www.elsevier.com/locate/ynimg

NeuroImage 39 (2008) 369–382

Motor and language DTI Fiber Tracking combined withintraoperative subcortical mapping for surgical removal of gliomas☆

Lorenzo Bello,a,⁎ Anna Gambini,b Antonella Castellano,b Giorgio Carrabba,a Francesco Acerbi,a

Enrica Fava,a Carlo Giussani,a Marcello Cadioli,b Valeria Blasi,b Alessandra Casarotti,a,c

Costanza Papagno,c Arun K. Gupta,d Sergio Gaini,a Giuseppe Scotti,b and Andrea Falinib

aNeurosurgery, Department of Neurological Sciences, University of Milano, Fondazione IRCCS Ospedale Maggiore Policlinico, Mangiagalli e Regina Elena,Via Francesco Sforza 35, 20122, Milan, ItalybNeuroradiology Department and CERMAC (Centro di Eccellenza Risonanza Magnetica ad Alto Campo),Scientific Institute and University Vita–Salute San Raffaele, Milan, Italy

cPsychology, University Milano Bicocca, Milan, ItalydDepartment of Imaging Sciences and Interventional Radiology, Sree Chitra Tirunal Institute for Medical Sciences and Technology, Trivandrum, India

Received 29 May 2007; revised 11 August 2007; accepted 20 August 2007Available online 29 August 2007

Preoperative DTI Fiber Tracking (DTI-FT) reconstruction of func-tional tracts combined with intraoperative subcortical mapping (ISM)is potentially useful to improve surgical procedures in gliomas locatedin eloquent areas. Aims of the study are: (1) to evaluate themodifications of fiber trajectory induced by the tumor; (2) to validatepreoperative DTI-FT results with intraoperative identification offunctional subcortical sites through direct subcortical stimulation; (3)to evaluate the impact of preoperative DTI-FT reconstructions in aneuronavigational setup combined with ISM technique on duration andmodalities of surgical procedures, and on functional outcome of thepatients.

Data are available on 64 patients (52 low-grade and 12 high-gradegliomas). DTI-FTwas acquired by a 3-TMR scanner with a single-shotEPI sequence (TR/TE 8986/80 ms, b=1000 s/mm) with gradientsapplied along 32 non-collinear directions. 3D Fast Field Echo (FFE) T1-weighted imaging (TR/TE 8/4 ms) was performed for anatomicguidance. The corticospinal tract (CST), superior longitudinal, inferiorfronto-occipital and uncinatus fasciculi were reconstructed. Data weretransferred to the neuronavigational system. Functional subcortical sitesidentified during ISM were correlated with fiber tracts depicted byDTI-FT.

In high-grade gliomas, DTI-FT depicted tracts mostly at the tumorperiphery; in low-grade gliomas, fibers were frequently located insidethe tumormass. There was a high correlation between DTI-FTand ISM(sensitivity for CST=95%, language tracts=97%). For a proper re-construction of the tracts, it was necessary to use a low FA threshold of

☆ This work was supported by grants from AIRC (Associazione ItalianaRicerca sul Cancro), Compagnia di San Paolo, Torino, Fondazione Banca delMonte di Lombardia, Pavia, and Fondazione ItaloMonzino,Milano (to L.B.).⁎ Corresponding author. Fax: +39 0259902239.E-mail address: lorenzo.bello@unimi.it (L. Bello).Available online on ScienceDirect (www.sciencedirect.com).

1053-8119/$ - see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.neuroimage.2007.08.031

fiber tracking algorithm and to position additional regions of interest(ROIs). The combination of DTI-FTand ISM decreased the duration ofsurgery, patient fatigue, and intraoperative seizures.

Combination of DTI-FT and ISM allows accurate identification ofeloquent fiber tracts and enhances surgical performance and safetymaintaining a high rate of functional preservation.© 2007 Elsevier Inc. All rights reserved.

Keywords: Gliomas; DTI Fiber Tracking; Motor and language subcorticalmapping

Introduction

Gliomas are the most common primary neoplasms of the centralnervous system. Surgery plays an important role in treatment,because it relieves symptoms, and provides accurate histologicaland molecular diagnosis. Goals of surgery are maximal tumorremoval associated with minimal post-operative morbidity andmaximal patient functional preservation (Bello et al., 2007; Berger,1996; Black, 1991). These principles apply particularly for gliomaslocated in eloquent areas. It is essential during surgical removal oflesions involving speech or motor areas or pathways to identifyfunctional brain tissue surrounding or involved by the tumor, toavoid removal of eloquent structures (Berger, 1996; Black, 1991;Duffau et al., 2002; Skirboll et al., 1996).

Cortical and subcortical mapping is considered the goldstandard for surgical removal of this type of lesions because itallows the identification and preservation during surgery of corticaland subcortical sites involved in these functions (Bello et al., 2007;Duffau et al., 2002; Keles et al., 2004). Various functional

370 L. Bello et al. / NeuroImage 39 (2008) 369–382

approaches such as fMRI or PET allow to locate functional site inthe pre-operative images and help the surgeon in the surgicalplanning and during surgery as well. A volume of literature isavailable on the correlation between cortical sites identified withfMRI and those found by direct cortical stimulation at the time ofsurgery, showing a variable sensitivity or specificity according tosurgical series or task used (Roux et al., 2001; Schiffbauer et al.,2001, 2002). On the contrary, limited data are available for sub-cortical sites of visualization.

Diffusion Tensor Imaging (DTI) and fiber tractography are MRtechniques based on the concept of anisotropic water diffusion inmyelinated fibers, which enable three-dimensional reconstructionand visualization of white matter tracts (Basser et al., 2000;Beaulieu, 2002; Burgel et al., 2006; Conturo et al., 1999; Mori et al.,2002b; Mori and van Zijl, 2002; Pierpaoli et al., 1996). DTI FiberTracking (DTI-FT) has been used for the visualization of specificfiber bundles, which are in the vicinity of brain lesions, or thoseinfluenced by them (Clark et al., 2003; Jellison et al., 2004; Mori etal., 2002a; Parmar et al., 2004; Schonberg et al., 2006). Space-occupying lesions, particularly infiltrating ones such as gliomas,may exert various effects on white matter tracts, which can bedepicted by DTI-FT. These techniques provide information aboutthe normal course, the displacement, or interruption of white mattertracts around a tumor, and widening of fiber bundles caused byedema or tumor infiltration can be detected (Clark et al., 2003; Moriet al., 2002a; Stadlbauer et al., 2007; Witwer et al., 2002). Somegroups have used DTI tractography to visualize the spatialrelationship between lesions adjacent to the sensorimotor systemand CST for presurgical planning and functional prediction (Hendleret al., 2003; Mori et al., 2002a). Others have used DTI-FT findingsfor intraoperative guidance by loading data provided by DTItractography into the neuronavigation system (Coenen et al., 2001;Nimsky et al., 2005a,b, 2006). In addition, some groups have alsotried to confirm the accuracy of the CST as depicted by DTI-FTusing electrophysiological monitoring (Kamada et al., 2005, 2007;Kinoshita et al., 2005; Mikuni et al., 2007; Okada et al., 2006b).While most studies showed the usefulness of this technique inneurosurgical planning, the validity of the information providedremains to be confirmed, as it has been shown as a potentialunderestimation of actual size of fiber bundles by DTI-FT(Kinoshita et al., 2005); moreover, some caveat have been reportedabout the accuracy of delineation of major white matter tracts, imageregistration and distortion when DTI-FT data were implanted into aneuronavigational system, thus recommending safety margins ofabout 5 mm to be taken into account when approaching eloquenttracts (Nimsky et al., 2005a,b, 2006).

The present study reports our experience in combining 3-T DTIFiber Tracking and intraoperative direct cortical and subcorticalmapping for motor and language functions. 3-T DTI-FT wasroutinely performed in patients harboring low- or high-gradegliomas involving speech or motor areas or pathways, reconstruct-ing the corticospinal tract (CST), and the tracts involved in thesemantic or phonologic loop for language (superior longitudinalfasciculus (SLF), inferior fronto-occipital fasciculus (IFO) anduncinatus (UNC) fasciculus); these datasets were transferred to theneuronavigational system, evaluating the feasibility of thecombined approach in a large series of patients, the reliability ofDTI-FT data in depicting the trajectories of the tracts and theirmodifications induced by the lesions by correlating trackingresults with those obtained during surgery and, finally, the impactof the availability of tractography data on duration and modalities

of surgical procedures, as well as on functional outcome of thepatients.

