,vk and vl are inter-

polated. This type of vertex is called an intersection vertex.

Once an intersection vertex is created, appropriate edges and

faces are also created to build the topology of the metamesh

Mm. These new edges and faces will allow Mm to combine

the topology of both Ms and Mt and to have a continuous

interpolation between the two models (see Figure 11).

Fig. 11: Example intersections between the mapped edges of

Ms (solid line) and Mt (dotted line). The intersection points

1, 2, 3 and 4 are created, as are appropriate edges (C1, 1D,

C2, 2F, ...) and faces (C12, F32, ...)

The second stage of the metamesh creation is the com-

putation of barycentric coordinates (BC) for every vertex of

Ms and Mt . To accomplish this computation, we first want to

know on which mapped face {i0, j0,k0} of M0t (resp. of M0

s)

a mapped vertex v0m of M0s (resp. M0

t ) lies on. The BC are a

unique triplet u,v,w such that v0m = uv0i+vv0j +wv0k. The face

that v0m lies on and its BC are saved. This type of vertex is

called a mesh vertex.

Thus, the metamesh is completely built and is composed

of a set of intersection vertices and mesh vertices. Interme-

diate models can now be easily obtained by interpolating


the positions of the vertices. The interpolation is possible

because we know, for each of them, an initial and a final po-

sition, as follows: for a mesh vertex coming from Ms, the ini-

tial position is its position in Ms. The final position is known

from the combination of its BC and the face of Mt that it lies

on. Inversely, for a mesh vertex coming from Mt , the initial

position is known using its BC and the face of Ms that it lies

on. The final position is its natural position in Mt . For an

intersection vertex, the initial position is known because of

its α coefficient and the edge of Ms that it lies on. The final

position is computed using its β coefficient and the edge of

Mt that it lies on.

Function MeshMorphing(3D Model m1, 3D Model m2) :

For m1 and m2 do

Cut mi into two parts according to its main axis

Map both parts of mi onto the unit disk

end For

Overlap the projections of m1 and m2

Deduce and compute the final metamesh

End

Algorithm 5: Pseudo-code for the mesh morphing ap-

proach

3.5 Tracking the tumor

Tumor tracking is the second goal of our method. It is im-

portant to know the location of the tumor to adjust the wave

beam accordingly. From this perspective, there are two main

differences between the tumor and the kidney. The first dif-

ference is that the tumor is not deformed by the respiratory

cycle; it only moves with the kidney. The second difference

is that the tumor is similar to an ellipsoid. In the segmenta-

tion step, the tumor is segmented separately from the kidney

and in a such way that the center of the tumor is known. Per-

forming another mesh morphing to obtain the tumor move-

ments (i.e., its tracking), would be inappropriate because its

shape remains the same from one breathing phase to an-

other. Moreover, the cost of the computational time would

be useless. A more convenient and fast way to accomplish

the same task is to interpolate the position of the tumor be-

cause we have the coordinates of its center for the inhale,

exhale and mid-cycle phases. We can use a quadratic Bezier

curve interpolation, which gives the tumor position for inter-

mediate phases. Thus, the 3D coordinates are known at any

time, and the tumor tracking is accomplished.

3.6 First results

The first results of the motion and deformation simulation

through mesh morphing are presented in Figure 12. This ap-

proach was first tested on three meshes that have roughly

the shape of a kidney for the inhale, exhale and mid-cycle

phases. Although these three models were modeled by hand

using classic 3D modeling software (Blender) and are to-

tally artificial, they attempt to simulate the type of motion

and deformation that a real kidney shows. Because images

alone are not truly relevant for an animated result, the reader

can refer to Online Resource 1 or the following link:

http://youtu.be/G-fdO93Yb5s .

