Journal of the Neurological Sciences 282 (2009) 55–60

Contents lists available at ScienceDirect

Journal of the Neurological Sciences

j ourna l homepage: www.e lsev ie r.com/ locate / jns

Voxel-wise assessment of progression of regional brain atrophy in relapsing-remittingmultiple sclerosis

Marco Battaglini a, Antonio Giorgio a,b, Maria L. Stromillo a, Maria L. Bartolozzi c, Leonello Guidi c,Antonio Federico a, Nicola De Stefano a,⁎a Department of Neurological and Behavioral Sciences, University of Siena, Italyb FMRIB Centre, Department of Clinical Neurology, University of Oxford, Oxford, UKc Neurology Unit, Empoli Hospital, Empoli, Italy

⁎ Corresponding author. Neurology & NeurometaboliBehavioral Sciences, University of Siena, Viale Bracci 2, 53233432; fax: +39 0577 233411.

E-mail address: destefano@unisi.it (N. De Stefano).

0022-510X/$ – see front matter © 2009 Elsevier B.V. Adoi:10.1016/j.jns.2009.02.322

a b s t r a c t

a r t i c l e i n f o

Article history:

Objective: To perform voxe Received 12 December 2008Received in revised form 21 January 2009Accepted 1 February 2009Available online 16 March 2009

Keywords:AtrophyBrainMultiple sclerosis

l-wise assessments of regional brain atrophy state and rate in subjects withrelapsing–remitting (RR) multiple sclerosis (MS).Background: Recently, attention has focused on defining the tissue compartments and regions within whichbrain atrophy occurs. These regional measures of brain volume changes may help to better define the natureof the pathology underlying MS. In this context, specific regional measures of grey matter (GM) volumechanges can be obtained by using the voxel-based morphometry (VBM) approach.Methods: Fifty-nine subjects with RR MS underwent conventional MRI at baseline and after a mean follow-upperiod of 3 years. Cross-sectionally, two VBM analyses (SPM-VBM, based on the Statistical ParametricMapping software package, and FSL-VBM, based on the FMRIB Software Library tools) were performed toassess cortical GM volumes in RR MS patients compared to 25 age- and sex-matched normal controls (NC).Longitudinally, FSL-VBM and the regional extension of the SIENA method (SIENAr) were both used to assessregional brain atrophy rate in the RR MS patients and its relationship with increases in T2-weighted whitematter (WM) lesion volume over the follow-up period.Results: Widespread decrease in cortical GM volume was found in the RR MS patients compared to NC. BothSPM-VBM and FSL-VBM showed similar involvement of cortical regions (frontal, temporal, parietal, occipitallobes and insula), with a close correlation between the numbers of significant voxels obtained with the twodifferent procedures (r=0.73, pb0.001). After 3-year follow-up, both FSL-VBM and SIENAr showed a furthersignificant reduction in GM volume in the lateral frontal and parietal cortices, bilaterally. Regional volumechanges also appeared significantly pronounced in correspondence to the increase in T2-weighted WM lesionvolume over the follow-up period.Conclusions: By using different methodologies, we showed similar widespread tissue loss in the cerebralcortex of patients with RR MS. This brain tissue loss further progresses over time, particularly in the fronto-parietal cortex and seems to be partially dependent upon the increase of lesion load.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

As expression of general tissue loss, brain atrophy is usuallyconsidered a global marker of irreversible disease progression [1,2]. Inpatients with multiple sclerosis (MS), brain atrophy occurs constantlyand increases progressively over time, with a moderately closerelation to the progression of clinical disability [1,3,4]. Therefore,global atrophy estimation has become widely accepted as a usefulmarker for charting disease progression and monitoring response totherapy in MS [1,5–7].

c Unit, Dept. Neurological and100 Siena, Italy. Tel.: +39 0577

ll rights reserved.

Due to the increasing interest in brain atrophy in MS, computedmethods for estimation of brain volumes have been developed over thepast few years [1,2,8]. By using these methods, global brain volumechanges can be estimated with great precision. Recently, attention hasfocused on defining the tissue compartments and regions within whichbrain atrophy occurs [9–20]. These regional measures of brain volumechangesmayhelp tobetterdefine thenatureof thepathologyunderlyingMS. If relative volume changes across the brain are not homogenous,then regionalmeasuresmay inprinciple detect greater local change thanmeasures averaged across the whole brain and thus provide moresensitive indices of change. This could improve correlationswith relatedpathological features in magnetic resonance imaging (MRI) data.