Material and methods

Patients

From November 2005 to January 2007, 64 patients with gliomaslocated in eloquent areas were submitted to pre-operative DTI fibertractography in our Institutions. Age, gender, clinical history, symp-toms as well as neurological status at admission were registered.Preoperatively, all these patients underwent a neuropsychologicalevaluation, baseline and volumetric MR studies, functional MRimaging (fMRI), and DTI-FT. Surgery was performed in all patientswith the aid of intraoperative cortical and subcortical mapping formotor and/or language functions. Volumetric scan analysis was usedfor establishing tumor location and topography as well as the volumeof the lesion. Tumor location was classified according to previouswork (Bello et al., 2007). Tumor volume was calculated on T2-weighted MRI scans for low-grade gliomas and on post contrast T1-weighted MRI scans for high-grade gliomas via a computerizedsystem (Bello et al., 2007; Duffau et al., 2002; Keles et al., 2004).Histology was classified according to the WHO brain tumor classi-fication. Extent of resection was measured on post-operativebaseline and post contrast MR performed immediately after surgeryor at 3 months, and classified as previously reported (Bello et al.,2007; Duffau et al., 2002; Keles et al., 2004).

Diffusion tensor imaging

MR imaging was performed preoperatively on a Philips Intera3.0-T (Best, The Netherlands) systemwith a maximum field gradientstrength of 80 mT/m. A multi-channel head coil was used for thereception of MR signal. DTI data were obtained using a single-shotecho planar imaging (EPI) sequence (TR 8986 ms; TE 80 ms) withparallel imaging (SENSitivity Encoding factor, R=2.5). Diffusiongradients were applied along 32 axes, using a b-value of 0 and1000 s/mm. A field of view of 240×240 mm2 and a data matrix of96×96 were used and this leaded to isotropic voxel dimensions(2.5×2.5×2.5 mm3). The data were interpolated in-plane to a matrixof 256×256 leading to voxel size of 0.94×0.94×2.5 mm3. Fifty-sixslices were obtained, with a thickness of 2.5 mm, with no gap. Thesequencewas repeated two consecutive times and datawere averagedoff-line to increase signal-to-noise ratio; thus, total time for diffusiontensor MR imaging was 10 min and 46 s. 3D Fast Field Echo (FFE)T1-weighted imaging (TR 8 ms; TE 4 ms; image resolution equal toDTI) was performed for anatomic guidance. Axial turbo-spin-echo(TSE) T2-weighted images (TR/TE 3000/85 ms; field of view[FOV], 230 mm; 22 slices; section thickness, 5/1-mm gap; matrix,512×512; SENSE factor, 1.5), 3D axial Fluid Attenuated InversionRecovery (FLAIR) images (TR/TE 10,000/110 ms; FOV, 230 mm;120 slices; section thickness, 1.5/0-mm gap; matrix, 224×256;SENSE factor, 2) and volumetric Inversion Recovery (IR) T1-weighted images (TR/TE 2000/10 ms; FOV, 230 mm; 22 slices;section thickness, 5/1-mm gap; matrix, 400×512; SENSE factor,1.5) were acquired for morphologic characterization of the lesion.

DTI data processing

DTI datasets were realigned off-line on a PC workstation usingthe AIR (Automatic Image Registration) software to correct artifacts

371L. Bello et al. / NeuroImage 39 (2008) 369–382

due to rigid body movement during scan acquisition (Woods et al.,1998).

Images were analyzed using DtiStudio version 2.02 software(Jiang H, Mori S, Department of Radiology, Johns HopkinsUniversity, Baltimore, MD, USA), and diffusion tensor elementswere calculated and diagonalized at each voxel, obtaining maineigenvector and fractional anisotropy (FA) maps. From theircombination, color maps were generated with conventional color-coding (Pajevic and Pierpaoli, 1999).

Fiber tracking

The CST and the subcortical tracts involved in the phonologic(SLF), or semantic loop (IFO, UNC) were reconstructed using theDtiStudio software, employing the “fiber assignment by continuoustracking” (FACT) method (Mori et al., 1999; Mori and van Zijl,2002; Xue et al., 1999). The eigenvector associated with the largesteigenvalue was assumed to represent the main fiber direction in thevoxel. An FA threshold of 0.1 and a turning angle N55° were used ascriteria to start and stop tracking.

Seed ROIs were chosen on the basis of previous anatomicknowledge, in sections perpendicular to the main course of the tracts(Wakana et al., 2004).

To reconstruct the CST, a ROI was placed on an axial section atthe level of subcortical white matter of the precentral gyri. For theIFO and UNC, a ROI was placed on a coronal section at the level ofthe anterior part of the external capsule at the junction of the frontaland temporal lobes, where the two tracts run in contiguity. Toreconstruct the SLF, a first ROI was placed on a coronal section at thelevel of a high-anisotropy region laterally to the central part of thelateral ventricle; a second ROI was placed in a peritrigonal site at thelevel of the descending branch of the tract.

For all the tracts reconstructed, eventual contaminating fiberswere removed.

Finally, volumetric pre-contrast T1-weighted or FLAIR imageswere coregistered to the mean of all diffusion-weighted imagesusing SPM2 software to obtain the superimposition of the whitematter tracts on T1-weighted anatomical images. This allowed tocompare the trajectories of the tracts in the involved hemisphere withthose of the contralateral unaffected hemisphere, and to evaluate theanatomical relationship between the tract and the tumor mass as wellas the effect exerted by the tumor on the tract of interest. Tracts werethen classified as unchanged, dislocated, or infiltrated/interrupted, asdescribed in previous works (Jellison et al., 2004; Witwer et al.,2002).

DTI-FT data transfer to neuronavigational system

DTI-FT data were saved as a compatible format (DICOM) to betransferred to Neuronavigational System (CRW, Radionics Inc.,Burlington, MA) using Medx Software (Medical Numerics, Inc.).The Neuronavigational System performed an automatic coregistra-tion between DTI-FT datasets and the preoperative MR imagesacquired with references applied on the skull of the patient by avoxel-by-voxel intensity matching non-linear algorithm. The resultwas the fusion in the same space of anatomical and functionaldatasets. As estimate for the clinical navigation accuracy, the targetregistration error localizing a separate fiducial, which was not usedfor registration, was documented. Furthermore, repeated landmarkchecks were performed during surgery to ensure overall ongoingclinical navigation accuracy.

Preoperative language neuropsychological evaluation

Patients were submitted to a battery of tests aimed to evaluateoral language production and comprehension, together withrepetition (Bello et al., 2007). Language dominance was evaluatedas well by submitting patients to the Edinburgh InventoryQuestionnaire and to fMRI. The following tasks were performed:Spontaneous speech; Oral-controlled Association by PhonemicCue; Famous face naming; Object Picture Naming; Action PictureNaming; Word Comprehension; Sentence Comprehension; Trans-coding tasks. In addition, the Token Test, the digit span, andcounting were performed. Ideomotor apraxia and face apraxia werealso assessed. The majority of the tests we used have beenstandardized on the normal population.

Intraoperative testing and stimulation protocol

Brain mapping was undertaken during asleep craniotomy whenonly a motor mapping was performed, or under asleep awakeanesthesia when motor, language, and/or visuospatial functionswere tested (Bello et al., 2007; Duffau et al., 2002). A neuronaviga-tional system was used in all the cases to define tumor location, todesign the bone flap, and to determine tumor area and its boundarieswhen the dura was opened. The threshold of current for evoking abrief afterdischarge in the electrocorticogram was established invarious points in the area of cortex to be mapped by the use of anOjemann cortical stimulator. Mapping was performed by using thelargest current that did not produce afterdischarge (ranging from 3.5to 16 mA). Subcortical stimulation by using the same currentthreshold was also applied during tumor resection, in a back andforth fashion (Bello et al., 2007; Keles et al., 2004). Throughout thecortical and subcortical language mapping, the electrocorticogramwas continuously monitored to signal afterdischarge spikes, both toreduce the chances of evoking a seizure by continued stimulation atthat current, and to avoid the chance that naming errors are caused bythe propagated effects of the current. We planned to resect tumortissue no more than 1 cm from eloquent cortex and preserving thesubcortical tracts, in order to maximally preserve functions. Thefinding of subcortical motor or language tracts constituted the limitof resection.

Motor mapping was performed at both cortical and subcorticallevel (Intraoperative subcortical mapping, ISM). Motor responseswere evaluated both clinically and by the use of a multichannelEMG recording.

Language mapping was also performed at both cortical andsubcortical level. For frontal lesions, Broca and oral namingessential sites were identified. Broca area was located as speecharrest during counting without simultaneous motor responses in themouth. To identify the sites involved in oral naming, the namingtasks (object picture naming, action picture naming and famouspeople naming) were repeated three (non-consecutive) times. Forlesions located in temporal areas, sites involved in oral naming,sentence and word comprehension were identified. Oral namingtasks (object picture naming, action picture naming and famouspeople naming), word comprehension and sentence comprehensiontasks, were repeated three (non-consecutive) times. During theintraoperative session we used blocks composed of 80 items com-posed of those items, which were correctly performed without anylatency three times out of three during the pre-operative languageexamination. Only positive sites in a reproducible errors wasrecorded were mapped with sterile numbered tickets. Subcortical

372 L. Bello et al. / NeuroImage 39 (2008) 369–382

stimulation was performed during oral naming or word comprehen-sion tests by using the same current threshold used for the corticalmapping. Subcortical stimulations was alternated with surgicalremoval and used when the resection became close to thesubcortical structures located in proximity to the cortical languagesites, all around the surgical cavity, and in its boundaries. In case oferrors, the patient was asked to rest and recover. Stimulation wasthen reapplied. Only those sites which stimulation induced a re-producible error at least three times were considered as subcorticalsites.