The results show a soft transition between the three mod-

els, which is what the method aims for. Although this tran-

sition is the ideal case (the models are very close to each

other, they have the same orientation and they are smoother

and flatter than real kidney models), no degenerated triangle

or non-coherent movement appears during the whole anima-

tion. Moreover, global and local deformations are well sim-

ulated, and the general motion is satisfied. The time com-

putation that is needed to obtain the dynamic model is ap-

proximately 5 seconds, which is fast and totally acceptable

for our medical environment. However, the three artificial

kidney models that we used have a moderate resolution (ap-

proximately 650 vertices, 2000 edges and 1300 faces).

4 Results of the whole method

4.1 Assessment

The results are evaluated through three sets of CT-Scan ac-

quisitions which spatial resolutions are 0.863⇥1.25⇥0.863

mm for the three of them. The kidney is present in approxi-

mately 160 slices. If we consider that we only have the slices

in which the kidney is present, the only user interactions dur-

ing the whole process are three mouse clicks, to define the

seeds that are necessary to segment the kidney for the three

sets.

Despite the refinement of the region-growing approach,

some errors are possible during the segmentation step. These

errors can occur when an organ is located right next to the

kidney or around its natural cavity, where various arteries

and veins are present. The choice of the Poisson surface re-

construction is relevant in this case. One advantage of this

method is that the final reconstruction can be more or less

accurate depending on the depth of the octree. By setting

this parameter to 5, we chose not to consider the details and

to approximate the point cloud; in other words, points com-

ing from segmentation errors will be ignored as long as they

are not frequent or are all located around the same region of


Fig. 12: Motion and deformation simulation of three artificial kidney models. Global and local deformations are both satis-

fied. The source and target models are displayed in red. The intermediate models are displayed in gray

the kidney.

To obtain a full animation, two mesh morphings are per-

formed: the first morphing is between M1 and M2, and the

second is between M2 and M3, which correspond to the in-

hale, mid-cycle and exhale phases. Figures 14 and 15 present

several intermediate models that were obtained while per-

forming the morphing from M1 to M2 to M3. Because the

results are not very explicit with frozen models, an animated

version can be seen on Online Resource 2 or at the following

URL: http://youtu.be/wXanzy87pOQ.

General movement and deformation of the kidney are re-

spected. The natural rotation of the principal axis of the or-

gan is present here, as is its enlargement. However, the local

deformations are not totally satisfactory, especially for the

tumor. The tumor on the morphed kidney is absorbed into a

part of the kidney and reappears from a different part, right

next to its original location. The natural deformation would

have been a smooth displacement between these two loca-

tions, almost like a translation. This outcome is due to the

morphing method itself and a way to overcome this problem

is discussed in the next section. Although these false defor-

mations are not very noticeable, they become obvious when

the tumor is displayed: it sticks out of the kidney model (Fig-

ures 13). The spatial resolution of the acquisition can also

affect the precision of our dynamic model; a low resolution

would lead to a sparse model. As positioning approxima-

tions are inherent to mesh morphing methods, they would

become more important and noticeable on a model defined

by a few vertices. The more vertices, the more attenuated the

approximations are.

The whole process takes less than 2 minutes and has

been tested on a laptop with Intel Core i7 processor and 4

Go of RAM. The longest part is the segmentation (1 minute),

more precisely, the final post-treatment when mathematical

morphology operators are applied (approximately 44 sec-

onds). The shortest part is the Poisson surface reconstruc-

tion because the depth of the octree is low (up to 3 seconds).

The morphing step is computed in 40 seconds for models

Fig. 13: Highlighting the local deformation problem. The in-

termediate model with the tumor (blue ellipsoid) presents a

local inaccuracy, especially for the tumor region (encircled)

composed of up to 2,300 vertices, 6,900 edges and 4,600

faces. Although these computational times do not allow real-

time use, they are acceptable for our medical environment

in which interventions used for our non-invasive tumor de-

struction (High Intensity Focused Ultrasound) are very long

(3 hours for uterine cancer). Moreover, the computational

time is needed only once, at the beginning. The animation

and the tracking are performed in real time because they

simply involve an interpolation between an initial and final

position of the vertices, as seen at the end of section 3.4.

