NeuroImage 59 (2012) 2661–2669

Contents lists available at SciVerse ScienceDirect

NeuroImage

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

Direct evidence of intra- and interhemispheric corticomotor network degeneration inamyotrophic lateral sclerosis: An automated MRI structural connectivity study

Stephen Rose a,b,c,⁎, Kerstin Pannek a,b, Christopher Bell a, Fusun Baumann a,d, Nicole Hutchinson d,Alan Coulthard e, Pamela McCombe a,d, Robert Henderson d

a The University of Queensland Centre for Clinical Research, Brisbane, Australiab The University of Queensland Centre for Advanced Imaging, Brisbane, Australiac The University of Queensland Centre for Medical Diagnostic Technologies in Queensland, The University of Queensland, St Lucia, Brisbane, Australiad Department of Neurology, Royal Brisbane and Women's Hospital, Brisbane, Australiae Department of Medical Imaging, Royal Brisbane and Women's Hospital, Brisbane, Australia

Abbreviations: ALS, Amyotrophic Lateral Sclerosis; Wgyrus; rh.preCG, right precentral gyrus; lh.PCG, left postgyrus; lh.caudMF, left caudual middle frontal gyrus; rh.⁎ Corresponding author at: UQ Centre for Clinical Res

E-mail address: Stephen.Rose@cai.uq.edu.au (S. Rose

1053-8119/$ – see front matter © 2011 Elsevier Inc. Alldoi:10.1016/j.neuroimage.2011.08.054

a b s t r a c t

a r t i c l e i n f o

Article history:Received 3 June 2011Revised 17 August 2011Accepted 19 August 2011Available online 26 August 2011

Keywords:ALSMNDHARDIDiffusion tractographyStructural connectivity reproducibilityVariability

Although the pathogenesis of amyotrophic lateral sclerosis (ALS) is uncertain, there is mounting neuroimag-ing evidence to suggest a mechanism involving the degeneration of multiple white matter (WM) motor andextramotor neural networks. This insight has been achieved, in part, by using MRI Diffusion Tensor Imaging(DTI) and the voxelwise analysis of anisotropy indices, along with DTI tractography to determine which spe-cific motor pathways are involved with ALS pathology. Automated MRI structural connectivity analyses,which probe WM connections linking various functionally discrete cortical regions, have the potential to pro-vide novel information about degenerative processes within multiple white matter (WM) pathways. Our hy-pothesis is that measures of altered intra- and interhemispheric structural connectivity of the primary motorand somatosensory cortex will provide an improved assessment of corticomotor involvement in ALS. To testthis hypothesis, we acquired High Angular Resolution Diffusion Imaging (HARDI) scans along with high res-olution structural images (sMRI) on 15 patients with clinical evidence of upper and lower motor neuron in-volvement, and 20 matched control participants. Whole brain probabilistic tractography was applied todefine specific WM pathways connecting discrete corticomotor targets generated from anatomical parcella-tion of sMRI of the brain. The integrity of these connections was interrogated by comparing the mean frac-tional anisotropy (FA) derived for each WM pathway. To assist in the interpretation of results, wemeasured the reproducibility of the FA summary measures over time (6 months) in control participants.We also incorporated into our analysis pipeline the evaluation and replacement of outlier voxels due tohead motion and physiological noise. When assessing corticomotor connectivity, we found a significant re-duction in mean FA within a number of intra- and interhemispheric motor pathways in ALS patients. The ab-normal intrahemispheric pathways include the corticospinal tracts involving the left and right precentral gyri(lh.preCG, rh.preCG) and brainstem (bs); right postcentral gyrus (rh.postCG) and bs; lh.preCG and left poste-rior cingulate gyrus (lh.PCG); rh.preCG and right posterior cingulate gyrus (rh.PCG); and the rh.preCGand right paracentral gyrus (rh.paraCG). The abnormal interhemispheric pathways included the lh.preCGand rh.preCG; lh.preCG and rh.paraCG; lh.preCG and right superior frontal gyrus (rh.supFG); lh.preCGand rh.postCG; rh.preCG and left paracentral gyrus (lh.paraCG); rh.preCG and left superior frontal gyrus(lh.supFG); and the rh.preCG and left caudal middle frontal gyrus (lh.caudMF). The reproducibility of themeasurement of these pathways was high (variation less than 5%). Maps of the outlier rejection voxels,revealed clusters within the corpus callosum and corticospinal projections. This finding highlights the impor-tance of correcting for motion artefacts and physiological noise when studying clinical populations. Our novelfindings, many of which are consistent with known pathology, show extensive involvement and degenera-tion of multiple corticomotor pathways in patients with upper and lower motor neuron signs and providesupport for the use of automated structural connectivity techniques for studying neurodegenerative diseaseprocesses.

M, white matter; FA, fractional anisotropy; FOD, fibreerior cingulate gyrus; rh.PCG, right posterior cingulate gypostCG, right postcentral gyrus; rh.supFG, right superiorearch, Royal Brisbane and Women's Hospital, Brisbane 4).

rights reserved.

© 2011 Elsevier Inc. All rights reserved.

orientation distribution; bs, brain stem; lh.preCG, left precentralrus; lh.paraCG, left paracentral gyrus; rh.paraCG, right paracentralfrontal gyrus; lh.supFG, left superior frontal gyrus.029, Australia. Fax: +61 7 33465599.

2662 S. Rose et al. / NeuroImage 59 (2012) 2661–2669

Introduction

Amyotrophic Lateral Sclerosis (ALS) is a progressive disease that istraditionally associated with loss of upper and lower motor neurons(UMN and LMN, respectively). Patients with ALS are defined accord-ing to strict criteria that require abnormalities of UMN and LMN(Brooks et al., 2000), although patients who fulfil these criteria mayshow heterogeneity of clinical features (Kiernan et al., 2011). ALS isuniformly fatal with paralysis and death normally occurring within3–5 years after onset of symptoms.