Correlation between DTI-FT data and perioperative subcorticalmotor and language findings

Correlation between DTI-FT data and subcortical sitesidentified during perioperative direct electrical stimulation wasperformed at the time of surgery and post-operatively. At the timeof surgery, the location of each subcortical site was recordedduring the various phases of resection by the use of neuronaviga-tion system on the preoperative MR scan in which DTI-FT datawere loaded. The distance between the point of response at ISMand the tract border was measured on transverse navigation imagesand was graded as correspondent (at the tract border or within thetract), close (within 1 cm) or distant (N1 cm). In addition, eachsubcortical site identified during subcortical mapping was markedwith a sterile numbered tag and a digital picture of the surgicalcavity was taken at the end of the resection. On the immediatepost-operative MR scan, the anatomical location of the subcorticalpathways was evaluated, i.e. the periphery of the surgical cavity,where the resection was stopped according to the functionalresponses elicited by intraoperative stimulation (Duffau et al.,2002). In addition, post-operative DT imaging was also performedin 10 representative patients. These were 3 patients operated for anF3 tumor, 3 patients operated for a dominant rolandic tumor, 2patients for a non-dominant rolandic tumor, and 2 patients operatedfor a left parietal tumor. These cases were chosen because due tothe location of the tumor, they allowed to reconstruct in the samecase the CST and most of the language tracts which were identifiedby ISM during surgery at the border of the surgical cavity (CST,SLF, IFO, and UNC in F3 tumors, CST and SLF in dominantrolandic tumors; CST in non-dominant rolandic tumors; CST, SLF,and IFO in left parietal tumors). Mean duration between surgeryand DTI images was 40 days (34–65 days). DT imaging sequenceand methods for tractography were equivalent to those forpreoperative tractography. The distance between the border ofthe tract and the nearest portion of the resection margin wasmeasured on transverse MR scan in which DTI-FT images weresuperimposed. Distances were graded as previously reported. Inaddition, pre- and post-operative tractography images were com-pared (Okada et al., 2006a,b).

The existence of a correspondence between the location of asubcortical site identified by ISM and that of the tract in DTI-FTimages was considered only when the distance between the locationof the subcortical site evoked by ISM and that identifiedintraoperatively by the use of neuronavigation system was gradedas adjacent to the tract in transverse neuronavigation images. In allthe other cases, the correspondence was considered null. Datagenerated intraoperatively were overlaid with those obtained onpost-operative scans after adjustment by visual inspection to correctfor small distortions inherent to intraoperative shift, to furtherconfirm these data.

Effect of DTI-FT on surgical procedures and patient functionaloutcome

The effect exerted by the availability of DTI-FT on the surgicalprocedure was studied by monitoring the duration of the time ofresection comprehensive of the duration of the subcortical mapping.In addition, we also looked at the occurrence of afterdischarge,electrical seizures, and clinical seizures comparing the frequency ofthese events during the procedures performed with the aid of DTI-FT to those performed by our group without tractography.Furthermore, for the cases in which the mapping was performedunder awake anesthesia, we also looked at the patient fatigue,defined as the need to stop the surgical procedures at least 5 timesfor at least 10 min due to patient fatigue and decreased ability tocollaborate and performing tasks. Patient functional outcome wasalso evaluated by submitting patient to neurological examinationand neuropsychological tests at 3 days, and 1 and 3 months aftersurgery.

Statistical analysis

Data were analyzed by the use of Prism 4 for Macintosh(GraphPad Software, Inc).

Differences in the duration of resection expressed as mean andstandard deviation were studied with t test. Differences in percentageof clinical and electrical seizure and of patient fatigue were studiedwith the Fisher’s exact test.

Sensitivity of DTI-FT for the identification of subcortical tractswas calculated as the ratio between the number of sites graded ascorrespondent in transverse neuronavigation images when locationof subcortical sites in ISM was reported to DTI-FT images (Truepositive – TP) and total number of sites as identified by ISM alone(TP+False Negative – FN). The estimation of specificity in thissetting can only be given as an approximate value. In fact,specificity is the ratio between the number of negative sites asidentified by ISM which corresponded to the absence of tract inDTI-FT images (True Negative – TN) and number of all negativesites identified with ISM (TN+False Positive – FP). DTI-FTprovides anatomical information and not functional ones. If thisissue is of relative importance for CST, it is of particular relevancefor language tracts. In fact, not all the tract as depicted in DTI-FTimages could be functionally relevant for language. This is clearlyaffecting the significance of FP, i.e. sites in which DTI-FT showedthe tract and in which ISM did not document any functionalresponse.

Results

Patients’ characteristics

Sixty-four patients were submitted to combined DTI-FT andISM. They were 40 males and 24 females, age ranging between 28and 58 years. All patients had a clinical history of a partial seizure. Atadmission, a mild motor or language deficit was observed in 6 and 5patients, respectively. In 23 cases, the location of the tumors wasrolandic, in 16 prerolandic, in 3 paralimbic, in 6 parietal, and in 16temporal. The mean calculated volume was 40,455 cm3 (SD ±33,894, median 34 cm3, min 11 cm3, max 161.8 cm3). 52 patientshad a low-grade glioma (oligodendroglioma or astrocytoma), 12 ahigh-grade glioma (anaplastic oligodendroglioma, anaplastic astro-cytoma, glioblastoma).

373L. Bello et al. / NeuroImage 39 (2008) 369–382

Fiber tracking and correlation with perioperative subcorticalmapping data

DTI was successfully performed in all 64 patients; DTI datasetscould be post-processed in all the cases, allowing the reconstructionof corticospinal and language tracts in all patients, independentlyfrom tumor location. Integration of preoperative fiber-tracking datain the neuronavigation system was technically feasible in allpatients. As estimate for the clinical navigation accuracy, the targetregistration error localizing a separate fiducial, which was not usedfor registration, was documented. The target registration error wasless than 2 mm (1.1±0.34 mm). During surgery, according to tumorlocation, all patients were submitted to cortical and subcorticalmotor and/or language mapping.

According to data reported in the literature, an FA threshold of0.2 was selected to start and stop the reconstruction algorithm for theeloquent tracts in the first 12 patients; in six of these an incompletecorrespondence was observed between tractography data andintraoperative subcortical mapping results, independently frombundle or tumor location. Therefore, data were reanalyzed with alower FA threshold (0.1): this allowed a better correspondence withISM results (Fig. 1), consequently a 0.1 FA threshold was adopted toanalyze subsequent DTI data.

Corticospinal tract (CST)

Table 1 shows CST findings according to tumor location.The CST was reconstructed in 57 patients. The correspondence

between DTI-FT and ISM was verified in 34 cases out of 57,because in this subgroup of patients the tumor was adjacent orinfiltrating the motor tract. Specifically it was verified in all 23rolandic tumors, in 5 out of 16 prerolandic, in 4 out of 6 parietal

Fig. 1. A case of left parietal low-grade glioma with infiltration of the motor area: Ahas been reconstructed with FA threshold of 0.2: it is depicted as dislocated antemapping identifies motor responses corresponding to the right leg with a good corarm and hand (red arrow) in an area of absence of fibers. (B) DTI data have been relacking fibers: a complete correspondence between DTI datasets and subcortical s

tumors, in 1 out of 9 temporal tumors, and in 1 out of 3 paralimbictumors. In the remaining cases, the trajectory of CST was far awayfrom the tumor border, and the correspondence with ISM in thesecases could not be verified. ISM identified 229 subcortical motorsites. The location of 218 sites was graded as correspondent to theCST. Eleven sites were identified in subcortical areas where notract was reconstructed by DTI-FT (FN). These belonged to apatient with a rolandic tumor (3 sites) and a patient with a leftparietal tumor (8 sites). Details for these cases are presented below.Sensitivity for motor tract identification was 95%. As forspecificity, no negative motor sites were identified by ISM in thelocation were DTI-FT depicted the CST (FP=0). Therefore, thespecificity for CST can be approximated to 100%.