Furthermore, our approach can be easily accelerated using

code optimization and GPU multithreading programming.

From a practical point of view, the dynamic model is

animated according to the breathing phase of the subject.

General anesthesia is almost always used for HIFU proto-

cols [71], which offers two advantages in our case: the first

one is that the patient is under artificial respiration. It is then

straightforward to know its breathing phase and to enslave

the dynamic model accordingly. The second advantage is

that interfering motions (like spams) that could affect the

kidney movement are avoided.

4.2 Comparison

Although the major part of the previous work in section

2.3 perform organ tracking, we consider only three of them

since they are closely related to our method: they present a

dynamic 3D model which position and/or shape correspond

to a given breathing phase. Other methods do not offer a


Fig. 14: Final results showing natural movements of the right kidney due to respiration. The source and target models that

are obtained from the reconstruction are displayed in red. The intermediate models are displayed in gray. Morphing from M1

to M2 is shown (from left to right)

Fig. 15: Morphing between M2 and M3 from a different point of view (a rotation of 180 degrees around the vertical axis).

The models are displayed in wireframe, and the tumor is visible (blue ellipsoid)

volume visualization and are therefore not suitable for an

objective comparison. This comparison is based on the fol-

lowing criteria and is presented on Table 2:

• Simulation: informs the kind of simulation the method

offers.

• Data/equipment needed: this criterion indicates if data is

necessary beforehand and/or the input the method needs,

as well as eventual special equipment.

• Drift correction: tells how the method deals with organ

drift over time and how its position is readjusted.

• Execution time: when possible, informs how long it takes

to get the organ tracking.

• Error estimation: when possible, this criterion indicates

the estimated error between the predicted position of the

organ and its real position.

5 Conclusions

We have presented a complete method that starts from three

CT-Scans or MRI acquisition and moves to kidney motion

simulation and tumor tracking. This method is divided into

3 major steps. The first step is kidney segmentation, which

is performed through a semi-automatic region-growing ap-

proach. This step requires a mouse click to define the seed.

Although the segmentation is then refined using histogram

analysis, small errors can remain. Fortunately, these errors

can be ignored during the second step, the surface recon-

struction. This step is performed through Poisson surface

reconstruction, which offers the possibility of computing a

smooth surface without considering all of the points of the

point cloud to reconstruct, resulting in the general shape of

the cloud. The third step is the mesh morphing among the

three kidney models, which corresponds to three breathing

phases. Two mesh morphings are performed here to make

soft transitions between the first and second model and be-

tween the second and third models. This step is fully auto-

matic and is based on mesh cutting, unit disk mapping and

metamesh creation. The output of the entire method is fully

geometric because it is a 3D model of the kidney for any

phase of the respiratory cycle. Although general deforma-

tion and movement of the kidney is simulated well, local

deformations are not precise enough, especially for tumors

near the surface. Our method could be used for other abdo-

minal organs like the spleen or the liver since they have a tis-

sue density close to the kidney. However, the shape of these

organs present regions of high curvature which could lead to

more important approximations during the mesh morphing

stage.

There are many prospects for our future work. The first

prospect is the validation of the model to quantify the er-

ror between its position and the real kidney and tumor po-

sition. This goal can be accomplished by using a fourth ac-

quisition for a given phase of the breathing cycle in which

the kidney and tumor are manually segmented by experts.

The comparison between these borders and the borders of

the dynamic model (for the same breathing phase) will pro-


Method SimulationData/equipment

neededDrift correction Execution time Error estimation

Arnold et al. [5]Only motion

simulation

• Built-up atlas

which consists in a

set of vector fields

corresponding to the

motion of the liver

for every phase of

the respiratory

cycle.

• Any equipment to

track the patient’s

breathing is

necessary.