In this context, specific regional measures of grey matter (GM)volume changes can be obtained by using the voxel-based morpho-metry (VBM) approach [21]. This is a widely used method for

56 M. Battaglini et al. / Journal of the Neurological Sciences 282 (2009) 55–60

assessing differences in regional volume or tissue “concentration”across subjects in conventional MR images. The procedure is relativelystraightforward and ismost commonly carried out using the StatisticalParametric Mapping (SPM) software package [15,18,19,22,23]. AVBM-style analysis can be also performed by using the FMRIB SoftwareLibrary (FSL) tools, a widely used package for neuroimaging studies[24–26]. Mainly, VBM analysis is used to make voxel-wise compar-isons of the local volume of GM across brain regions betweentwo groups of subjects. In some occasions, VBM has been usedlongitudinally to assess indirectly progression rates of GM atrophy[15,22].

Another increasingly popular approach that can measure rates ofbrain volume change more directly is the SIENA method (StructuralImage Evaluation, using Normalization, of Atrophy, also part of FSL[8]). This is a strictly longitudinal method, which uses two MR time-points to estimate directly the local shifts in brain edges across theentire brain. The SIENA method has been extended (SIENAr) to allowthe voxel-wise statistical analysis of brain atrophy across subjects,which results in a regional analysis of difference in brain volumeoccurring over time between two groups of subjects [27,28].

Fig. 1. Cross-sectional VBM analysis between the 59 RR-MS patients and the 25 age- and seyellow shows significant clusters of voxels (corrected for multiple comparisons across spacecompared to NC. Values of t-statistic within significant clusters range from 2 to 4. Backgrou

In this study, two VBM analyses (based on SPM and FSL) were usedto compare cross-sectionally cortical GM volumes between relapsing–remitting (RR) MS patients and normal controls (NC). Furthermore,FSL-VBM and SIENAr were both used to assess voxel-wise regionalbrain volume changes over time in the group of RR MS patients,followed up for a mean period of 3 years. The relevance of regionalatrophyprogressionwasevaluated, also in relation to the extent ofwhitematter (WM) demyelinating lesions.

2. Materials and methods

2.1. Study population

We studied 59 patients (36 females and 23 males; age range=17–60 years, median=35 years) with RR MS [29]. All patients underwentidentical clinical andMRprocedures at baseline andafter amean follow-up time of 3 years (range 2–4.8 years). At baseline, the median diseaseduration was 1.8 years (range 0.1–17 years). The median ExpandedDisability Status Scale (EDSS) score [30]was 1.5 (range: 0–5) at baselineand 1.5 (range: 1–5) at the end of the follow-up. Forty-six out of 59 RR

x-matched normal controls (NC) obtained with SPM-VBM (A) and FSL-VBM (B). Red-at pb0.05). In both procedures, cortical GM volume is similarly lower in RR-MS patientsnd image is the MNI152. All images are shown in radiological convention.

Fig. 2. Scatter-plot illustrating the correlation between the total number of significantvoxels of reduced GM volume in RR-MS patients compared to normal controls obtainedwith SPM-VBM and FSL-VBM procedures (Spearman's rank coefficient=0.73,pb0.001). Six cortical regions showed significant voxels with FSL-VBM but not withSPM-VBM (posterior division of the left temporal fusiform cortex, left superior parietallobule, anterior division of the left inferior temporal gyrus, right lingual gyrus, rightoccipital pole and right temporal fusiform cortex) whereas four cortical regions showedsignificant voxels with SPM-VBM but not with FSL-VBM (anterior division of theparahippocampal gyrus and precentral gyrus, both bilaterally).

57M. Battaglini et al. / Journal of the Neurological Sciences 282 (2009) 55–60

MS patients started a treatment with beta-interferons or glatirameracetate within the first year of observation. During the follow-up period21 out of 59 MS patients experienced clinical relapses and in theseoccasions they were treated with 1 g methylprednisolone/day i.v. for 3or 5 days. All of the MS patients had been corticosteroid treatment-freefor at least 1 month before the baseline and follow-up scans. For eachpatient, neurological evaluation (which included the rating of disabilityusing the EDSS score[30]) was performed within 24 h of the MRexaminationbyanexperiencedobserver (MLB)whowaskept blinded tothe MRI results. For the cross-sectional part of this study, we alsoincluded25age- andsex-matchednormal controls (NC)whounderwentthe sameMRprotocol of the RR-MS patients. The studywas approved bythe Ethic Committee of the Faculty ofMedicine of the University of Sienaand informed consent was obtained from all participating subjects.