In patients with a rolandic tumor, pre-operative DTI-FT showedCST at the posterior border of the tumor, mainly infiltrated (56.5%)or dislocated (34.7%) by the tumormass. Less frequently (8.6%), theCST was adjacent to the posterior border and unchanged by thetumor, which had a small volume. ISM identified the CST in allthe cases: when the CSTwas documented by DTI-FT adjacent to thetumor border, or infiltrated or interrupted and inside the tumor mass,ISM identified the CST in the same location. In addition, ISMshowed that the CST had a similar extension along the surgicalcavity as depicted by DTI-FT (Fig. 1B and Supplementary data).Only in one case that ISM documented 3 motor sites inside asubcortical area of the tumor in which no tracts were identified byDTI-FT even after the reduction of FA threshold to 0.1 (Fig. 2).

In the group of patients with a tumor involving the prerolandicregion, CST was depicted by DTI-FT as unchanged and notadjacent to the tumor mass in 68.7% of cases. In these small sizetumors, ISM did not document any motor response at the peripheryof the tumor. In the remaining 5 patients in whom the tumorvolume was larger, DTI-FT showed CST dislocated posteriorly,

xial T1-weighted images with superimposed CST fibers (yellow). (A) CSTriorly and interrupted in the anterolateral portion of the tumor. Subcorticalrespondence with DTI-FT CST reconstruction (white arrow) and to the rightanalyzed with FA threshold of 0.1 allowing the reconstruction of previouslytimulation has been reached (white arrows).

Table 1Corticospinal tract (CST): DTI Fiber Tracking data, concordance with intraoperative subcortical mapping and post-operative deficits

Location Pre op DTI N N Intraop Concordance Deficits

YES NO Early Late

Rolandic, N=23 Unchanged 2 2 2 – 2 –Dislocated 8 8 8 – 8 –Infiltrated 13 13 12 1 12 –

Prerolandic, N=16 Unchanged 11 – – – – –Dislocated 5 5 5 – 5 –Infiltrated – – – – – –

Paralimbic, N=3 Unchanged 2 – – – – –Dislocated 1 1 1 – – –Infiltrated – – – – – –

Parietal, N=6 Unchanged 2 – – – – –Dislocated 2 2 2 – 2 –Infiltrated 2 2 2 a – 2 –

Temporal, N=9 b Unchanged 6 – – – – –Dislocated 3 1 1 – 1 –Infiltrated – – – – – –

Data are presented according to location. N=number of cases observed. The modification of the tract in relationship to the tumor mass as depicted by DTI FiberTracking, is indicated as Unchanged, Dislocated, or Infiltrated/Interrupted. The second column (N Intraop) reports the number of cases in which DTI FiberTracking depicted the tract as Adjacent or inside the tumor mass, in which it was possible to directly test the concordance with subcortical mappingintraoperatively. Concordance is described as YES when there is concordance between the location of the tract as documented by DTI and subcortical mapping.NO indicates absence of concordance.a The case (left parietal) in which a good concordance was reached by the modification of the FA threshold. Post-operative deficits are indicated as early (at

3 days) and late (1 month after surgery).b Number of patients with temporal tumors in which, due to the posterior and deep extension of the tumor, the CST was reconstructed.

374 L. Bello et al. / NeuroImage 39 (2008) 369–382

and adjacent to the tumor border, in the same location where ISMidentified motor responses.

In all 6 patients with a parietal lobe tumor, DTI-FT located theCSTadjacent to the anterior border of the tumor. In small tumors, thetract was unchanged. In middle-sized tumors, the tract wasdislocated, or infiltrated and interrupted and inside the tumor mass.In the patients with unchanged or dislocated CST, the tract waslocated by ISM in a similar position (Fig. 3). In one case, ISMidentified 8 motor sites corresponding to the right arm and hand inareas of absence of fibers. The reanalysis of the DTI images with FAthreshold of 0.1 allowed the reconstruction of previously lackingfibers and to reach a complete correspondence between DTI datasetsand ISM findings (Fig. 1).

CSTwas reconstructed also in 9 cases of temporal tumors, whichhad a posterior and deep extension. In one case, in which thereconstructed bundle appeared adjacent to the tumor mass, whereISM evoked motor responses.

In one of the cases of paralimbic tumors, CST was depicted byDTI-FTas dislocated and in relationship with the posterior border ofthe tumor mass. DTI-FT evoked motor responses in the sameposition.

Language tracts

Superior longitudinal fasciculus (SLF)SLF was found unchanged in small tumors (11.6%), dislocated

(16.6%) or infiltrated and interrupted (66.8%) in those with a middleor large volume. Table 2 shows SLF findings according to tumorlocation.

ISM induced phonemic paraphasias in 169 subcortical sitescorresponding to the reconstructed position of the SLF. Theanatomical distribution of SLF as depicted by DTI-FT was larger

than the functional one identified by ISM, as a large part of theunchanged SLF was functionally silent in terms of languagefunction, allowing the removal during resection. When functionalsubcortical areas were found, their location was graded ascorrespondent to fibers depicted by DTI-FT, either in cases ofdislocated, infiltrated, or unchanged tract. Moreover, in cases of SLFlocated inside the tumor mass with FT interruption, ISM evoked nolanguage responses in areas of lacking fibers. A mismatch betweenDTI-FT and ISM was observed in 3 cases out of 8 tumors involvingF3 gyrus, in which DTI-FT failed to demonstrate fibers at theanterior border of the tumor, in an area corresponding to the anteriorportion of the fasciculus: ISM evoked in these areas functionalresponses (Fig. 4). This observation induced a revision of SLFreconstruction modalities: the positioning of a further ROI in thesubcortical area of F3 gyrus allowed a complete correspondencebetween FTand stimulation data in all the cases involving this region(Fig. 5). Therefore, sensitivity for SLF was 98.2%. As forSpecificity, in 12 cases, ISM did not evoke responses were DTI-FT showed the presence of the tract (FP). Nevertheless, not all theSLF is functionally relevant for language. Therefore, specificitycannot be estimated.

Inferior fronto-occipital fasciculus (IFO)This bundle appeared as unchanged in 19.4% of the cases (small

tumor volume), dislocated in 41.6%, infiltrated and interrupted in38.8% of patients (middle or large size tumors). Table 3 shows IFOfindings according to tumor location.

ISM induced semantic paraphasias in 26 subcortical sitescorresponding to the reconstructed position of the tract. Twenty-three sites were graded as correspondent to the tract, either dis-located, infiltrated, or unchanged (Fig. 5D). ISM evoked nolanguage disturbances in areas of lacking fibers when the tract was

Fig. 2. A case of right rolandic low-grade glioma infiltrating the motor area. (A, B) Axial T1-weighted images with superimposed CST fibers (yellow). The centerdepicts the intraoperative photograph of the resection cavity. The white tags represent the cortical and subcortical sites at which electrical stimulation elicitedmotor responses during surgery. In Fig. 1A, FA threshold of 0.2 fails to reconstruct CST in the posterior area of the tumor where ISM finds functional responsescorresponding to left arm and fingers (blue arrows). Good correspondence is shown in other areas corresponding to the leg and arm (white arrows). In Fig. 1B, thereduction of FA threshold to 0.1 leads only to a partial correction (white arrow), as the lesion determines a marked reduction of FA below the selected threshold.

375L. Bello et al. / NeuroImage 39 (2008) 369–382

depicted as interrupted inside the tumor mass. A mismatch wasobserved in 3 of 8 cases of F3 tumors, in which semantic paraphasiaswere induced in a subcortical area located inferiorly to the tumormass, where no fibers were shown by DTI-FT. In these cases, DTI-FT showed fibers just below the functional site (Fig. 6). Sensitivitywas 88.6%. As for specificity, no negative language sites wereidentified by ISM in the location were DTI-FT depicted the presenceof the IFO (FP=0). Therefore, the specificity for IFO can beapproximated to 100%.

Uncinatus fasciculus (UNC)The uncinatus was depicted as unchanged in 26.4% of cases

(small tumor volume), infiltrated in 36.8%, and interrupted in36.8% (middle or large size tumors).

ISM induced semantic paraphasias in 9 subcortical sitescorresponding to the reconstructed position of the tract. In 8 thelocation was graded as correspondent to the tract.When the tract wasfound interrupted and located inside the tumor mass, no languageresponses were evoked by ISM. A mismatch was observed in onecase: a temporal tumor in which DTI-FT showed some infiltratedfibers inside the tumor mass. At surgery, ISM evoked no languagedisturbances in the same area. Sensitivity was 89%. As for speci-

ficity, no negative language sites were identified by ISM in thelocation were DTI-FT depicted the presence of the UNC (FP=0).Therefore, the specificity for UNC can be approximated to 100%.

DTI-FT, subcortical mapping and histology

In high-grade gliomas, DTI-FT showed that tracts were mostlydislocated (50%) or infiltrated and interrupted (37.5%) by the tumormass, and no tracts were documented inside the tumor mass.Subcortical mapping demonstrated functional subcortical areas inthe same location, specifically at the tumor border. In addition, whenthe tract was found interrupted, no response was evoked by ISM inthe same areas.