One slice is

acquired every 60 s

for the exhale phase

and is used to

compute

misposition of the

simulated model

1.3 s (computer

configuration is not

specified)

1.1 mm

Hostettler et al. [26]Only motion

simulation

• In addition of the

organ

reconstruction, the

segmentation of the

skin, viscera volume

and diaphragm is

necessary.

• Markers are

positioned onto the

patient’s skin to

track the breathing

cycle.

Not mentioned

Real-time using a

core 2Duo T7700,

2.4 GHz, 2Go

RAM, GeForce

8600M GT 512 MB

2 to 3 mm

Preiswerk et al. [54]Only motion

simulation

• Statistical model

obtained by

analyzing

differences between

the vector fields of

several manually-

segmented

models.

• Breathing cycle is

know at any time by

tracking a 3

surrogate markers,

which may be

implanted

electromagnetic

beacons.

Implicitly

considered by the

method

Real-time

(computer

configuration is not

specified)

1.2 mm

Our methodBoth motion and

deformation

• Input necessary

consist of three

acquisitions for

inhale, mid-cycle

and exhale phases.

• Any equipment to

track the patient’s

breathing is

necessary.

Registration of the

simulated model

using two

orthogonal flash

acquisitions (future

work)

A computation time

of 40 s is necessary

beforehand using an

Intel Core i7

processor and 4 Go

of RAM. Then,

motion and

deformation are

computed in real

time

N/A (future work)

Table 2: Comparison between our method and other dynamic modeling approaches used for organ tracking.

vide a scientific clue to how far our model is and the robust-

ness of the method. The second prospect is the improvement

of the segmentation step because errors can still occur. An

approach based on mathematical morphology pre-treatment

and watershed algorithm is considered. Although watershed

is known for over-segmented results, the pre-treatment al-

lows it to have an acceptable segmentation, and the first re-

sults are encouraging. Nevertheless, CT-Scan acquisitions

will most likely not be used anymore because there is no

tumor heating control, which is fundamental for HIFU ther-

apy. MRI acquisitions will be used instead. The contrasts in

these images are greater than the contrasts in CT-Scan im-

ages. Thus, the segmentation part will work better, giving

better results and reducing errors. The third prospect will

address mesh morphing, especially to correct the animation

of some local deformation. A way to overcome this problem


would be to impose the parts of the model with close cur-

vature morph into each other. Finally, the last prospect is to

consider the natural drift of organ motion and to adapt our

dynamic model consequently. A way to apply these move-

ments could be based on [15,14] and would consist in seg-

menting the kidney for two orthogonal slices at a given breath-

ing phase. As only two slices are involved, their acquisition

(called flash acquisition) as well as their segmentation can

be performed in real time. Thus, the dynamic model could

be registered according to these kidney borders and would

lead to its readjustment.

Acknowledgements This work is granted by the group Novartis and

by the Foundation ”Sante, Sport et Developpement Durable”, presided

over Pr. Yvon Berland. The authors would like to thank the persons in-

volved in the Kidney Tumor Tracking (KiTT) project : Christian Coulange

for his precious help, Marc Andre, Frederic Cohen Philippe Souteyrand

and Julien Frandon for their wise advice and for providing CT scan

data, and Pierre-Henri Rolland for his support.

References

1. Ahmed, M., Yamany, S., Mohamed, N., Farag, A., Moriarty, T.:

A modified fuzzy C-means algorithm for bias field estimation and

segmentation of MRI datac-means algorithm for bias field estima-

tion and segmentation of mri data. IEEE Transactions on Medical

Imaging 21(3) (2002)

2. Akkouche, S., Galin, E.: Adaptive implicit surface polygonization

using marching triangles. Computer Graphics Forum 20, 67–80

(2001)

3. Alexa, M., Cohen-Or, D., Levin, D.: As-rigid-as-possible shape

interpolation. Proceedings of Computer Graphics and Interactive

Techniques (2000)