2.2. MR examinations

All participants were examined on a Philips Gyroscan scanner(Philips Medical Systems, Best, The Netherlands) operating at 1.5 T,using the sameMR protocol. For each scan, theMR protocol included atransverse dual-echo, turbo spin-echo sequence (TR/TE1/TE2=2075/30/90 ms, 256×256 matrix, 1 signal average, 250×250 mmfield of view) yielding proton density (PD) and T2-weighted (T2-W)images with 50 contiguous 3 mm-thick slices, acquired parallel to theline connecting the anterior and posterior commissures. Subsequently,transverse T1-weighted (T1-W) images (TR/TE=35 ms/10, 256×256matrix,1 signal average, 250×250mm field of view)were acquired. Thissequence yielded image volumes of 50 slices, 3-mm thick, oriented tomatch exactly the PD/T2 acquisition. Nomajor hardware upgrades werecarried out on the scanner during the time of the study, and monthlyquality assurance sessions confirmed the stability of measurementsthroughout the duration of the study.

2.3. MR data analysis

2.3.1. Lesion volumesClassification of T2-WandT1-W lesionvolume(LV)wasperformed in

each patient by a single observer (MLS) employing a segmentationtechnique based on user-supervised local thresholding (Jim 4.0, XinapseSystem, Leicester, UK). For the T2-W LV classification, lesion borderswere determined primarily on PD-weighted images. Information fromT2-W and T1-W images was also considered. Hypointense WM T1-Wlesionswere defined as those lesionswith signal intensity between thatof the GM and the cerebrospinal fluid (CSF) on T1-W scans [31]. Thevalues of both T2-W and T1-W total brain LV was calculated bymultiplying lesion area by slice thickness.

2.3.2. Cross-sectional VBM-based analysesAll VBM-based analyses in the current study were carried out on

the cortical GM. Deep GM was masked out because of the relativelylow spatial resolution of the T1-W images. We assessed the presenceof topographic differences in cortical GM volume between the groupof RR-MS patients and that of NC by using the VBM approach of bothSPM05 (SPM-VBM) and FSLv4.0 (FSL-VBM). Both procedures requirespatial normalization, segmentation into different tissue classes,smoothing and statistical inference in order to obtain statisticalmaps, where the value of each voxel represents the statisticalsignificance of the difference in cortical GM volume for that voxel.

As a recent study [32] pointed out that T1-WGMvolume is reducedin the presence of high T1-W hypointense LV, before running the twoVBM-based analyses each T1-W hypointense lesion of each subjectwas filled up with the mean intensity value of the normal appearingWM present in the same slice of the lesion.

The SPM analysis was performed by using the classical integratedgenerative modelling approach supplied with the software [33]. This

includes brain extraction, tissue segmentation (with prior probabilitymaps used by default), bias correction and non-linear registration.

In FSL, the analysiswas carried out by using separate tools providedby the package in order to achieve in sequence brain extraction [34,35],tissue segmentation (using prior probability maps) [36], and non-linear registration [37].

Both procedures provided images representing partial volumeestimates of GM, WM and CSF. A study-specific GM template wascreated (from GM images of 25 RR-MS patients and 25 NC) in order toreduce the effect of inter-subject structural variability during theregistration step, as previously suggested [38]. This templatewas usedin both SPM and FSL procedures to non-linearly register the native GMimages. In each procedure, each voxel of these registered GM imagewas then divided by the Jacobian determinant of the warp field(modulation step). Each modulated-normalized cortical GM imagewas smoothed with an isotropic Gaussian kernel with a sigma of4.5 mm (~10 mm FWHM).

Differences in cortical GM volume between RR-MS patients and NCwere assessed separately for SPM-VBM and FSL-VBM using an unpairedt-test within the framework of the general linear model (GLM).Statistical inference was carried out using permutation testing [39](5000 permutations) as implemented in the “randomise” program, alsopart of FSL [24]. Cluster-size thresholding (tN2) at p-values b0.05, fullyadjusted for multiple comparisons across space, was used.

The Spearman rank order correlation (SROC) was used to assessthe relationship between the numbers of significant voxels obtainedby the two VBM-based procedures across the different corticalregions.

2.3.3. Longitudinal VBM-style and SIENAr analysesThe FSL-VBM approach was also used longitudinally to assess

changes in cortical GM volume during the 3-year follow-up in RR-MSpatients. In this case, the FSL-VBM procedure described above wascarried out on each pair of T1-W images in order to obtain 59 pairs ofmodulated-normalized cortical GM images. Then, a new image wasgenerated for each patient by subtracting the normalized-modulatedcortical GM image at the end of the follow-up from the homologousbaseline image. All of the “subtracted” cortical GM images were

Fig. 3. Longitudinal FSL-VBM (A) and SIENAr (B) analyses in the whole group of 59 RR-MS patients. Yellow-orange shows voxels of significant GM volume decrease. Note that theseappear relatively more pronounced in regions with increasing T2-W lesion volume (shown in red) over the follow-up period (mean of 3 years), as also suggested by the significantcorrelation found between atrophy rates and T2-W lesion volume changes (seeMaterials andmethods for details). The color overlay is created on top of theMNI152 image. All imagesare shown in radiological convention. In the two methods, note the similar cortical regions showing significant volume decreases over the follow-up period. See the Results sectionfor details.