In low-grade gliomas, DTI-FTshowed that the tracts weremostlyinfiltrated and interrupted (62.5%) or dislocated (25%) by the tumormass. In addition, 50% of the tracts were documented inside thetumor mass. ISM demonstrated functional areas either at the tumorborder or inside the tumor mass. When functional subcortical areaswere documented at the tumor border by ISM, these areas cor-responded to those documented as infiltrated by DTI-FT. Whentracts were found interrupted inside the tumor mass, in some cases,ISM documented the presence of functional areas.

Fig. 3. A case of right parietal low-grade glioma with anterior dislocation of CST: snapshots from neuronavigational system with bilateral CST fibers (white)superimposed on post contrast T1-weighted images. Panel A shows positive correspondence between DTI and subcortical mapping whereas panel B showsnegative correspondence in an area in which both subcortical mapping and fiber tracking do not identify motor functional fibers. (C) A comparison between pre-(left column) and post-operative (right column) T1 post contrast weighted images shows the respect of the functional boundaries for resection.

376 L. Bello et al. / NeuroImage 39 (2008) 369–382

Effect of the combination of DTI-FT and ISM on surgicalprocedure and patient functional outcome

The mean duration of the resection, comprehensive of sub-cortical mapping (excluding craniotomy and closure), decreasedfrom 1 h and 40 min to 1 h and 3 min when subcortical motormapping was performed (pb0.0001), from 2 h and 30min to 1 h and10 min when language and motor mapping were applied(pb0.0001).

A decrease in the number of electrical seizure (30% vs. 15.6%),with a significant difference when considering clinical seizure(22.5% vs. 7.8%) was observed (p=0.0662 and p=0.0306,respectively). In addition, in the cases in which the mapping wasperformed under awake anesthesia, a significant decrease (from28.6% to 6.6%, p=0.301) in patient fatigue was also documented.

94% and 96% of patients in which a motor or language tract wasidentified by ISM at the time of surgery experienced a decrease inmotor or language function immediately after surgery, which lastedfor 1 week on the average. At 1 month of follow-up, 89% ofpatients with a motor lesion had a normal motor exam and 88% ofthose with lesions involving speech areas or pathways had a normallanguage.

Discussion

This study demonstrates the feasibility of the combined approachof DTI-FT and ISM in a large series of patients undergoing gliomasurgery, and the effective reliability of tractography data by showinga good congruency between the two methods. Furthermore, thepositive impact of the availability of DTI-FT datasets in theneuronavigational system on the surgical procedures has beenunderlined by the decrease of the duration of the procedures and ofpatients’ fatigue, as well as the occurrence of clinical seizures.

Most of previous studies dealing with visualization of fiber tractsin neuronavigational setups combined with ISM were performedusing 1.5-T machines (Kamada et al., 2005; Kinoshita et al., 2005;Nimsky et al., 2005a, 2006). We performed all DTI studies using a3.0-T system: this choice was based on the fact that higher magnetstrength increases the signal to noise ratio and improves thedemonstration of fiber tracts (Hunsche et al., 2001; Jaermann et al.,2004; Nagae-Poetscher et al., 2004; Okada et al., 2006a). To ourknowledge, there is only one study of tractography combined withISM in which DT images were obtained using a 3.0-T system: itfocused on CST visualization, performed in a small number ofpatients (Okada et al., 2006b). The strength of the present study is the

Table 2SLF in left tumors

Location Pre op DTI N NIntraop

Concordance Deficits

YES NO Early Late

Rolandic, N=11 Unchanged 4 4 4 – 4 –Dislocated – – – – – –Infiltrated 7 7 7 – 7 –

Prerolandic, N=16 Unchanged 3 2 2 – 2 –Dislocated 1 1 1 – 1 –Infiltrated 12 12 12 a – 12 –

Temporal, N=15 b Unchanged – – – – – –Dislocated 6 6 6 – 6 –Infiltrated 9 9 9 – 9 –

DTI Fiber Tracking data, concordance with intraoperative subcortical mapping and post-operative deficits. Data are presented according to location. N=numberof cases observed. The modification of the tract in relationship to the tumor mass as depicted by DTI Fiber Tracking, is indicated as Unchanged, Dislocated, orInfiltrated/Interrupted. The second column (N Intraop) reports the number of cases in which DTI Fiber Tracking depicted the tract as Adjacent or inside the tumormass, in which it was possible to directly test the concordance with subcortical mapping intraoperatively. Concordance is described as YES when there isconcordance between the location of the tract as documented by DTI and subcortical mapping. NO indicates absence of concordance.a Three cases in which a good concordance was reached by the modification of the FA threshold. Post-operative deficits are indicated as early (at 3 days) and

late (1 month after surgery).b Number of patients with a left temporal tumor in which the tract was reconstructed.

377L. Bello et al. / NeuroImage 39 (2008) 369–382

application of 3.0-T DTI-FT on a large and homogeneous group ofpatients, evaluating both motor and language functions. Higherspatial resolution DTI-FTwith a short acquisition time at 3.0-T MRwith parallel imaging (Jaermann et al., 2004) allowed to implement aclinically feasible study routinely applicable. In all patients, high-quality images were obtained to be processed by means of DTI-FT;the CST and the SLF, IFO and uncinatus fasciculus, were re-constructed by using the “fiber assignment by continuous tracking”(FACT) method (Mori et al., 1999; Mori and van Zijl, 2002; Xue et

Fig. 4. Two cases of F3 low-grade gliomas: SLF fibers (yellow) superimposed onpanels A and B depicts the intraoperative photograph of the resection cavity in thestimulation elicited language responses (phonemic paraphasias) during surgery. In bwhere subcortical mapping elicits phonemic paraphasias or anemia (blue arrowspositioning of a further ROI (red) anteriorly to the tumor showing some infiltrated

al., 1999), up to now the more clinically used of all tractographyreconstruction algorithms. According to the data in literature, a FAthreshold of 0.2 was chosen to start and stop tracking in the set of thefirst 12 patients. The correlation with ISM data showed anincomplete correspondence between the two methods: after thispreliminary result, the FA threshold was lowered to the lowest limitempirically chosen in order to avoid the inclusion of non-pertinentstructures (i.e. CSF, gray matter). In most of previous DTI studies,the FA threshold values used for fiber tract reconstruction were

3D FLAIR (A, B) and on T1-weighted images (C, D). The picture betweenfirst case. The white tag represents the subcortical site at which subcorticaloth cases, DTI failed to demonstrate fibers in the anterior portion of the tumorin panels A and C). (B, D) Exact correspondence was reached through thefibers connected to the main tract (red arrow).

Fig. 5. A case of left F3 low-grade glioma: snapshots from the neuronavigational system with SLF fibers (white) superimposed on T1 post contrast weightedimages and intraoperative picture. The picture in the center depicts the intraoperative photograph of the resection cavity. The white tags represent the subcorticalsites at which subcortical stimulation elicited language responses during surgery. A complete correspondence between SLF (A, B, C) and IFO-UNC (D) fibersand functional areas (inducing respectively phonemic and semantic paraphasias) in superficial and deep portions of peritumoral white matter is shown. (E) Acomparison between pre- (left column) and post-operative (right column) T1 post contrast weighted images shows the respect of the functional boundaries forresection.

378 L. Bello et al. / NeuroImage 39 (2008) 369–382

chosen rather arbitrarily, ranging from 0.2 up to 0.3 (Jellison et al.,2004; Mori et al., 2002b; Stieltjes et al., 2001; Yamada et al., 2003).Recently, Stadlbauer et al. thoroughly evaluated the effects ofdifferent FA thresholds on the reconstruction of fibers in the vicinityof gliomas by correlating fiber tracking results with histopatholo-

Table 3Inferior fronto occipitalis fasciculus in left tumors

Location Pre op DTI N NIntraop

Parietal, N=6 Unchanged 1 –Dislocated 3 –Infiltrated 2 2

Prerolandic, N=14 Unchanged 6 2Dislocated 4 4Infiltrated 6 6

Temporal, N=14 c Unchanged – –Dislocated 8 8Infiltrated 6 6

DTI Fiber Tracking data, concordance with intraoperative subcortical mapping andof cases observed. The modification of the tract in relationship to the tumor mass aInfiltrated/Interrupted. The second column (N Intraop) reports the number of cases imass, in which it was possible to directly test the concordance with subcorticalconcordance between the location of the tract as documented by DTI and subcorta Two cases in which semantic paraphasias were induced during surgery by elec

DTI fiber tracking failed to show the IFO. The tract was shown more inferiorly andb A case in which semantic paraphasias were induced during surgery by electrica

fiber tracking failed to show the IFO. The tract was shown more inferiorly andoperative deficits are indicated as early (at 3 days) and late (1 month after surgerc Number of patients with a left temporal tumor in which the tract was reconst

gical findings from stereotactic biopsies (Stadlbauer et al., 2007);they demonstrated that an increase in the FA threshold increases thedistance between the reconstructed fiber bundles and the tumor and,conversely, reduces the number of detectable fiber bundles, so an FAthreshold in the range of 0.15 to 0.2 represent the best compromise to

Concordance Deficits

YES NO Early Late

– – – –– – – –2 – 2 –– 2 a – –4 – 4 –5 1 b 6 –– – – –8 – 8 –6 – 6 –

post-operative deficits. Data are presented according to location. N=numbers depicted by DTI Fiber Tracking, is indicated as Unchanged, Dislocated, orn which DTI Fiber Tracking depicted the tract as Adjacent or inside the tumormapping intraoperatively. Concordance is described as YES when there isical mapping. NO indicates absence of concordance.trical stimulation in the inferior medial border of F3 tumors in a place whereunchanged. These patients did not experience post op semantic paraphasias.l stimulation in the inferior medial border of F3 tumors in a place where DTIinfiltrated. This patient did experience post op semantic paraphasia. Post-y).ructed.