Although ALS was first described over 140 years ago, progress re-garding our understanding of the aetiology is still limited with manyproposed models of pathogenesis (Ilieva et al., 2009; Rothstein,2009). However, it is recognised that ALS pathology in the brain iswidespread and involves motor networks other than those compris-ing the primary corticospinal tracts (CST) (Eisen and Weber, 2001),as highlighted in recent neuroimaging reviews (Agosta et al.,2010a; Turner and Modo, 2010). Such information supports the con-cept of motor neuron degeneration being a progressive process thatspreads contiguously through multiple corticomotor networks,which potentially explains the wide variation in clinical features forpatients who present with mixed UMN and LMN signs (Ravits andLa Spada, 2009). Central to these findings has been the use of diffu-sion tensor imaging (DTI), which enables the assessment of the integ-rity of white matter (WM) pathways by measuring the preferreddirection of water diffusion along WM fibre tracts (Basser et al.,1994; Beaulieu 2002).

With regard to analysis strategies, numerous studies have takenadvantage of the ability to study WM tract degeneration using a vox-elwise analysis of DTI-derived fractional anisotropy (FA) measures(Abe et al., 2004; Agosta et al., 2007; Ciccarelli et al., 2009; Filippiniet al., 2010; Keller et al., 2011; Metwalli et al., 2010; Sage et al.,2007, 2009; Senda et al., 2011; Stanton et al., 2009). QuantitativeFA measures are believed to reflect changes in myelination, fibredensity and packing (Concha et al., 2010; Mädler et al., 2008). Alter-native approaches have made use of diffusion tensor tractography(DTT) which enables the study of specific WM pathways associatedwith ALS neuropathology (Aoki et al., 2005; Agosta et al., 2010b;Blain et al., 2011; Hong et al., 2008; Sage et al., 2009; Sato et al.,2010; Senda et al., 2009; van der Graaff et al., 2011). A useful exten-sion to these studies is the use of DTT, employing predefined corti-cal targets derived from structural MRI (sMRI) to investigatestructural connectivity of corticomotor networks in ALS (Ciccarelliet al., 2006; Verstraete et al., 2010). Although these previousstudies were limited to connectivity analyses of the primary motorcortex in ALS, the concept of using multiple cortical target regionshas significant appeal as connections between various cortical andsubcortical regions can be investigated in a completely automatedfashion (Hagmann et al., 2008; Johansen-Berg and Rushworth,2009).

An important issue that is rarely addressed when applying DTI inclinical populations is the effect of image artefacts, induced by headmotion and physiological noise, on summary measures of WM integ-rity, such as FA. Well known sources of image artefact, such as thoseinduced by eddy currents and susceptibility effects, can be reducedwith appropriate acquisition and post processing techniques (Jones,2010). However a recent study has highlighted the deleterious im-pact of subtle head movement and cardiac pulsatile motion on an-isotropy indices (Walker et al., 2011). Voxels containing theseartefacts can be identified (RESTORE, Chang et al., 2005), andreplaced (Morris et al., 2011) within the processing pipeline. Onewould expect this problem to be common when scanning patientswith neurodegenerative disease using DTI due to lack of headcontrol.

The objective of this study was to investigate the integrity of thecorticomotor pathways associated with the primary motor and

somatosensory cortex in patients with sporadic ALS presenting withmixed UMN and LMN signs using a fully automated structural con-nectivity approach. Our specific hypothesis is that an analysisof intra- and interhemispheric structural connectivity, based on ameasure of mean FA derived from all streamlines defining theWM connections associated with the precentral and postcentralgyri, will provide new insight into corticomotoneuron involvementin ALS. To achieve this goal, we have employed a strategy combiningHigh Angular Resolution Diffusion Imaging (HARDI) with constrainedspherical deconvolution which describes diffusion in complexWM networks using a fibre orientation distribution function (FOD)(Tournier et al., 2007). Whole brain probabilistic tractography wasapplied to define specific WM pathways connecting discrete cortico-motor targets generated from anatomical parcellation of high-resolu-tion sMRI of the brain (Pannek et al., 2010). A schematic diagramoutlining our automated pipeline is provided in Fig. 1. Importantly,within this pipeline we have endeavoured to integrate strategies toreduce artefacts generated from head motion and physiologicalnoise. To enable comparison with previous studies, we have also in-vestigated a voxelwise analysis of FA maps using Tract Based SpatialStatistics (TBSS, Smith et al., 2006). Furthermore, to assist in the inter-pretation of results, the reproducibility of the FA encoded corticomo-tor connections involving the precentral and postcentral gyri was alsoevaluated in control participants.

Materials and methods

Participants

Fifteen patients with probable or definite ALS as defined by the re-vised EL Escorial criteria (Brooks et al., 2000) were recruited into thestudy, see Table 1. These patients were all typical ALS with progres-sive muscular atrophy and primary lateral sclerosis variants excluded.All patients were referred from the multidisciplinary ALS clinical ser-vice at the Royal Brisbane and Women's Hospital and were clinicallyclassified for the stage of disease using the Amyotrophic Lateral Scle-rosis Functional Rating Scale-Revised (ALSFRS-R). Patients were alsoclassified according to the site of disease onset as either limb-onsetor bulbar-onset and the duration of the disease, in months, was deter-mined from the onset of symptoms. From the ALSFRS and disease du-ration, a disease progression score was obtained as used by others(Ciccarelli et al., 2006). Most patients were receiving riluzole treat-ment. Twenty matched healthy control participants were alsorecruited. The control participants had no history of hypertension orcerebrovascular disease and were not on any medications. To gaugereproducibility of connectivity measures, 7 control subjects under-went scanning on two occasions separated by a period of 6 months.All subjects gave their informed written consent to participatein the study as approved by the local Human Research EthicsCommittee.