58 M. Battaglini et al. / Journal of the Neurological Sciences 282 (2009) 55–60

smoothed with an isotropic Gaussian kernel with a sigma of 4.5 mm(~10 mm FWHM) and fed into the voxel-wise statistics (see below).

Baseline and follow-up T1-W images for each patient were alsoanalyzed with the SIENA method [8,40]. SIENA uses a complexprocedure to find brain/non-brain edge points. Then the perpendi-cular edge displacement (between the two time-points) is estimatedat brain edge points. To allow for a voxel-wise statistical analysisacross multiple subjects, standardized cerebral edge flow images wereobtained [27,28]. The subject-specific edge flow images (whichencode amount and direction of perpendicular local brain edgemotion between the two scans) were spatially dilated, affinetransformed to standard space, masked with a standard space brainedge image, smoothed with an isotropic Gaussian kernel with a sigmaof 4.5 mm (~10 mm FWHM) and remasked. These steps producedstandard space flow images that were submitted to voxel-wisestatistical analysis in order to detect voxels with significant brainedge negative shifts (representing regional volume reductions) overthe follow-up period.

Statistical inference for both longitudinal analyses (FSL-VBM andSIENAr) was performed with the aim to assess: (i) the brain regionswith significant volume reduction between baseline and follow-upscans; and (ii) the brain regions where volume reduction significantlycorrelated with increasing T2-W LV over the follow-up period. Adesign matrix within the GML framework was used, with T2-W LVchanges as regressor and age as covariate of no interest. Cluster-sizethresholding (tN2) corrected for multiple comparisons across space atpb0.05 was carried out using permutation testing [39] (5000permutations) as implemented in the “randomise” program.

In both cross-sectional and longitudinal methods, the anatomicallocation of the significant cortical GM clusters was determined byusing predefined standard space masks (see http://www.fmrib.ox.ac.uk/fsl/) as provided by the Harvard–Oxford cortical structural atlas(containing 48 regions for each hemisphere).

3. Results

3.1. Cross-sectional data

The two procedures (SPM-VBM and FSL-VBM) showed diffusecortical GM decrease in the RR-MS patients compared to NC (Fig. 1Aand B). Significant voxels were located in the frontal, temporal,parietal, occipital and insular cortices of both hemispheres.

This decrease in GM volume was located in similar cortical regionsusing the two different procedures (Fig. 1A and B), although the totalnumber of significant GM voxels was higher with FSL-VBM than withSPM-VBM (n=18,381 and 15,519, respectively). More specifically,when the 96 regions provided by the Harvard–Oxford corticalstructural atlas were used to investigate specific local overlappingbetween SPM-VBM and FSL-VBM outputs, we found that the 89% (86out of 96) of the total cortical regions were shared by both theprocedures. Moreover, there was a close positive correlation betweenthe numbers of significant voxels obtained with the two differentprocedures (r=0.73, pb0.001, Fig. 2).

3.2. Longitudinal data

The FSL-VBM analysis of cortical GM changes in the 59 RR-MSpatients over 3-year follow-up showed significant decrease of corticalGM volume in both hemispheres, mainly involving the frontal lobe(paracingulate gyrus, anterior division of the cingulate gyrus, inferiorand superior frontal gyri, precentral gyrus), the parietal lobe (pre-cuneous and posterior division of the cingulate gyrus) and the occipitallobe (cuneous cortex, intracalcarine and lateral occipital cortices)(Fig. 3A).

When the analysis was performed using the SIENAr procedure,significant regional volume reductions over the follow-up periodwerepronounced around the lateral and third ventricles. However,significant changes were also found in cortical regions of the frontaland parietal lobes, consistent with the results of the FSL-VBM analysis(Fig. 3B).

In the group of patients, the mean T2-W LV was 5.9±6.7 cm3

at baseline and 6.7±6.7 cm3 at follow-up. The increase over time inWMT2-W LV appeared more pronounced alongside the significant volumedecrease, especially when, as with the SIENAr approach, the volumedecrease in the lateral ventricles was considered (Fig. 3A and B).