Fig. 6. A case of left F3 low-grade glioma: IFO fibers (yellow) aresuperimposed on T1-weighted images. Subcortical mapping elicits semanticparaphasias inferiorly to the tumor, in areas of absence of fibers (red crosses):DTI depicts IFO fibers just below functional areas. This is explained by brainshift occurring in deep regions after tumor resection.

379L. Bello et al. / NeuroImage 39 (2008) 369–382

gain fiber tract reconstructions that reflect the structural situation inand around an infiltrating intrinsic brain tumor. However, thisretrospective study did not investigate the functionality of the fiberbundles: in the present study a valid reconstruction of these func-tional fibers with a trackability threshold of 0.1 is demonstrated, andthis is the first result that underlines the importance of the combinedapproach to evaluate the reliability of tracking from a functionalpoint of view.

Some authors have reported the combination of cortical andsubcortical stimulation and corticospinal tractography (Berman etal., 2004; Kamada et al., 2005; Kinoshita et al., 2005; Okada et al.,2006b) integrated in a neuronavigational system; intraoperativesubcortical white matter stimulation during awake surgery hasrecently been applied in the evaluation of fronto-temporal languagefibers depicted by fiber-tracking in two cases of gliomas located inthe dominant fronto-temporal area (Kamada et al., 2007). In thepresent study, fiber tracking data were systematically transferred intothe neuronavigational system to correlate tractography results anddata obtained during perioperative subcortical motor and languagemapping. One of the major drawbacks affecting the reliability ofDTI-FT in functional neuronavigation is intraoperative brain shifteffect, well-investigated by other groups (Nimsky et al., 2005a,b)who recommended a safety margin of about 5 mm whenapproaching, for example, the pyramidal tract. Considering thisproblem, repeated landmark checks were performed during surgeryto ensure overall ongoing clinical navigation accuracy. In addition,resection was attempted in order to maximally ensure themaintenance of accuracy of the neuronavigation system. In case offrontal tumors located in the proximity of CST, resection was startedfrom the posterior border where the CST was located and, after itsidentification, the tract was followed inside the tumor mass.Afterwards the remaining anterior part of the tumor was removed.

Similarly, in case of parietal tumors, resection was started from theanterior border following the same principle. This reduced the brainshift in the most critical areas of the tumor. Indeed, the accuracy ofDTI-FT loaded on neuronavigation system was higher in case ofsmall or medium size tumors, which represent the majority of thoseincluded in this study. In these cases, the use of a craniotomy limitedto the minimum necessary to expose the tumor area and a limitedportion of the surrounding brain, allowed to minimize brain shift andto maintain the accuracy of the information loaded onto theneuronavigation system. Furthermore, post-operative DTI-FTimages showed that resection margins were coincident with thetracts (see Supplementary data). Although the absolute amount ofshifting correlated with the tumor volume, and in small tumors greatdeformations are unlikely to occur, inward or outward shifting ofmajor white matter tracts during surgery seems to be unpredictable(Nimsky et al., 2005b). The use of intraoperative MR DTI-FT cancompensate for these changes (Nimsky et al., 2006).

The location of the CST as depicted by DTI-FT depended ontumor location and volume. Generally, when the tract was in closerelationship with the tumor ISM identified the tract in the sameposition, with similar extension both in cases of dislocated tractsand in cases of infiltrated/interrupted tractography. Nevertheless, inspite of the relatively low anisotropy threshold used in this study, inone case functional motor fibers were found by mapping insubcortical areas of the tumor where DTI-FT failed to demonstratethe presence of fibers. Adding a ROI in the area of lacking fibersproduced only a partial correction of the correspondence, as thepoint-by-point anisotropy values in the tumor area were all belowthe threshold of 0.1, thus indicating that interruption of tractographydoes not always mean real disruption of fibers. Indeed, thedisorganization of normal architecture of myelinated fibers by theneoplastic growth without disruption of bundles could determine alowering of FA values below the usual values selected for startingand terminating tracking (Goebell et al., 2006; Inoue et al., 2005;Stadlbauer et al., 2006; Tropine et al., 2004; Wieshmann et al.,1999; Witwer et al., 2002). The opposite effect of fiber tractsdepicted by DTI-FT inside the tumor mass and not documented byISM was very rare, despite of the low threshold used; in the onlycase observed these interrupted fibers could be safely resected.These previous examples underlines the importance of thecombined approach with direct subcortical stimulation to reach asafe and effective resection of the tumor in those cases in whichDTI-FT results are of non-univocal interpretation.

Similarly to the CST, the trajectory and size of language tractswere also influenced by tumor volume and location. In addition, thecorrelation between DTI-FT reconstruction and ISM data dependedon the location, size and function of the tract. A Large tract as SLFwas depicted by DTI-FT as dislocated or infiltrated by the tumor.When language sites were identified, these corresponded to the areasdepicted by DTI-FT; likewise, no language responses were evokedin areas of lacking fibers in cases of SLF located inside the tumormass with FT interruption. Smaller tracts as IFO or UNC weremostly depicted by DTI-FT as dislocated or unchanged, in the samesites in which ISM elicited semantic paraphasias.When the tract wasdepicted inside the tumor mass with FT interruption, no responseswere elicited by ISM in areas of lacking fibers. Globally considered,these data indicate a good concordance between the trajectory ofthese tracts as reconstructed by DTI-FT and ISM.

An additional result of the present analysis is the concordancebetween DTI-FT data and subcortical mapping findings with tumorgrade. In high-grade gliomas, tracts were mostly depicted by fiber

380 L. Bello et al. / NeuroImage 39 (2008) 369–382

tracking as unchanged or dislocated by the tumor mass, and in closerelationship with its border. In low-grade gliomas tracts weretypically located by DTI-FT inside the tumor mass and frequentlyinfiltrated. Prevalent displacement of fiber bundles by high-gradelesions has been previously reported (Clark et al., 2003; Holodny etal., 2001; Tropine et al., 2004), as well as infiltration in low-gradegliomas (Jbabdi et al., 2005). Data shown in the present study seemto be reproducible and in good concordance with these works;however, as concern the interpretation of interruption of trackingaccording to histological findings, whereas a good correspondencebetween tractography data and ISM was shown in high-gradegliomas, in which functional correlates suggest that the interruptionof the tracts was real, in low-grade gliomas, a certain degree ofcaution should be taken, as in some cases, ISM documented thepresence of functional fibers inside the tumor mass.

The analysis of the impact of the combined approach on surgicalprocedures and patients outcome showed that when DTI-FTinformation was available, the duration of the mapping and thenumber of intraoperative clinically evident seizures were reduced.The patient fatigue was also significantly decreased, particularlyduring surgical removal of large tumors requesting both speech andmotor function mapping. The clinical relevance of this approach wasalso confirmed by the patients’ functional examinations. Most ofpatients in which subcortical tracts were encountered by subcorticalmapping at the time of surgery, developed immediate neurologicaldeficits. The majority of these deficits were transient, andneurological and neuropsychological examination at 1 month fromsurgery showed that the majority of patients recovered, and werecapable to regain a normal useful function and life.

This study has some limitations. The first one is technical andmainly due to inability of DTI-FT to resolve white matter archi-tecture in areas in which more than one fiber population occupies thesame voxel (Jones, 2003; Pierpaoli et al., 1996;Wiegell et al., 2000),as in presence of tumor infiltration. In these voxels, the distributionof fiber orientations is nearly random so FA values can be sig-nificantly reduced, even if functionality of fibers could bemaintained. These partial volume effects are certainly influencedby the relatively elevated dimensions of sampled voxels as regards tothe microscopic dimension of axons (Basser et al., 2000), and canaffect the reliability of tracking by early stopping the reconstructionalgorithm, so they have to be kept in mind to avoid the resection offunctional fibers.