MRI protocol

MRI data was acquired using a 3 T Siemens TimTrio (Siemens,Erlangen, Germany) using commercial sequences from VB17 Neuroapplications and Diffusion Tensor Imaging options and a 12channel head coil. Along with a number of radiological scans, ahigh-resolution structural image was acquired for each participantusing a 1 mm3 isotropic 3D T1 MPRAGE (FOV 24×25.6×17.6 cm,TR/TE/TI 2300/2.26/900 ms, flip angle 9). The imaging time was9:14 min. The diffusion imaging parameters were: 60 axial slices,FOV 30×30 cm, TR/TE 9200/112 ms, 2.5 mm slice thickness, acquisi-tion matrix 128×128 with a 2.3 mm in plane image resolution,an acceleration factor of 2 employing the GRAPPA parallel acquisi-tion technique (Griswold et al., 2002), and a diffusion encoding gra-dient strength of b=3000 s mm−2. Sixty diffusion–weighted

Fig. 1. Schematic diagram of the automated image-processing pipeline. The structural MRI (sMRI) data is functionally parcellated into 66 cortical regions using Freesurfer. Withregard to the raw HARDI data, the fibre orientation distribution (FOD) is calculated using constrained spherical deconvolution and probabilistic tractography performed usingMRtrix to generate the 3D whole brain tractograms. Connectivity indices, based on measures of mean FA derived from each anatomical connection, were generated for statisticalanalysis. Connections of the left precentral gyrus derived using this approach are given in the figure.

2663S. Rose et al. / NeuroImage 59 (2012) 2661–2669

images were acquired at each location consisting of 1 low (b=0)and 64 high diffusion weighted images. The acquisition time for thediffusion dataset was 9:40 min. A field map for diffusion data was ac-quired using two 2D gradient recalled echo images (TE1/TE24.76/7.22 ms) to assist the correction for distortion due to susceptibil-ity inhomogeneity.

MRI analysis

Structural images

Cortical parcellation was performed on structural images with theFreesurfer image analysis suite (http://surfer.nmr.mgh.harvard.edu).

Table 1Participant characteristics.

Controls ALS

Age, y, mean±SD (range) 53±12 (41–73) 57±14 (41–75)Male:female 16:4 12:3Handedness for writing, R:L 20:0 15:0Disease duration, mo,mean±SD (range)

NA 24±11.6 (9–40)

ALSFRS-R, mean±SD (range) NA 39±5.6 (29–46)Disease progression,a

mean±SD (range)NA 0.47±0.37 (0.09–1.36)

Site of onsetb Bulbar/RUL/LUL/LLL/RLL NA 3/4/3/4/1

a Disease progression=(48 — ALSFRS-R score)/disease duration.b Site of onset: RUL — right upper limb; LUL — left upper limb; LLL — left lower limb;

and RLL — right lower limb.

2664 S. Rose et al. / NeuroImage 59 (2012) 2661–2669

Intensity inhomogeneity of the images was corrected (Sled et al.,1998) and non-brain tissue was removed using a hybrid watershed/surface deformation procedure (Fischl et al., 2004). The cerebral cor-tex was parcellated into 33 units per hemisphere based on gyral andsulcal structure (Desikan et al., 2006; Fischl et al., 2004) along with amask delineating brainstem structures, which comprised the cerebralpeduncles. It has been shown previously that this automated corticalparcellation procedure is comparable with manual delineation of re-gions in terms of accuracy (Desikan et al., 2006). A terminationmask for tractography was generated from the structural image toprevent streamlines from crossing the cortical folds (Pannek et al.,2010).

Diffusion processing

Diffusion weighted images were corrected for subject motionusing the method described by Bai and Alexander (2008). Briefly,the diffusion tensor was calculated from the (uncorrected) diffusiondata. Using the diffusion tensor information, a synthetic image wasgenerated for every volume of the diffusion series. Every volume ofthe raw diffusion data was then aligned with the corresponding syn-thetic volume using a six degrees-of-freedom registration performedwith ANTS (http://picsl.upenn.edu/ANTS), with the appropriate ad-justment of the b-matrix (Leemans and Jones, 2009; Rohde et al.,2004). Susceptibility distortions were corrected using the fieldmapemploying FUGUE (Jenkinson 2003) and PRELUDE (Jenkinson 2004)in raw image space; both part of FMRIB Software Library FSL, with sig-nal intensity correction (Jones and Cercignani 2010). Motion artefactswere identified and replaced using the DROP-R method (Morriset al., 2011) in conjunction with the registration method by Bai andAlexander (2008) described above. For visual assessment, the voxelsidentified as outliers due to motion and physiological noise weremapped onto an average FAmodel constructed from all control partic-ipants data sets. The fibre orientation distribution (FOD) was estimat-ed using the constrained spherical deconvolution method within theMRtrix package (http://www.brain.org.au/software) (Tournier et al.,2007). Fibre tracking was performed using MRtrix. Fifty probabilisticstreamlines were seeded for each voxel over the entire brain volumeto generate a whole brain tractography map. A schematic diagramoutlining the automated image-processing pipeline is provided inFig. 1.

Connectivity metrics

Connectivity metrics were generated using our previously pub-lished technique (Pannek et al., 2010). In brief, b=0 diffusion imageswere linearly registered (6 degrees of freedom) to the structuralimage and the inverse transformation applied to the parcellated cor-tical masks maintaining the higher resolution. A terminationmask was applied to prevent streamlines from crossing corticalfolds. Corticocortical and brainstem connections for the precentral

and postcentral gyri were then generated by hit-testing every stream-line's terminal ends with every cortical parcellation (Pannek et al.,2010). To provide a summary measure, a mean FA value was derivedfrom all streamlines defining a specific WM connection for all of themajor motor pathways associated with the precentral and postcentralgyri.

Statistical analyses

Reproducibility of connectivity measures in control participants

The reproducibility over time of mean FA for each corticocorticalor corticospinal connection for each of the precentral and postcentralgyri was assessed by evaluating the average absolute relative differ-ence of this measure, as expressed by the following equation.

Δ ¼ 2x A1–A2j j= A1 þ A2ð Þ

where A is the mean FA for each connection from scans 1 and 2, re-spectively. We have used a similar approach to measure the repro-ducibility of quantitative average pathlength maps generated usingwhole brain tractography (Pannek et al., 2011).

Group comparisons

To detect significant group differences in precentral and postcen-tral connectivity between the ALS and control participants, a non-parametric Mann–Whitney U-Test (employing an FDR of 10%) wasapplied to the FA summary measures.