4. Discussion

In this study, by using different voxel-wise procedures formeasurement of brain volumes, we demonstrated progressiveregional brain volume reductions in a group of RR-MS patients. Atbaseline, independently of the VBM procedure used, significantdecreases in cortical GM volume were found bilaterally in mostcortical regions (i.e., frontal, temporal, parietal, occipital and insular

59M. Battaglini et al. / Journal of the Neurological Sciences 282 (2009) 55–60

cortices) of RR-MS patients compared to normal controls. This volumedecrease progressed further over a 3-year follow-up period, mainly inthe frontal and parietal regions. Moreover, decreases in regionalvolumes, especially around the lateral ventricles, were in close spatialrelationship with thoseWM regions showing increases in T2-W lesionvolume between baseline and end of follow-up.

Over the past years, brain atrophy measurement has become animportant marker for chartingMS progression [41]. Notably, measuresof cortical GM volume have appeared to be more sensitive to changes,providing relevant information on the presence of tissue loss innormal appearing brain tissue [22,42,43]. The most used approach forassessment of GM volume is the VBM analysis [21]. This has beensuccessfully used in several MS studies to assess localization of GMloss cross-sectionally [18,23,44]. These studies, as well as other studiesusing different approaches [16,45,46], have showed the presence of awidespread cortical GM loss in MS patients, with slightly differentpatterns of regional distribution in relation to clinical stage anddisease phenotype [16,23,45]. Our present study confirms previousreports by showing, in a group of patients with established RR MS,widespread cortical GM loss involving the frontal, temporal, parietaland occipital lobes and the insula. Interestingly, the significant tissueloss was similar in the cortical GM of both hemispheres, which wasnot always the case in previous studies on RR MS patients [18,23].Because of the relatively low spatial resolution of the T1-W images(i.e., 3-mm slice thickness) in this study, structures of the deep GMwere masked out and not considered for the VBM-based analyses.However, previous cross-sectional studies have not found significantGM volume reductions in the deep GM of RR-MS patients [18,23,44].

In this study, the cross-sectional regional analysis of cortical GMvolumes was done by using two different VBM-based procedures suchas SPM-VBM [21] and FSL-VBM [24]. Similarly to a previous studycomparing the two procedures in schizophrenic patients [26], wefound a few more significant clusters with FSL-VBM than with SPM-VBM, analyzing the datawith the same statistical model. Thismight bedue to small differences in the segmentation and/or registration stepsused in the two procedures (see Methods for details). However,overall the two methods provided similar information, with as muchas 89% of the total cortical regions shared by both methods and a veryclose positive correlation between the numbers of significant voxelsobtained with the two different procedures. This finding reinforcesthe confidence in our results, suggesting that large and diffusedecreases in GM volume are really occurring in RR-MS patients, evenin a group with short disease duration.

The FSL-VBM procedure was also used here longitudinally, over afollow-up period as long as 3 years, to assess indirectly progressionrates of cortical GM atrophy in RR-MS patients. In this case, an imagewas generated for each patient by subtracting the normalized-modulated cortical GM image at the end of the follow-up from thehomologous baseline image and the changes in cortical GM volumesoccurring over the 3-year follow-up period were assessed. Resultsshowed decreases over time in the cortical GM volume of bothhemispheres, mainly involving frontal, parietal and occipital lobes(Fig. 3A). Similar results were found in a very recent 1-year longitudinalstudy involving 151 RR-MS patients and using SPM-VBM [22]. Incontrast, significant GM volume decreases were found only in thethalami and right lateral prefrontal cortex after 2-year follow-up in arelatively small group of early RR MS who did not show significantvolume decreases at baseline [47].

Furthermore, to increase the confidence in our results, we alsoanalyzed longitudinal data by using a recent extension of the SIENAmethod. This allows the investigation of atrophy rates in different brainregions (SIENAr) without the need of intensity-based segmentationalgorithms [27,28]. Due to the nature of its direct local measurement ofbrain edge shifts, SIENAr does not suffer from the well-known problemof ambiguity of interpretation (i.e., volume changes due to real tissueloss or systematic alignment errors) that can be found in the VBM

approach [48,49]. In the current study, SIENAr analysis revealed thatsubstantial cortical atrophy was also present in the frontal and parietallobes, consistentwith the results of the FSL-VBManalysis (Fig. 3A andB).It is important to note that the extent of abnormality was different inthe VBM-style and SIENAr results. This may be due to differences invariability between the segmentation-based VBM-style approach andthe edge-flow SIENAr method, particularly with the relatively thick3-mm slice thickness used. Interestingly, the SIENAr analysis alsoshowed significant rates of change in periventricular regions (this wasnot assessed in the VBM-style analysis, where subcortical masking wasused). Atrophy rates were in general correlated with increasing WMT2-W lesion volume over the follow-up period, adding to a recent studyin suggesting thepresenceof a closeassociationbetween theprogressionof regional atrophy andWM lesion progression in MS [22].