The second drawback is the placement of anatomical seed ROIs,that in this study is based on anatomical landmarks; however, braintumors often alter the known anatomy of fiber pathways, so whitematter mapping using seed ROI based on known normal anatomicallocations might be misleading. Using fMRI landmarks for seed ROIselection may enhance tracking performance (Schonberg et al.,2006) and allow to selectively isolate functionally distinct fiberbundles included in major tracts (Makris et al., 2005), as it has beenshown for arcuatus fasciculus as part of SLF (Kamada et al., 2007).

The third limitation is brain shift, that in this study is consideredand minimized, but it cannot be eliminated, so it is important toconsiderate intraoperatively the accuracy of tract localization, andto take into account safety margins during resection. One of themost frequently proposed solution of this problem could be theadoption of intraoperative MR imaging to update tractographydata: nevertheless, this implies instrumental drawbacks in term ofcost and prolongation of surgical procedures, whereas combinationof tractography datasets with ISM in neuronavigational system hasshown to reduce the duration of interventions.

Conclusion

Globally considered, our data indicate a good concordancebetween DTI-FT data and those obtained during subcorticalmapping. This large series is mainly composed of patients withlow-grade gliomas: these tumors display an infiltrative modality ofgrowth, along short and long connecting fibers, which can be eitherdisplaced, disrupted or invaded by the tumor, and in whichvisualizing the trajectory of the tracts is important for planningand performing surgery. When used in combination with subcorticalmapping, DTI-FT offers the opportunity to quickly find the fibersassociated with motor or language functions during surgery. Theclinical relevance of this combined approach comes from the factthat it further enhances surgical safety maintaining a high rate offunctional preservation.

Acknowledgments

We would like to thank Susumu Mori, Peter Van Zijl, HangyiJiang and the colleagues of F.M. Kirby Research Center forFunctional Brain Imaging, Kennedy Krieger Institute, Baltimore(MD) for teaching DTI data processing and for allowing the use ofthe software DtiStudio. We also thankMr. Antonio Ladislao andMr.Riccardo Biffi for their technical support with intraoperative andMRpictures.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.neuroimage.2007.08.031.

References

Basser, P.J., Pajevic, S., Pierpaoli, C., Duda, J., Aldroubi, A., 2000. In vivofiber tractography using DT-MRI data. Magn Reson Med 44 (4),625–632.

Beaulieu, C., 2002. The basis of anisotropic water diffusion in the nervoussystem – a technical review. NMR Biomed. 15 (7–8), 435–455.

Bello, L., Gallucci, M., Fava, M., Carrabba, G., Giussani, C., Acerbi, F.,Baratta, P., Songa, V., Conte, V., Branca, V., et al., 2007. Intraoperativesubcortical language tract mapping guides surgical removal of gliomasinvolving speech areas. Neurosurgery 60 (1), 67–80 (discussion 80–2).

Berger, M.S., 1996. Minimalism through intraoperative functional mapping.Clin. Neurosurg. 43, 324–337.

Berman, J.I., Berger, M.S., Mukherjee, P., Henry, R.G., 2004. Diffusion-tensor imaging-guided tracking of fibers of the pyramidal tract combinedwith intraoperative cortical stimulationmapping in patients with gliomas.J. Neurosurg. 101 (1), 66–72.

Black, P.M., 1991. Brain tumors: Part 1. N. Engl. J. Med. 324 (21),1471–1476.

Burgel, U., Amunts, K., Hoemke, L., Mohlberg, H., Gilsbach, J.M., Zilles,K., 2006. White matter fiber tracts of the human brain: three-dimensionalmapping at microscopic resolution, topography and intersubject varia-bility. NeuroImage 29 (4), 1092–1105.

Clark, C.A., Barrick, T.R., Murphy, M.M., Bell, B.A., 2003. White matterfiber tracking in patients with space-occupying lesions of the brain: a newtechnique for neurosurgical planning? NeuroImage 20 (3), 1601–1608.

Coenen, V.A., Krings, T., Mayfrank, L., Polin, R.S., Reinges, M.H., Thron,A., Gilsbach, J.M., 2001. Three-dimensional visualization of thepyramidal tract in a neuronavigation system during brain tumor surgery:first experiences and technical note. Neurosurgery 49 (1), 86–92(discussion 92–3).

381L. Bello et al. / NeuroImage 39 (2008) 369–382

Conturo, T.E., Lori, N.F., Cull, T.S., Akbudak, E., Snyder, A.Z., Shimony,J.S., McKinstry, R.C., Burton, H., Raichle, M.E., 1999. Trackingneuronal fiber pathways in the living human brain. Proc. Natl. Acad. Sci.U. S. A. 96 (18), 10422–10427.

Duffau, H., Capelle, L., Sichez, N., Denvil, D., Lopes, M., Sichez, J.P., Bitar,A., Fohanno, D., 2002. Intraoperative mapping of the subcorticallanguage pathways using direct stimulations. an anatomo-functionalstudy. Brain; J. Neurol. 125 (Pt 1), 199–214.

Goebell, E., Paustenbach, S., Vaeterlein, O., Ding, X.Q., Heese, O., Fiehler,J., Kucinski, T., Hagel, C., Westphal, M., Zeumer, H., 2006. Low-gradeand anaplastic gliomas: differences in architecture evaluated withdiffusion-tensor MR imaging. Radiology 239 (1), 217–222.

Hendler, T., Pianka, P., Sigal, M., Kafri, M., Ben-Bashat, D., Constantini, S.,Graif, M., Fried, I., Assaf, Y., 2003. Delineating gray and white matterinvolvement in brain lesions: three-dimensional alignment of functionalmagnetic resonance and diffusion-tensor imaging. J. Neurosurg. 99 (6),1018–1027.

Holodny, A.I., Schwartz, T.H., Ollenschleger, M., Liu, W.C., Schulder, M.,2001. Tumor involvement of the corticospinal tract: diffusion magneticresonance tractography with intraoperative correlation. J. Neurosurg. 95(6), 1082.

Hunsche, S., Moseley, M.E., Stoeter, P., Hedehus, M., 2001. Diffusion-tensor MR imaging at 1.5 and 3.0 T: initial observations. Radiology 221(2), 550–556.

Inoue, T., Ogasawara, K., Beppu, T., Ogawa, A., Kabasawa, H., 2005.Diffusion tensor imaging for preoperative evaluation of tumor grade ingliomas. Clin. Neurol. Neurosurg. 107 (3), 174–180.

Jaermann, T., Crelier, G., Pruessmann, K.P., Golax, X., Netsch, T., vanMuiswinkel, A.M., Mori, S., van Zijl, P.C., Valavanis, A., Kollias, S.,et al., 2004. SENSE-DTI at 3 T. Magn Reson Med 51 (2), 230–236.

Jbabdi, S., Mandonnet, E., Duffau, H., Capelle, L., Swanson, K.R.,Pelegrini-Issac, M., Guillevin, R., Benali, H., 2004. Simulation ofanisotropic growth of low-grade gliomas using diffusion tensor imaging.Magn Reson Med 54 (3), 616–624.

Jellison, B.J., Field, A.S., Medow, J., Lazar, M., Salamat, M.S., Alexander,A.L., 2004. Diffusion tensor imaging of cerebral white matter: a pictorialreview of physics, fiber tract anatomy, and tumor imaging patterns.AJNR Am. J. Neuroradiol. 25 (3), 356–369.

Jones, D.K., 2003. Determining and visualizing uncertainty in estimates offiber orientation from diffusion tensor MRI. Magn Reson Med 49 (1),7–12.

Kamada, K., Todo, T., Masutani, Y., Aoki, S., Ino, K., Takano, T., Kirino, T.,Kawahara, N., Morita, A., 2005. Combined use of tractography-integrated functional neuronavigation and direct fiber stimulation.J. Neurosurg. 102 (4), 664–672.

Kamada, K., Todo, T., Masutani, Y., Aoki, S., Ino, K., Morita, A., Saito, N.,2007. Visualization of the frontotemporal language fibers by tracto-graphy combined with functional magnetic resonance imaging andmagnetoencephalography. J. Neurosurg. 106 (1), 90–98.

Keles, G.E., Lundin, D.A., Lamborn, K.R., Chang, E.F., Ojemann, G.,Berger, M.S., 2004. Intraoperative subcortical stimulation mappingfor hemispherical perirolandic gliomas located within or adjacent tothe descending motor pathways: evaluation of morbidity andassessment of functional outcome in 294 patients. J. Neurosurg.100 (3), 369–375.

Kinoshita, M., Yamada, K., Hashimoto, N., Kato, A., Izumoto, S., Baba, T.,Maruno, M., Nishimura, T., Yoshimine, T., 2005. Fiber-tracking does notaccurately estimate size of fiber bundle in pathological condition: initialneurosurgical experience using neuronavigation and subcortical whitematter stimulation. NeuroImage 25 (2), 424–429.

Makris, N., Kennedy, D.N., McInerney, S., Sorensen, A.G., Wang, R.,Caviness Jr., V.S., Pandya, D.N., 2005. Segmentation of subcomponentswithin the superior longitudinal fascicle in humans: a quantitative, invivo, DT-MRI study. Cereb. Cortex (New York, N.Y.: 1991) 15 (6),854–869.