Correlation of connectivity indices with clinical measures

Pearson correlations between summary connectivity indices andALSFRS-R scores and disease progression measures, within the ALSpatient group, were also investigated. Correlations were correctedfor multiple comparisons using FDR.

Voxelwise FA tract differences

To enable comparison of anisotropy changes associated in our pa-tient group with those from other studies (Ciccarelli et al., 2009;Filippini et al., 2010; Sage et al., 2009), changes in FA measures with-in WM tracts, which have previously been attributed to ALS neuro-pathology, were also measured using TBSS (Smith et al., 2006). Inbrief, this voxelwise approach relies on the registration of each sub-jects FA map to the FMRIB58 template. An FA skeleton was then gen-erated which represents WM tracts that are common to all subjects.Differences in FA between the ALS and control participants were de-termined using a permutation-based modelling approach (Nicholsand Holme, 2002). The resultant statistical maps were thresholdedat pb0.05, corrected for multiple comparisons using threshold-freecluster enhancement (Smith and Nichols, 2009). Permutation model-ling was also employed to investigate the correlation between FAmeasures and clinical scores (ALSFRS-R, disease progression), withinthe patient group.

Results

Demographic and clinical summary measures for the 15 ALS pa-tients are presented in Table 1. All patients had clinical signs ofupper and lower motor neuron involvement. 3 patients had bulbaronset, 7 had upper limb onset and 5 had lower limb onset. Therewas no significant difference in mean age between the ALS and con-trol participants (pb0.14) and all participants were right handed.The gender ratio was similar for both groups. Results obtained forthe voxelwise analysis of FA measures using TBSS are given in Fig. 2.

Fig. 2. TBSS analysis of FA data. Results of the voxelwise TBSS analysis of FA maps generated for the ALS and control participants. The blue regions show where there was a signif-icant reduction in FA value in the ALS patients, corrected for multiple comparisons. Significant involvement of the corpus callosum was observed, which extended rostrally, andbilaterally to the motor cortices. With respect to CST involvement, there was a significant reduction in FA in rostral regions of the CST bilaterally, and in more caudal regions withinthe right hemisphere in ALS patients compared to controls.

2665S. Rose et al. / NeuroImage 59 (2012) 2661–2669

There was a significant reduction in FA values, pb0.05 corrected formultiple comparisons, within the corpus callosum extending rostrallyand bilaterally to the motor cortices in the ALS patients compared tothe control participants. With respect to corticospinal tracts, in theALS group there were significant decreases in FA within voxels inthe superior-rostral regions of the CST bilaterally, and in caudal re-gions only within the right hemisphere.

The statistical analysis of the connectivity derived FA summarymeasures, corrected for multiple comparisons, revealed a number ofcorticomotor connections with a significant reduction in mean FAvalue in ALS patients compared to control participants. As documen-ted in Table 2, the intrahemispheric connections include corticospinaltracts involving the left and right precentral gyri (lh.preCG, rh.preCG)and brainstem (bs); right postcentral gyrus (rh.postCG) and bs;lh.preCG and left posterior cingulate gyrus (lh.PCG); rh.preCG andright posterior cingulate gyrus (rh.PCG); and the rh.preCG and rightparacentral gyrus (rh.paraCG). A significant decrease in FA summarymeasure was also found in a number of interhemispheric connections,namely the lh.preCG and rh.preCG; lh.preCG and rh.paraCG; lh.preCGand right superior frontal gyrus (rh.supFG); lh.preCG and rh.postCG;

Table 2Significant differences in FA summary measures (mean and standard deviation (SD), reprodated with the precentral and postcentral gyri in ALS participants compared to controls.

WM connection FA (controls)

Mean SD

Intrahemisperic lh.preCG–bs 0.448 0.023lh.preCG–lh.PCG 0.412 0.037rh.preCG–bs 0.450 0.026rh.preCG–rh.paraCG 0.363 0.028rh.preCG–rh.PCG 0.406 0.029rh.postCG–bs 0.444 0.022

Interhemisperic lh.preCG–rh.preCG 0.419 0.020lh.preCG–rh.paraCG 0.455 0.026lh.preCG–rh.supFG 0.444 0.017lh.preCG–rh.postCG 0.417 0.022rh.preCG–lh.paraCG 0.451 0.026rh.preCG–lh.supFG 0.424 0.190rh.preCG–lh.caudMF 0.392 0.021

bs: brain stem; lh.preCG: left precentral gyrus; rh.preCG: right precentral gyrus; lh.PCG:paracentral gyrus; rh.paraCG: right paracentral gyrus; lh.caudMF: left caudual middle fronlh.supFG: left superior frontal gyrus.

rh.preCG and left paracentral gyrus (lh.paraCG); rh.preCG and leftsuperior frontal gyrus (lh.supFG); and the rh.preCG and left caudalmiddle frontal gyrus (lh.caudMF). There were no connections wherethere was an increase in FAmeasure in ALS compared to control partic-ipants. Representative diagrams showing these corticomotor connec-tions along with box–whisker plots of summary FA measures areprovided in Fig. 3.

With respect to the precision of structural connectivity measuresfor these corticomotor regions, we found that the mean FA summarymeasure derived from all streamlines defining each WM connectionwere highly reproducible (b5% variation) over time (6 months) with-in the control participants, see Table 2. With respect to possible corre-lations between FA summary measures and clinical scores (ALSFRS-Rand disease progression score) for both the TBSS and connectivity an-alyses, no correlation reached a level of statistical significance aftercorrecting for multiple comparisons.