In summary, by using conventional T1-W images such as thoseused in clinical studies and different voxel-wise methodologies, weexplored regional atrophy state and rate in a group of patients with RRMS.We found a diffuse pathological decrease in cortical GM volume inthe RR-MS patients compared to normal controls. More importantly,the greatest rates of change were located in the lateral fronto-parietalcortex of both hemispheres. Given the role of these cortical regions inimportant processes such as attention [50], motor control [51] andworking memory [52], these results emphasize the relevance of thispattern of injury in furthering our understanding of cognitive dysfunc-tion inMSpatients [53–55]. Proximal axonal transectionswithin corticallesions [56,57], neuronal cytotoxicity [58] and apoptosis [56] all couldcontribute to this relevant pathological process in the cortical GMof MS patients, most likely with a regionally variable susceptibility [59].However, our data also show that WM lesions do play a role indetermining local brain tissue loss in MS, contributing with theother different processes to the development of atrophy and probablybeing responsible for the differences in the distribution and burden ofpathology.

Acknowledgements

The authors gratefully acknowledged the support from Associa-zione Italiana Sclerosi Multipla (NDS).

References

[1] Miller DH, Barkhof F, Frank JA, Parker GJ, Thompson AJ. Measurement of atrophy inmultiple sclerosis: pathological basis, methodological aspects and clinicalrelevance. Brain 2002;125:1676–95.

[2] Fox NC, Schott JM. Imaging cerebral atrophy: normal ageing to Alzheimer's disease.Lancet 2004;363:392–4.

[3] De Stefano N, Iannucci G, Sormani MP, Guidi L, Bartolozzi ML, Comi G, et al.MR correlates of cerebral atrophy in patients with multiple sclerosis. J Neurol2002;249:1072–7.

[4] Sanfilipo MP, Benedict RH, Sharma J, Weinstock-Guttman B, Bakshi R. Therelationship between whole brain volume and disability in multiple sclerosis: acomparison of normalized gray vs. white matter with misclassification correction.NeuroImage 2005;26:1068–77.

[5] Bermel RA, Bakshi R. The measurement and clinical relevance of brain atrophy inmultiple sclerosis. Lancet Neurol 2006;5:158–70.

[6] Zivadinov R. Can imaging techniques measure neuroprotection and remyelinationin multiple sclerosis? Neurology 2007;68:S72–82.

[7] Giorgio A, Battaglini M, Smith SM, De Stefano N. Brain atrophy assessment inmultiple sclerosis: importance and limitations. Neuroimaging Clin N Am2008;18:675–86.

[8] Smith SM, Zhang Y, Jenkinson M, Chen J, Matthews PM, Federico A, et al. Accurate,robust and automated longitudinal and cross-sectional brain change analysis.NeuroImage 2002;17:479–89.

[9] Dalton CM, Chard DT, Davies GR, Miszkiel KA, Altmann DR, Fernando K, et al. Earlydevelopment of multiple sclerosis is associated with progressive grey matteratrophy in patients presenting with clinically isolated syndromes. Brain2004;127:1101–7.

[10] Chard DT, Griffin CM, Parker GJ, Kapoor R, Thompson AJ, Miller DH. Brain atrophyin clinically early relapsing–remitting multiple sclerosis. Brain 2002;125:327–37.

[11] De Stefano N, Matthews PM, Filippi M, Agosta F, De Luca M, Bartolozzi ML, et al.Evidence of early cortical atrophy in MS: relevance to white matter changes anddisability. Neurology 2003;60:1157–62.

60 M. Battaglini et al. / Journal of the Neurological Sciences 282 (2009) 55–60

[12] Quarantelli M, Ciarmiello A, Morra VB, Orefice G, LarobinaM, Lanzillo R, et al. Braintissue volume changes in relapsing–remitting multiple sclerosis: correlation withlesion load. NeuroImage 2003;18:360–6.

[13] Tedeschi G, Lavorgna L, Russo P, Prinster A, Dinacci D, Savettieri G, et al. Brainatrophy and lesion load in a large population of patients with multiple sclerosis.Neurology 2005;65:280–5.

[14] TiberioM,ChardDT, AltmannDR,DaviesG,GriffinCM,RashidW, et al. Grayandwhitematter volume changes in early RRMS: a 2-year longitudinal study. Neurology2005;64:1001–7.

[15] Sepulcre J, Sastre-Garriga J, Cercignani M, Ingle GT, Miller DH, Thompson AJ.Regional gray matter atrophy in early primary progressive multiple sclerosis: avoxel-based morphometry study. Arch Neurol 2006;63:1175–80.