Mikuni, N., Okada, T., Enatsu, R., Miki, Y., Hanakawa, T., Urayama, S.,Kikuta, K., Takahashi, J.A., Nozaki, K., Fukuyama, H., et al., 2007.

Clinical impact of integrated functional neuronavigation and subcorticalelectrical stimulation to preserve motor function during resection ofbrain tumors. J. Neurosurg. 106 (4), 593–598.

Mori, S., van Zijl, P.C., 2002. Fiber tracking: principles and strategies – atechnical review. NMR Biomed. 15 (7–8), 468–480.

Mori, S., Crain, B.J., Chacko, V.P., van Zijl, P.C., 1999. Three-dimensionaltracking of axonal projections in the brain by magnetic resonanceimaging. Ann. Neurol. 45 (2), 265–269.

Mori, S., Frederiksen, K., van Zijl, P.C., Stieltjes, B., Kraut, M.A.,Solaiyappan, M., Pomper, M.G., 2002a. Brain white matter anatomy oftumor patients evaluated with diffusion tensor imaging. Ann. Neurol. 51(3), 377–380.

Mori, S., Kaufmann, W.E., Davatzikos, C., Stieltjes, B., Amodei, L.,Fredericksen, K., Pearlson, G.D., Melhem, E.R., Solaiyappan, M.,Raymond, G.V., 2002b. Imaging cortical association tracts in the humanbrain using diffusion-tensor-based axonal tracking. Magn Reson Med 47(2), 215–223.

Nagae-Poetscher, L.M., Jiang, H., Wakana, S., Golay, X., van Zijl, P.C.,Mori, S., 2004. High-resolution diffusion tensor imaging of the brainstem at 3 T. AJNR Am. J. Neuroradiol. 25 (8), 1325–1330.

Nimsky, C., Ganslandt, O., Hastreiter, P., Wang, R., Benner, T., Sorensen,A.G., Fahlbusch, R., 2005a. Preoperative and intraoperative diffusiontensor imaging-based fiber tracking in glioma surgery. Neurosurgery 56(1), 130–137 (discussion 138).

Nimsky, C., Ganslandt, O., Hastreiter, P., Wang, R., Benner, T., Sorensen,A.G., Fahlbusch, R., 2005b. Intraoperative diffusion-tensorMR imaging:shifting of white matter tracts during neurosurgical procedures – initialexperience. Radiology 234 (1), 218–225.

Nimsky, C., Ganslandt, O., Fahlbusch, R., 2006. Implementation of fibertract navigation. Neurosurgery 58 (4 Suppl 2) ONS-292-303 (discussionONS-303-4).

Okada, T., Miki, Y., Fushimi, Y., Hanakawa, T., Kanagaki, M., Yamamoto,A., Urayama, S., Fukuyama, H., Hiraoka, M., Togashi, K., 2006a.Diffusion-tensor fiber tractography: intraindividual comparison of 3.0-Tand 1.5-T MR imaging. Radiology 238 (2), 668–678.

Okada, T., Mikuni, N., Miki, Y., Kikuta, K., Urayama, S., Hanakawa, T.,Fushimi, Y., Yamamoto, A., Kanagaki, M., Fukuyama, H., et al.,2006b. Corticospinal tract localization: integration of diffusion-tensortractography at 3-T MR imaging with intraoperative white matterstimulation mapping – preliminary results. Radiology 240 (3),849–857.

Pajevic, S., Pierpaoli, C., 1999. Color schemes to represent the orientation ofanisotropic tissues from diffusion tensor data: Application to whitematter fiber tract mapping in the human brain. Magn Reson Med 42 (3),526–540.

Parmar, H., Sitoh, Y.Y., Yeo, T.T., 2004. Combined magnetic resonancetractography and functional magnetic resonance imaging in evaluationof brain tumors involving the motor system. J. Comput. Assist. Tomogr.28 (4), 551–556.

Pierpaoli, C., Jezzard, P., Basser, P.J., Barnett, A., Di Chiro, G., 1996.Diffusion tensor MR imaging of the human brain. Radiology 201 (3),637–648.

Roux, F.E., Ibarrola, D., Tremoulet, M., Lazorthes, Y., Henry, P., Sol, J.C.,Berry, I., 2001. Methodological and technical issues for integratingfunctional magnetic resonance imaging data in a neuronavigationalsystem. Neurosurgery 49 (5), 1145–1156 (discussion 1156–7).

Schiffbauer, H., Ferrari, P., Rowley, H.A., Berger, M.S., Roberts, T.P., 2001.Functional activity within brain tumors: a magnetic source imagingstudy. Neurosurgery 49 (6), 1313–1320 (discussion 1320–1).

Schiffbauer, H., Berger, M.S., Ferrari, P., Freudenstein, D., Rowley, H.A.,Roberts, T.P., 2002. Preoperative magnetic source imaging for braintumor surgery: a quantitative comparison with intraoperative sensoryand motor mapping. J. Neurosurg. 97 (6), 1333–1342.

Schonberg, T., Pianka, P., Hendler, T., Pasternak, O., Assaf, Y., 2006.Characterization of displaced white matter by brain tumors usingcombined DTI and fMRI. NeuroImage 30 (4), 1100–1111.

Skirboll, S.S., Ojemann, G.A., Berger, M.S., Lettich, E., Winn, H.R., 1996.

382 L. Bello et al. / NeuroImage 39 (2008) 369–382

Functional cortex and subcortical white matter located within gliomas.Neurosurgery 38 (4), 678–684 (discussion 684–5).

Stadlbauer, A., Ganslandt, O., Buslei, R., Hammen, T., Gruber, S., Moser, E.,Buchfelder, M., Salomonowitz, E., Nimsky, C., 2006. Gliomas:histopathologic evaluation of changes in directionality and magnitudeof water diffusion at diffusion-tensor MR imaging. Radiology 240 (3),803–810.

Stadlbauer, A., Nimsky, C., Buslei, R., Salomonowitz, E., Hammen, T.,Buchfelder, M., Moser, E., Ernst-Stecken, A., Ganslandt, O., 2007.Diffusion tensor imaging and optimized fiber tracking in gliomapatients: histopathologic evaluation of tumor-invaded white matterstructures. NeuroImage 34 (3), 949–956.

Stieltjes, B., Kaufmann, W.E., van Zijl, P.C., Fredericksen, K., Pearlson,G.D., Solaiyappan, M., Mori, S., 2001. Diffusion tensor imaging andaxonal tracking in the human brainstem. NeuroImage 14 (3), 723–735.

Tropine, A., Vucurevic, G., Delani, P., Boor, S., Hopf, N., Bohl, J., Stoeter, P.,2004. Contribution of diffusion tensor imaging to delineation of gliomasand glioblastomas. J. Magn. Reson. Imaging: JMRI 20 (6), 905–912.

Wakana, S., Jiang, H., Nagae-Poetscher, L.M., van Zijl, P.C., Mori, S., 2004.Fiber tract-based atlas of human white matter anatomy. Radiology 230(1), 77–87.

Wiegell, M.R., Larsson, H.B., Wedeen, V.J., 2000. Fiber crossing in humanbrain depicted with diffusion tensor MR imaging. Radiology 217 (3),897–903.

Wieshmann, U.C., Clark, C.A., Symms, M.R., Franconi, F., Barker, G.J.,Shorvon, S.D., 1999. Reduced anisotropy of water diffusion in structuralcerebral abnormalities demonstrated with diffusion tensor imaging.Magn. Reson. Imaging 17 (9), 1269–1274.

Witwer, B.P., Moftakhar, R., Hasan, K.M., Deshmukh, P., Haughton, V.,Field, A., Arfanakis, K., Noyes, J., Moritz, C.H., Meyerand, M.E., et al.,2002. Diffusion-tensor imaging of white matter tracts in patients withcerebral neoplasm. J. Neurosurg. 97 (3), 568–575.

Woods, R.P., Grafton, S.T., Holmes, C.J., Cherry, S.R.,Mazziotta, J.C., 1998.Automated image registration: I. General methods and intrasubject,intramodality validation. J. Comput. Assist. Tomogr. 22 (1), 139–152.

Xue, R., van Zijl, P.C., Crain, B.J., Solaiyappan, M., Mori, S., 1999. In vivothree-dimensional reconstruction of rat brain axonal projections bydiffusion tensor imaging. Magn Reson Med 42 (6), 1123–1127.

Yamada, K., Kizu, O., Mori, S., Ito, H., Nakamura, H., Yuen, S., Kubota, T.,Tanaka, O., Akada, W., Sasajima, H., et al., 2003. Brain fiber trackingwith clinically feasible diffusion-tensor MR imaging: initial experience.Radiology 227 (1), 295–301.