With regard to the impact of motion and physiological noise onthe diffusion data, many of the outlier voxels were found to be locatedin clusters within the corpus callosum, and around corticospinal tractprojections, see Fig. 4.

ucibility) for the intra- (top) and interhemispheric (bottom) WM connections associ-

FA (MND) p-Value Reproducibility

Mean SD(%)

0.424 0.019 0.0108 2.160.386 0.031 0.0108 0.620.418 0.017 0.0005 1.610.337 0.027 0.0133 2.380.375 0.033 0.0036 0.970.424 0.017 0.0016 0.770.396 0.018 0.0051 1.920.4112 0.029 0.0002 1.630.422 0.025 0.0079 2.20.393 0.023 0.0079 1.200.416 0.029 0.0025 1.320.386 0.033 0.0022 1.780.368 0.026 0.0210 1.09

left posterior cingulate gyrus; rh.PCG: right posterior cingulate gyrus; lh.paraCG: lefttal gyrus; rh.postCG: right postcentral gyrus; rh.supFG: right superior frontal gyrus;

Fig. 3. Box–whisker plots showing changes in mean FA measures for the precentral and postcentral gyri. There was a significant reduction in mean FA between ALS patients andcontrol participants, corrected for multiple comparisons, for intrahemispheric connections including the left and right precentral gyri (lh.preCG, rh.preCG) and brainstem (bs);right postcentral gyrus (rh.postCG) and bs; lh.preCG and left posterior cingulate gyrus (lh.PCG); rh.preCG and right posterior cingulate gyrus (rh.PCG); and the rh.preCGand right paracentral gyrus (rh.paraCG). A significant decrease in FA summary measure was also found in a number of interhemispheric connections, namely the lh.preCG andrh.preCG; lh.preCG and rh.paraCG; lh.preCG and right superior frontal gyrus (rh.supFG); lh.preCG and rh.postCG; rh.preCG and left paracentral gyrus (lh.paraCG); rh.preCG andleft superior frontal gyrus (lh.supFG); and the rh.preCG and left caudal middle frontal gyrus (lh.caudMF). Representative images showing the anatomical location of these connec-tions are also given within the figure. For each WM pathway, the box and whisker plot presented on the left represents data from control participants (higher mean FA value),whilst the plots given on the right represent mean FA measures derived from ALS patients.

2666 S. Rose et al. / NeuroImage 59 (2012) 2661–2669

Discussion

The results of our voxelwise analysis of FA maps corroborates pre-vious findings (Agosta et al., 2010a; Turner and Modo, 2010), show-ing motor neuron degeneration in multiple corticomotor networksin ALS, which potentially explains the wide variation in clinical fea-tures for patients who present with mixed UMN and LMN signs(Ravits and La Spada, 2009). However, a limitation of this approachis the difficulty of identifying which motor networks are primarily in-volved with ALS pathology. To overcome this constraint, we investi-gated the change in FA within a number of key intra- andinterhemispheric corticomotor pathways using an automated struc-tural connectivity strategy. Importantly, we have endeavoured to ad-dress many of the pitfalls reported to be associated with diffusionimaging and connectivity analyses (Jones 2010; Jones and Cercignani,2010), namely (i) the correction of inherent diffusion image distor-tions, head motion and physiological noise artefacts which adverselyimpact on tractography performance and accuracy of diffusivity mea-sures, (ii) employing HARDI and a higher order diffusion model alongwith probabilistic tractography to help resolve complex crossing fibrenetworks, (iii) using quantitative measures of FA to compare tract in-tegrity, rather than streamline number which has yet to be validatedas a possible connectivity index for groupwise analyses, and (iv)showing that our connectivity derived summary measures are

reproducible over time in matched control participants. Althoughmany previous studies have used lower b values to investigate WMchanges associated with MND pathology, in a recent study (Panneket al., 2010) employing HARDI using two b values (1000 and3000 s mm−2) with a higher order diffusion model (Tournier et al.,2007) along with probabilistic tractography, we found that the repro-ducibility of connectivity measures of the major WM commissuralpathways was improved using the higher b value. We believe thatthe use of higher b values is important for connectivity analyses to as-sist in the resolution of complex WM fibre networks.

Employing this analysis strategy revealed a number of novel find-ings. First, we found abnormalities of multiple corticospinal tract pro-jections from both the precentral and postcentral gyri in ALS subjects.Although degeneration of the CSTs was evident in the TBSS analysis,and a common finding in many studies, our connectivity analysisrevealed the loss in integrity of CSTs projecting from the precentralgyri bilaterally, and onlywithin the right CST projecting from the post-central gyrus. One previous tractography study has reported reducedFA measures within the CSTs projecting from only the right postcen-tral gyrus in ALS (Sage et al., 2007). Thus connectivity analyses maybe useful complementary techniques to voxelwise approaches foridentifying specific WM pathways associated with disease processes.

With regard to other intrahemispheric connectivity changes, anumber of studies have reported reduced FA in frontal lobe motor

Fig. 4. Effect of head motion and physiological noise on the fit of the diffusion tensor. Outlier voxels, due to head motion and physiological noise are overlaid on an average FA modelgenerated from the control participants. The outlier voxel map, was generated using the following formula ((#outliers/#degrees of freedom) ×100), as recently reported (Walkeret al., 2011). The outlier voxels are clustered within cerebellum, the mid and posterior regions of the corpus callosum where interhemispheric motor tracts cross the midline andCSTs projecting from the motor cortex.

2667S. Rose et al. / NeuroImage 59 (2012) 2661–2669

regions using voxelwise analyses, as highlighted in recent review ar-ticles (Agosta et al., 2010a; Turner and Modo, 2010). Our study ex-tends this work to show involvement of corticomotor pathwayslinking the rh.preCG and right paracentral gyrus (rh.paraCG). Inter-estingly, the FA was significantly reduced, bilaterally, within theWM pathways connecting the preCG with posterior cingulate cortex(PCC). Recently, Beckmann et al. (2009), using a strategy involvingprobabilistic tractography and atlas-based parcellation of the cortexthat matches known architecture in the macaque brain, showed sig-nificant connectivity between the preCG and PCC in the humanbrain. Our independent connectivity analysis confirms this finding.A meta-analysis of functional MRI (fMRI) studies utilising motor par-adigms has reported consistent activation within posterior regions ofthe cingulate cortex (Beckmann et al., 2009; Picard and Strick, 2001)indicating that the PCC plays an integral role within motor neural net-works. Thus, the loss in the integrity of this motor network may con-tribute to the clinical features presented by patients with mixed UMNand LMN signs.