[16] Sailer M, Fischl B, Salat D, Tempelmann C, Schonfeld MA, Busa E, et al. Focalthinning of the cerebral cortex in multiple sclerosis. Brain 2003;126:1734–44.

[17] Chen JT, Narayanan S, Collins DL, Smith SM, Matthews PM, Arnold DL. Relatingneocortical pathology to disability progression in multiple sclerosis using MRI.NeuroImage 2004;23:1168–75.

[18] Prinster A, Quarantelli M, Orefice G, Lanzillo R, Brunetti A, Mollica C, et al. Greymatter loss in relapsing–remitting multiple sclerosis: a voxel-based morphometrystudy. NeuroImage 2006;29(3):859–67.

[19] Pagani E, Rocca MA, Gallo A, Rovaris M, Martinelli V, Comi G, et al. Regional brainatrophy evolves differently in patients with multiple sclerosis according to clinicalphenotype. AJNR Am J Neuroradiol 2005;26:341–6.

[20] Jasperse B, Vrenken H, Sanz-Arigita E, de GV, Smith SM, Polman CH, et al. Regionalbrain atrophy development is related to specific aspects of clinical dysfunction inmultiple sclerosis. NeuroImage 2007;38:529–37.

[21] Ashburner J, Friston KJ. Voxel-based morphometry — the methods. NeuroImage2000;11:805–21.

[22] Bendfeldt K, Kuster P, Traud S, Egger H, Winklhofer S, Mueller-Lenke N, et al.Association of regional gray matter volume loss and progression of white matterlesions in multiple sclerosis — a longitudinal voxel-based morphometry study.NeuroImage 2009;45(1):60–7.

[23] Ceccarelli A, Rocca MA, Pagani E, Colombo B, Martinelli V, Comi G, et al. A voxel-basedmorphometry study of greymatter loss inMS patients with different clinicalphenotypes. NeuroImage 2008;42:315–22.

[24] Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TE, Johansen-BergH, et al. Advances in functional and structural MR image analysis andimplementation as FSL. NeuroImage 2004;23(Suppl 1):S208–19.

[25] Giorgio A,Watkins KE, Douaud G, James AC, James S, De Stefano N, et al. Changes inwhite matter microstructure during adolescence. NeuroImage 2008;39:52–61.

[26] Douaud G, Smith S, Jenkinson M, Behrens T, Johansen-Berg H, Vickers J, et al.Anatomically related grey and white matter abnormalities in adolescent-onsetschizophrenia. Brain 2007;130:2375–86.

[27] De Stefano N, Jenkinson M, Guidi L, Bartolozzi ML, Federico A, Smith SM. Voxel-level cross-subject statistical analysis of brain atrophy in early relapsing remittingMS patients. Proc Int Soc Magn Reson Med 2003;2:2625.

[28] Bartsch AJ, Homola G, Biller A, Smith SM, Weijers HG, Wiesbeck GA, et al.Manifestations of early brain recovery associated with abstinence from alcohol-ism. Brain 2007;130:36–47.

[29] Lublin FD, Reingold SC. Defining the clinical course of multiple sclerosis: results ofan international survey. National Multiple Sclerosis Society (USA) AdvisoryCommittee on Clinical Trials of New Agents in Multiple Sclerosis. Neurology1996;46:907–11.

[30] Kurtzke JF. Rating neurologic impairment in multiple sclerosis: an expandeddisability status scale (EDSS). Neurology 1983;33:1444–52.

[31] van Waesberghe JH, van Walderveen MA, Castelijns JA, Scheltens P, Nijeholt GJ,Polman CH, et al. Patterns of lesion development inmultiple sclerosis: longitudinalobservations with T1-weighted spin-echo and magnetization transfer MR. AJNRAm J Neuroradiol 1998;19:675–83.

[32] Nakamura K, Fisher E. Segmentation of brain magnetic resonance images formeasurement of gray matter atrophy in multiple sclerosis patients. NeuroImage2009;44(3):769–76.

[33] Ashburner J, Friston KJ. Unified segmentation. NeuroImage 2005;26:839–51.[34] Smith SM. Fast robust automated brain extraction. HumBrainMapp 2002;17:143–55.

[35] Battaglini M, Smith SM, Brogi S, De Stefano N. Enhanced brain extraction improvesthe accuracy of brain atrophy estimation. NeuroImage 2008;40:583–9.

[36] Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hiddenMarkov random field model and the expectation-maximization algorithm. IEEETrans Med Imag 2001;20:45–57.

[37] Rueckert D, Sonoda LI, Hayes C, Hill DL, Leach MO, Hawkes DJ. Nonrigidregistration using free-form deformations: application to breast MR images.IEEE Trans Med Imag 1999;18:712–21.