4. Amrani, M., Crespin, B., Shariat, B.: Skeletal implicit surface re-

construction and hybrid model for organs interactions simulation

(2003)

5. Arnold, P., Preiswerk, F., Fasel, B., Salomir, R., Scheffler, K., Cat-

tin, P.: 3d organ motion prediction for mr-guided high intensity

focused ultrasound. MICCAI 6892, 623–630 (2011)

6. Atkins, S., Mackiewich, B.: Fully automatic segmentation of the

brain in MRI. IEEE Transactions on Medical Imaging 17(1), 98 –

107 (1998)

7. Bardinet, E., Cohen, L., Ayache, N.: A parametric deformable

model to fit unstructured 3D data. Computer Vision and Image

Understanding 71, 39 – 54 (1995)

8. Bardinet, E., Cohen, L., Ayache, N.: Tracking and motion analy-

sis of the left ventricle with deformable superquadrics. Medical

Image Analysis 1(2), 129 – 149 (1996)

9. Barequet, G., Shapiro, D., Tal, A.: Multilevel sensitive reconstruc-

tion of polyhedral surfaces from parallel slices. The Visual Com-

puter 16(2), 116 – 133 (2000)

10. Boes, J., Weymouth, T., Meyer, C.: Multiple organ definition in CT

using a bayesian approach for 3D model fitting. Vision Geometry

2573, 244 – 251 (1995)

11. Boissonnat, J.D.: Shape reconstruction from planar cross-sections.

Rapport de Recherche INRIA 546 (1986)

12. Boscolo, R., Brown, M., McNitt-Gray, M.: Medical image seg-

mentation with knowledge-guided robust active contours. Radio

Graphics 22, 437 – 448 (2002)

13. Cotin, S., Delingette, H., Ayache, N.: Real-time elastic deforma-

tions of soft tissues for surgery simulation. Rapport de Recherche

INRIA 3511 (1998)

14. Cresson, T., Branchaud, D., Chav, R., Godbout, B., de Guise, J.A.:

3d shape reconstruction of bone from two x-ray images using

2d/3d non-rigid registration based on moving least-squares defor-

mation. Proceedings of the SPIE, Medical Imaging 2010: Image

Processing 7623 (2010)

15. Cresson, T., Chav, R., Branchaud, D., Humbert, L., Godbout, B.,

Aubert, B., Skalli, W., De Guise, J.: Coupling 2d/3d registration

method and statistical model to perform 3d reconstruction from

partial x-rays images data. Annual International Conference of the

IEEE in Engineering in Medicine and Biology Society pp. 1008–

1011 (2009)

16. Delingette, H.: Simplex meshes: a general representation for 3D

shape reconstruction. Rapport de Recherche INRIA 2214 (1994)

17. Ekoule, A.B., Peyrin, F.C., Odet, C.L.: A triangulation algorithm

from arbitrary shaped multiple planar contours. ACM Transac-

tions on Graphics 10(2), 182 – 199 (1991)

18. Fritsch, D., Pizer, S., Yu, L., Johnson, V., Chaney, E.: Localiza-

tion and segmentation of medical image objects using deformable

shape loci. Information Processing in Medical Imaging 1230, 127

– 140 (1997)

19. Fuchs, H., Kedem, Z.M., Uselton, S.P.: Implicit surfaces for semi-

automatic medical organs reconstruction. Communications of the

ACM 20(10) (1977)

20. Galin, E., Akkouche, S.: Fast surface reconstruction from contours

using implicit surfaces. Implicit Surfaces’98 Conference 3, 139 –

144 (1998)

21. Gao, J., Kosaka, A., Kak, A.: A deformable model for human or-

gan extraction. International Conference on Image Proessing 3

(1998)

22. Ginneken, B., Frangi, A., Staal, J., ter Haar Romeny, B., Viergever,

M.: Active shape model segmentation with optimal features. IEEE

Transactions on Medical Imaging 21(8) (2002)