Recent neuroimaging studies have highlighted the role of the cor-pus callosum involvement in the potential spread of ALS neuropathol-ogy (Agosta et al., 2007; Ciccarelli et al., 2009; Filippini et al., 2010;Sage et al., 2009; Verstraete et al., 2010). Once again our voxelwiseTBSS analysis of FA measures corroborates these findings. However,the results obtained from our automated structural connectivity anal-ysis specifically show that interhemispheric connections of the leftpreCG with the right preCG, right postcentral gyrus (postCG), rightparaCG and right superior frontal gyrus (supFG) are compromisedin ALS. The integrity of a number of connections between the rightpreCG and left supFG and left caudal middle frontal gyrus (caudMF)was also reduced in the patient group. One other connectivitystudy has reported degeneration to the callosal fibres connectingthe precentral gyri in ALS using cortical defined tractography targets(Verstraete et al., 2010). However, this study was limited to connec-tions anatomically linking the precentral gyri. Our novel findingsshowing a reduction in FA within multiple colossal motor pathwaysare supported by pathological studies that report the presence ofdegenerating fibres within the preCG and paraCG, including the cor-pus callosum (Smith 1960) and clinical studies showing that mirrormovements are often found in subjects with early ALS (Bartels et al.,2008; Karandreas et al., 2007; Krampfl et al., 2003), indicating im-paired interhemispheric inhibition. Furthermore, transcranial mag-netic stimulation (TMS) studies have shown deficient transcallosalinhibition in ALS patients at the very earliest stages of disease pro-gression (Witttstock et al., 2007). Our findings give support for the re-cent concept of the corpus callosum being the conduit for the spread

of pathology from one hemisphere to the other (Eisen 2009) andshow that multiple callosal motor pathways are involved with thisprocess.

There are a number of limitations with this study. We studied pa-tients with established ALS and to demonstrate spread of pathologyit would be necessary to study patients in a serial fashion from theearly stages of disease. Thus, it is not possible to determine the focior precise pattern of spread of disease. However, because of thesound reproducibility of the technique, we believe connectivity ana-lyses are ideal tools to investigate these important questions. Fronto-temporal dementia (FTD) was an exclusion criteria for this study;however our patients did not receive extensive neurophysiologicalassessment. Therefore we cannot exclude that some patients withsubtle cognitive impairment may have influenced our results.Our technique is based on the use of cortical targets to identify cor-ticocortical and corticospinal connections. The choice of the precen-tral gyrus is based on the hallmark feature in ALS of thedegeneration of motor neurons (Betz cells) within the primarymotor cortex (Kiernan and Hudson, 1991), along with the knownwidespread degeneration of motor fibres projecting from the post-central gyrus (Smith 1960). However, these cortical targets are asso-ciated with diverse motor functions, not all of which may be affectedwith ALS pathology. To overcome this problem, one could use anfMRI motor task to identify cortical targets that are directly associat-ed with perturbed limb function for driving the connectivity analysisin ALS patients. In this study we used FA to investigate the integrityof WM pathways. Although the precise neural correlates of alteredFA measures are unknown, strong evidence suggests that a reduc-tion in FA is a useful marker of axonal degeneration, reflectingchanges in mediation, fibre density and packing (Concha et al.,2010; Mädler et al., 2008). As the measure of FA is derived fromthe tensor model, we cannot rule out some contamination withinthis measure of anisotropy from crossing fibres. To reduce this affect,we have employed a higher order diffusion model to extract themajor WM pathways projecting from the precentral and postcentralgyrus, and have statistically compared the mean FA measure derivedover the entire WM pathway. In both our TBSS and connectivityanalysis, we found no correlations between FA summary measuresand clinical scores (ALSFRS-R and disease progression) that reacheda level of statistical significance after correcting for multiple compar-isons. This finding is in keeping with a number of other studies thatfailed to find a correlation between CST damage and clinical mea-sures (Agosta et al., 2010b; Ciccarelli et al., 2006). The range of clin-ical scores, and number of study participants may have impacted onthe significance levels of these correlations. Due to our modest

2668 S. Rose et al. / NeuroImage 59 (2012) 2661–2669

patient numbers, this study needs to be expanded to include a muchlarger MND cohort, scanned in a serial fashion, to fully evaluate thisautomated analysis strategy and altered connectivity profiles associ-ated with MND pathology.

As our HARDI acquisition is approximately 9:30 min in duration,involuntary head movement by the participants is highly likely. Fur-thermore, we did not acquire our diffusion data using cardiac gatingtechniques so our raw data will also contain artefacts from cardiacpulsatile motion. For this reason we identified outliers and replacedthese voxels (Morris et al., 2011) to reduce image artefacts thatwould impact on tractography precision and accuracy of the anisotro-py measures (Walker et al., 2011). Interestingly, as demonstrated inFig. 4, the outlier voxels are clustered within the middle and posteriorregions of the corpus callosum where interhemispheric motor tractscross the midline, and within regions surrounding the CSTs projec-tions. It should be noted that we employed the standard tensormodel to identify outlier voxels and that this approach may overesti-mate the number of outliers within voxels containing complex WMpopulations. This may help explain the number of outliers withinWM regions at the level of the corona radiata. However, this approachalso identified outliers in many locations coincident with thosereported by Walker et al., 2011, namely within a number of WMtracts surrounding the ventricles including projections into the cere-bellum, corticospinal tracts and regions within the splenium of thecorpus callosum. Such regions would be most affected by pulsatilemotion of the brain from the cardiac cycle. These findings, in general,highlight the importance of identifying and compensating for motionand physiological noise from diffusion imaging studies in clinicalpopulations. Further improvements in assessing outliers might beachieved using DROP-R (Morris et al., 2011) with a higher order dif-fusion model in place of the standard tensor model.

In summary, we have shown that a primary feature of neurode-generation in patients with mixed UMN and LMN phenotypes is aloss in FA within multiple intra- and interhemispheric motor path-ways involving the preCG and postCG. Many of these findings relateto known pathology, clinical features, and TMS experimental resultsregarding inhibition of motor information. The results from thisstudy provide support for the use of automated structural connectiv-ity measures to identify non-invasive markers of ALS neuropathologyin larger serial imaging studies.