[38] Good CD, Johnsrude IS, Ashburner J, Henson RN, Friston KJ, Frackowiak RS. A voxel-basedmorphometric study of ageing in 465 normal adult humanbrains. NeuroImage2001;14:21–36.

[39] Nichols TE, Holmes AP. Nonparametric permutation tests for functional neuroima-ging: a primer with examples. Hum Brain Mapp 2002;15:1–25.

[40] Smith SM, De Stefano N, Jenkinson M, Matthews PM. Normalized accuratemeasurement of longitudinal brain change. J Comput Assist Tomogr 2001;25:466–75.

[41] De Stefano N, Battaglini M, Smith SM. Measuring brain atrophy in multiplesclerosis. J Neuroimaging 2007;17(Suppl 1):10S–5S.

[42] Valsasina P, Benedetti B, Rovaris M, Sormani MP, Comi G, Filippi M. Evidence forprogressive gray matter loss in patients with relapsing–remitting MS. Neurology2005;65:1126–8.

[43] Fisher E, Lee JC, Nakamura K, Rudick RA. Gray matter atrophy in multiple sclerosis:a longitudinal study. Ann Neurol 2008;64:255–65.

[44] Morgen K, Sammer G, Courtney SM, Wolters T, Melchior H, Blecker CR, et al.Evidence for a direct association between cortical atrophy and cognitiveimpairment in relapsing–remitting MS. NeuroImage 2006;30:891–8.

[45] Charil A, Dagher A, Lerch JP, Zijdenbos AP, Worsley KJ, Evans AC. Focal corticalatrophy in multiple sclerosis: relation to lesion load and disability. NeuroImage2007;34:509–17.

[46] Calabrese M, Atzori M, Bernardi V, Morra A, Romualdi C, Rinaldi L, et al. Corticalatrophy is relevant inmultiple sclerosis at clinical onset. J Neurol 2007;254:1212–20.

[47] Audoin B,DaviesGR, Finisku L, ChardDT, ThompsonAJ,MillerDH. Localization of greymatter atrophy in early RRMS: a longitudinal study. J Neurol 2006;253:1495–501.

[48] Bookstein FL. “Voxel-based morphometry” should not be used with imperfectlyregistered images. NeuroImage 2001;14:1454–62.

[49] Ashburner J, Friston KJ. Why voxel-based morphometry should be used. Neuro-Image 2001;14:1238–43.

[50] Woldorff MG, Hazlett CJ, Fichtenholtz HM, Weissman DH, Dale AM, Song AW.Functional parcellation of attentional control regions of the brain. J Cogn Neurosci2004;16:149–65.

[51] Sahyoun C, Floyer-Lea A, Johansen-Berg H, Matthews PM. Towards an under-standing of gait control: brain activation during the anticipation, preparation andexecution of foot movements. NeuroImage 2004;21:568–75.

[52] Derrfuss J, Brass M, von Cramon DY. Cognitive control in the posterior frontolateralcortex: evidence from common activations in task coordination, interferencecontrol, and working memory. NeuroImage 2004;23:604–12.

[53] Amato MP, Bartolozzi ML, Zipoli V, Portaccio E, Mortilla M, Guidi L, et al.Neocortical volume decrease in relapsing–remitting MS patients with mildcognitive impairment. Neurology 2004;63(1):89–93.

[54] BlinkenbergM, Rune K, Jensen CV, RavnborgM, Kyllingsbaek S, Holm S, et al. Corticalcerebralmetabolism correlateswithMRI lesion load and cognitive dysfunction inMS.Neurology 2000;54:558–64.

[55] Christodoulou C, Krupp LB, Liang Z, Huang W, Melville P, Roque C, et al. Cognitiveperformance andMRmarkers of cerebral injury in cognitively impairedMS patients.Neurology 2003;60:1793–8.

[56] Peterson JW, Bo L,Mork S, ChangA, TrappBD. Transectedneurites, apoptotic neurons,and reduced inflammation in cortical multiple sclerosis lesions. Ann Neurol2001;50:389–400.

[57] Geurts JJ, Bo L, Pouwels PJ, Castelijns JA, Polman CH, Barkhof F. Cortical lesions inmultiple sclerosis: combined postmortem MR imaging and histopathology. AJNRAm J Neuroradiol 2005;26:572–7.

[58] Giuliani F, Goodyer CG, Antel JP, Yong VW. Vulnerability of human neurons to Tcell-mediated cytotoxicity. J Immunol 2003;171:368–79.

[59] Bo L, Vedeler CA, NylandHI, Trapp BD,Mork SJ. Subpial demyelination in the cerebralcortex of multiple sclerosis patients. J Neuropathol Exp Neurol 2003;62:723–32.