23. Gregory, A., State, A., Lin, M., Manocha, D., Livingston, M.:

Feature-based surface decomposition for correspondence and

morphing between polyhedra. Proceedings of Computer Anima-

tion pp. 64 – 71 (1998)

24. Hartmann, E.: A marching method for the triangulation of sur-

faces. The Visual Computer 14, 95–108 (1998)

25. Held, K., Kops, E.R., Krause, B., III, W.W., Kikinis, R., Muller-

Gartner, H.W.: Markov random field segmentation of brain MR

images. IEEE Transactions on Medical Imaging (1997)

26. Hostettler, A., Nicolau, S., Soler, L., Remond, Y., Marescaux, J.:

A real-time predictive simulation of abdominal organ positions in-

duced by free breathing. International Symposium on Biomedical

Simulation pp. 89 – 97 (2008)

27. Ju, T., Warren, J., Carson, J., Eichele, G., Thaller, C., Chiu, W.,

Bello, M., Kakadiaris, I.: Building 3D surfce networks from 2D

curve networks with application to anatomical modeling. Proceed-

ings of Pacific Graphics 21, 764 – 773 (2005)

28. Kanai, T., Suzuki, H., Kimura, F.: Metamorphosis of arbitrary tri-

angular meshes. Proceedings of Computer Graphics and Applica-

tion 20(2) (2000)

29. Kapur, T., Grimson, E., Kikinis, R., Wells, W.: Enhanced spatial

priors for segmentation of magnetic resonance imagery. Medical

Image Computing and Computer-Assisted Intervention (1998)

30. Kaus, M., Warfield, S., Nabavi, A., Black, P., Jolesz, F., Kikinis,

R.: Automated segmentation of MR images of brain tumors. Ra-

diology 218, 586 – 591 (2001)

31. Kazhdan, M., Bolitho, M., Hoppe, H.: Poisson surface reconstruc-

tion. Eurographics Symposium on Geometry Processing (2006)

32. Kent, J., Carlson, W., Parent, R.: Shape transformation for poly-

hedral objects. Computer Graphics 26(2) (1992)

33. Klein, R., Schilling, A., Strasser, W.: Reconstruction and simpli-

fication of surfaces from contours. Proceedings of The Seventh


Pacific Conference on Computer Graphics and Applications pp.

198 – 207 (1999)

34. Lee, A., Dobkin, D., Sweldens, W., Schroder, P.: Multiresolution

mesh morphing. Proceedings of Computer Graphics and Interac-

tive Techniques (1999)

35. Lee, A., Sweldens, W., Schroder, P., Cowsar, L., Dobkin, D.:

Maps: Multiresolution adaptive parameterization of surfaces. Pro-

ceedings of SIGGRAPH pp. 95 – 104 (1998)

36. Lee, C.C., Chung, P.C., Tsai, H.M.: Identifying multiple abdomi-

nal organs from CT image series using a multimodule contextual

neural network and spatial fuzzy rules. IEEE Transactions on In-

formation Technology in Biomedecine 7(3) (2003)

37. Lei, T., Sewchand, W.: A new stochastic model based image seg-

mentation tech for CT images. IEEE Transactions on Medical

Imaging 11(1) (1992)

38. Leonardi, V., Mari, J.L., Vidal, V., Daniel, M.: A morphing ap-

proach for kidney dynamic modeling - from 3d reconstruction to

motion simulation. 20th International Conference in Central Eu-

rope on Computer Graphics, Visualization and Computer Vision,

WSCG 2012 pp. 179 – 187 (2012)

39. Liang, Z., MacFall, J., Harrington, D.: Parameter estimation and

tissue segmentation from multispectral MR images. IEEE Trans-

actions on Medical Imaging 13(3) (1994)

40. Lim, S., Haron, H.: Surface reconstruction techniques: a review.

Artificial Intelligence Review pp. 1–20 (2012). URL http://

dx.doi.org/10.1007/s10462-012-9329-z41. Lin, D.T., Lei, C.C., Hung, S.W.: Computer-aided kidney segmen-

tation on abdominal CT images. IEEE Transactions on Informa-

tion Technology in Biomedecine 10(1) (2006)

42. Lin, J.S., Cheng, K.S., Mao, C.W.: Multispectral magnetic reso-

nance images segmentation using fuzzy hopfield neural network.