Acknowledgments

We wish to acknowledge the Motor Neuron Disease Research In-stitute of Australia for their funding support and the significant con-tribution by the late Dr Jonathan Chalk in the important planningstages of this project.

References

Abe, O., Yamada, H., Masutani, Y., Aoki, S., Kunimatsu, A., Yamasue, H., Fukuda, R., Kasai,K., Hayashi, N., Masumoto, T., Mori, H., Soma, T., Ohtomo, K., 2004. Amyotrophiclateral sclerosis: diffusion tensor tractography and voxel-based analysis. NMRBiomed. 17 (6), 411–416.

Agosta, F., Pagani, E., Rocca, M.A., Caputo, D., Perini, M., Salvi, F., Prelle, A., Filippi, M.,2007. Voxel-based morphometry study of brain volumetry and diffusivity in amyo-trophic lateral sclerosis patients with mild disability. Hum. Brain Mapp. 28 (12),1430–1438.

Agosta, F., Chiò, A., Cosottini, M., De Stefano, N., Falini, A., Mascalchi, M., Rocca, M.A.,Silani, V., Tedeschi, G., Filippi, M., 2010a. The present and the future of neuroimag-ing in amyotrophic lateral sclerosis. AJNR Am. J. Neuroradiol. 31 (10), 1769–1777.

Agosta, F., Pagani, E., Petrolini, M., Caputo, D., Perini, M., Prelle, A., Salvi, F., Filippi, M.,2010b. Assessment of white matter tract damage in patients with amyotrophiclateral sclerosis: a diffusion tensor MRI imaging tractography study. AJNR Am.J. Neuroradiol. 31, 1457–1461.

Aoki, S., Iwata, N.K., Masutani, Y., Yoshida, M., Abe, O., Ugawa, Y., Masumoto, T., Mori,H., Hayashi, N., Kabasawa, H., Kwak, S., Takahashi, S., Tsuji, S., Ohtomo, K., 2005.Quantitative evaluation of the pyramidal tract segmented by diffusion tensor trac-tography: feasibility study in patients with amyotrophic lateral sclerosis. Radiat.Med. 23 (3), 195–199.

Bai, Y., Alexander, D.C., 2008. Model-based registration to correct for motion betweenacquisitions in diffusion MR imaging. IEEE International Symposium on BiomedicalImaging: from Nano to Macro, 1-4, pp. 947–950.

Bartels, C., Mertens, N., Hofer, S., Merboldt, K.D., Dietrich, J., Frahm, J., Ehrenreich, H.,2008. Callosal dysfunction in amyotrophic lateral sclerosis correlates withdiffusion tensor imaging of the central motor system. Neuromuscul. Disord. 18 (5),398–407.

Basser, P.J., Mattiello, J., LeBihan, D., 1994. Estimation of the effective self-diffusion ten-sor from the NMR spin echo. J. Magn. Reson. B 103, 247–255.

Beaulieu, C., 2002. The basis of anisotropic water diffusion in the nervous system — atechnical review. MNR BioMed. 15, 435–455.

Beckmann, M., Johansen-Berg, H., Rushworth, M.F.S., 2009. Connectivity-based parcel-lation of the human cingulated cortex and its relation to functional specialization.J. Neurosci. 29, 1175–1190.

Blain, C.R.V., Brunton, S., Williams, V.C., Leemans, A., Turner, M.R., Anderson, P.M.,Catani, M., Stanton, B.R., Ganesalingham, J., Jones, D.K., Williams, S.C.R., Leigh,P.N., Simmons, A., 2011. Differential corticospinal tract degeneration in homozy-gous ‘D90A’ SOD-1 ALS and sporadic ALS. J. Neurol. Neurosurg. Psychiatry 82 (8),843–849.

Brooks, B.R., Miller, R.G., Swash, M., Munsat, T.L., 2000. El Escorial revisited: revised cri-teria for the diagnosis of amyotrophic lateral sclerosis. Amyotroph. Lateral SclerOther Motor Neuron. Disord. 1, 293–299.

Chang, L.-C., Jones, D.K., Pierpaoli, C., 2005. RESTORE: robust estimation of tensors byoutlier rejection. Magn. Reson. Med. 53 (5), 1088–1095.

Ciccarelli, O., Behrens, T.E., Altmann, D.R., Orrell, R.W., Howard, R.S., Johansen-Berg, H.,Miller, D.H., Matthews, P.M., Thompson, A.J., 2006. Probabilistic diffusion tracto-graphy: a potential tool to assess the rate of disease progression in amyotrophiclateral sclerosis. Brain 129, 1859–1871.

Ciccarelli, O., Behrens, T.E., Johansen-Berg, H., Talbot, K., Orrell, R.W., Howard, R.S.,Nunes, R.G., Miller, D.H., Mathews, P.M., Thompson, A.J., Smith, S., 2009. Investiga-tion of white matter pathology in ALS and PLS using tract-based spatial statistics.Hum. Brain Mapp. 30, 615–624.

Concha, L., Livy, D.L., Beaulieu, C., Wheatley, B.M., Gross, D.W., 2010. In vivo diffusiontensor imaging and histopathology of the fimbria–fornix in temporal lobe epilepsy.J. Neurosci. 30 (3), 996–1002.

Desikan, R.S., Ségonne, F., Fischl, B., Quinn, B.T., Dickerson, B.C., Blacker, D., Buckner,R.L., Dale, A.M., Maguire, R.P., Hyman, B.T., Albert, M.S., Killiany, R.J., 2006. An au-tomated labelling system for subdividing the human cerebral cortex on MRIscans into gyral based regions of interest. NeuroImage 31 (3), 968–980.

Eisen, A., 2009. Amyotrophic lateral sclerosis — evolutionary and other perspectives.Muscle Nerve 40, 297–304.

Eisen, A., Weber, M., 2001. The motor cortex and amyotrophic lateral sclerosis. MuscleNerve 24, 564–573.