International Journal of Biomedical Computing 42, 205 – 214

(1996)

43. Lorensen, W.E., Cline, H.E.: Marching cubes : A high resolution

3D surface reconstruction algorithm. SIGGRAPH (1987)

44. Mari, J.L.: Approaches dedicated to the modeling of complex

shapes - application to medical data. Journal of Medical Infor-

matics and Technologies 7, 3 – 15 (2004)

45. Murphy, M., Chang, S., Gibbs, I., Le, Q.T., Hai, J., Kim, D., Mar-

tin, D., Adler, J.: Patterns of patient movement during frameless

image-guided radiosurgery. International Journal of Radiation On-

cology Biology Physics 55(5), 1400 – 1408 (2003)

46. Nakamoto, M., Ukimura, O., Gill, I., Mahadevan, A., Miki, T.,

Hashizume, M., Sato, Y.: Realtime organ tracking for endoscopic

augmented reality visualization using miniature wireless magnetic

tracker. Medical Imaging and Augmented Reality pp. 359 – 366

(2008)

47. Nicolau, S., Pennec, X., Soler, L., Ayache, N.: Clinical evaluation

of a respiratory gated guidance system for liver punctures. Medi-

cal Image Computing and Computer-Assisted Intervention pp. 77

– 85 (2007)

48. Noe, K.O., de Senneville, B.D., Elstrom, U.V., Tanderup, K.,

Sorensen, T.S.: Acceleration and validation of optical flow based

deformable registration for image-guided radiotherapy. Acta On-

cology 47(7), 1286 – 1293 (2008)

49. Olbricha, B., Trau, J., Wiesner, S., Wicherta, A., Feussner, H.,

Navab, N.: Respiratory motion analysis: Towards gated augmenta-

tion of the liver. Computer Assisted Radiology and Surgery 1281,

248 – 253 (2005)

50. Pavlovich, C., Walther, M., Choyke, P., Pautler, S., Chang, R.,

Linehan, M., Wood, B.: Percutaneous radio frequency ablation of

small renal tumors: Initial results. The Journal of Urology 167,

10–15 (2002)

51. Pham, D., Prince, J.: Adaptative fuzzy segmentation of magnetic

resonance images. IEEE Transactions on Medical Imaging (1998)

52. Pihuit, A., Palombi, O., Cani, M.P.: Reconstruction implicite de

surfaces 3D a partir de regions 2D dans des plans paralleles. 22e

Journees de l’Association Francophone d’Informatique Graphique

(2009)

53. Polthier, K.: Conjugate harmonic maps and minimal surfaces.

Tech. rep., Technische University of Berlin (2000)

54. Preiswerk, F., Arnold, P., Fasel, B., Cattin, P.: A bayesian frame-

work for estimating respiratory liver motion from sparse measure-

ments. Abdominal Imaging. Computational and Clinical Applica-

tions 7029, 207 – 214 (2012)

55. Rajapakse, J., Giedd, J., Rapoport, J.: Statistical approach to seg-

mentation of single-channel cerebral MR images. IEEE Transac-

tions on Medical Imaging 16(2) (2001)

56. Reyes, M., Malandain, G., Koulibaly, P.M., Ballester, M.G., Dar-

court, J.: Respiratory motion correction in emission tomography

image reconstruction. Medical Image Computing and Computer-

Assisted Intervention pp. 396 – 376 (2005)

57. Rohlfing, T., Maurer, C., O’Dell, W., Zhong, J.: Modeling

liver motion and deformation during the respiratory cycle using

intensity-based free-form registration of gated MR images. Med-

ical Imaging 2001: Visualization, Image-Guided Procedures, and

Display pp. 337 – 348 (2001)

58. Rueckert, D., Sonoda, L.I., Hayes, C., Hill, D.L.G., Leach, M.O.,

Hawkes, D.J.: Nonrigid registration using free-form deformations:

Application to breast MR images. IEEE Transactions on Medical

Imaging 18(8) (1999)

59. Rusko, L., Bekes, G., Fidrich, M.: Automatic segmentation of the

liver from multi- and single-phase contrast-enhanced CT images.

Medical Image Analysis 13, 871 – 882 (2009)

60. Sarrut, D., Boldea, V., Miguet, S., Ginestet, C.: Simulation of four-

dimensional CT images from deformable registration between in-

hale and exhale breath-hold CT scans. Medical Physics 33(3)

(2006)

61. Schweikard, A., Glosser, G., Bodduluri, M., Murphy, M., Adler,

J.: Robotic motion compensation for respiratory movement during

radiosurgery. Journnal of Computer-Aided Surgery (2000)

62. Shirato, H., Shimizu, S., Kitamura, K., Nishioka, T., Kagei, K.,

Hashimoto, S., Aoyama, H., Kunieda, T., Shinohara, N., Dosaka-

Akita, H., Miyasaka, K.: Four dimensional treatment planning anf

fluoroscopic real-time tumor tracking radiotherapy for moving tu-

mor. International Journal of Radiation Oncology Biology Physics

48, 435 – 442 (2000)

63. Tsingos, N., Bittar, E., Gascuel, M.P.: Implicit surfaces for semi-

automatic medical organs reconstruction. Computer Graphics In-

ternational pp. 3 – 15 (1995)

64. Vilarino, D., Cabello, D., Balsi, M., Brea, V.: Image segmentation

based on active contours using discrete time cellular neural net-

works. Journal of VLSI Signal Processing Systems 23, 403 – 414

(1999)

65. Villard, P.F.: Simulation du mouvement pulmonaire pour un traite-

ment oncologique. Ph. D. Thesis at University Claude-Bernard

Lyon 1 (France) (2006)

66. Wells, W., Grimson, W., Kikinis, R., Jolesz, F.A.: Adaptive seg-

mentation of MRI data. IEEE Transactions on Medical Imaging

15, 429 – 442 (1996)

67. Wrazidlo, W., Brams, H., Lederer, W., Schneider, S., Geiger, B.,

Fischer, C.: An alternative method of three-dimensional recon-

struction from two-dimensional CT and MR data sets. European

Journal of Radiology 12, 11 – 16 (1991)

68. Wu, J., Poehlman, S., Noseworthy, M., Kamath, M.: Texture fea-

ture based automated seeded region growing in abdominal MRI

segmentation. Journal of Biomedical Science and Engineering 2,

1 – 8 (2009)

69. Xu, S.B., Lu, W.X.: Surface reconstruction of 3D objects in com-

puterized tomography. Computer Vision, Graphics and Image Pro-

cessing 44, 270 – 278 (1988)

70. Yan, H.B., Hu, S.M., Martin, R.: 3d morphing using strain field

interpolation. Computer Science and Technology 1 (2007)


71. Yao, C.L., Trinh, T., Wong, G.T.C., Irwin, M.G.: Anaesthesia for

high intensity focused ultrasound (hifu) therapy. Journal of the

Association of Anaesthetists of Great Britain and Ireland 63, 865–

872 (2008)

72. Zhang, Y., Brady, M., Smith, S.: Segmentation of brain MR

images through a hidden markov random field model and the

expectation-maximization algorithm. IEEE Transactions on Med-

ical Imaging 20(1) (2001)

ablation radioguidee des masses renales - Thèses

Documents

Transcript of ablation radioguidee des masses renales - Thèses