Filippini, N., Douaud, G., Mackay, C.E., Knight, S., Talbot, K., Turner, M.R., 2010. Corpuscallosum involvement is a consistent feature of amyotrophic lateral sclerosis.Neurology 75, 1645–1652.

Fischl, B., van der Kouwe, A., Destrieux, C., Halgren, E., Segonne, F., Salat, D.H., Busa, E.,Seidman, L.J., Goldstein, J., Kennedy, D., Caviness, V., Makris, N., Rosen, B., Dale, A.M.,2004. Automatically parcellating the human cerebral cortex. Cereb. Cortex 14 (1),11–22.

Griswold, M.A., Jakob, P.M., Heidemann, R.M., Nittka, M., Jellus, V., Wang, J.M., Kiefer, B.,Haase, A., 2002. Generalized autocalibrating partially parallel acquisitions (GRAP-PA). Magn. Reson. Med. 47 (6), 1202–1210.

Hagmann, P., Cammoun, L., Gigandet, X., Meuli, R., Honey, C.J., Wedeen, V.J., 2008.Mapping the structral core of human cerebral cortex. PloS One 6, E159.

Hong, Y.H., Sung, J.J., Kim, S.M., Park, K.S., Lee, K.W., Chang, K.H., Song, I.C., 2008. Diffu-sion tensor tractography-based analysis of the pyramidal tract in patients withamyotrophic lateral sclerosis. J. Neuroimaging 18 (3), 282–287.

Ilieva, H., Polymenidou, M., Cleveland, D.W., 2009. Non-cell autonomous toxicity inneurodegenerative disorders: ALS and beyond. J. Cell Biol. 187, 761–772.

Jenkinson, M., 2003. A fast, automated, n-dimensional phase unwrapping algorithm.Magn. Reson. Med. 49 (1), 193–197.

Jenkinson, M., 2004. Improving the registration of B0-distorted EPI images using calcu-lated cost function weights. Tenth Annual Meeting of the Organization for HumanBrain Mapping.

Johansen-Berg, H., Rushworth, M.F.S., 2009. Using diffusion imaging to study humanconnectional anatomy. Annu. Rev. Neurosci. 32, 75–94.

Jones, D.K., 2010. Challenges and limitations of quantifying brain connectivity in vivowith diffusion MRI. Imaging Med. 2 (3), 341–355.

Jones, D.K., Cercignani, M., 2010. Twenty-five pitfalls in the analysis of diffusion MRIdata. NMR Biomed. 23, 803–820.

Karandreas, N., Papaopoulou, M., Kokotis, P., Papapostolou, A., Tsivgoulis, G., Zambelis,T., 2007. Impaired interhemispheric inhibition in amyotrophic lateral sclerosis.Amyotroph. Lateral Scler. 8, 112–118.

Keller, J., Vymazal, J., Ridzon, P., Rusina, R., Kulistak, P., Malikova, H., Rulseh, A.M., Keler,O., Jech, R., 2011. Quantitative brain MR imaging in amyotrophic lateral sclerosis.Magn. Reson. Mater. Phy. 24, 67–76.

Kiernan, J.A., Hudson, A.J., 1991. Changes in sizes of cortical and lower motor neuronsin amyotrophic lateral sclerosis. Brain 114, 843–853.

Kiernan, M.C., Vucic, S., Cheah, B.C., Turner, M.R., Eisen, A., Hardiman, O., Burrell, J.R.,Zoing, M.C., 2011. Amyotrophic lateral sclerosis. Lancet 377 (9769), 942–955.

Krampfl, K., Petri, S., Gotz, F., Mohammadi, B., Bufler, J., 2003. Amyotrophic lateral scle-rosis (ALS) and mirror movements in a patient with polymicrogyria. Amyotroph.Lateral Scler. 4, 266–269.

Leemans, A., Jones, D.K., 2009. The B-matrix must be rotated when correcting for sub-ject motion in DTI data. Magn. Reson. Med. 61 (6), 1336–1349.

2669S. Rose et al. / NeuroImage 59 (2012) 2661–2669

Mädler, B., Drabycz, S.A., Kolind, S.H., Whittall, K.P., MacKay, A.L., 2008. Is diffusion anisot-ropy an accuratemonitor ofmyelination?Correlationofmulticomponent T2 relaxationand diffusion tensor anisotropy in human brain. Magn. Reson. Imaging 26, 874–888.

Metwalli, N.S., Benatar, M., Nair, G., Usher, S., Hu, X., Carew, J.D., 2010. Utility of axialand radial diffusivity from diffusion tensor MRI as markers of neurodegenerationin amyotrophic lateral sclerosis. Brain Res. 1348, 156–164.

Morris, D., Nossin-Manor, R., Taylor, M.J., Sled, J.G., 2011. Preterm neonatal diffusionprocessing using detection and replacement of outliers prior to resampling. Magn.Reson. Med. 66 (1), 92–101.

Nichols, T.E., Holme, A.P., 2002. Nonparametric permutation tests for functional neuro-imaging: a primer with examples. Hum. Brain Mapp. 15, 1–25.

Pannek, K., Mathias, J.L., Bigler, E.D., Brown, G., Taylor, J.D., Rose, S.E., 2010. An automat-ed strategy for the delineation and parcellation of commissural pathways suitablefor clinical populations utilising high angular resolution diffusion imaging tracto-graphy. NeuroImage 50 (3), 1044–1053.

Pannek, K., Mathias, J.L., Bigler, E.D., Brown, G., Taylor, J.D., Rose, S.E., 2011. The averagepathlength map: a diffusion MRI tractography-derived index for studying brain pa-thology. NeuroImage 55, 133–141.

Picard, N., Strick, P.L., 2001. Imaging thepremotor areas. Curr. Opin. Neurobiol. 11, 663–672.Ravits, J.M., La Spada, A.R., 2009. ALS motor phenotype heterogeneity, focality, and

spread. Neurology 73, 805–811.Rohde, G.K., Barnett, A.S., Basser, P.J., Marenco, S., Pierpaoli, C., 2004. Comprehensive

approach for correction of motion and distortion in diffusion-weighted MRI.Magn. Reson. Med. 51 (1), 103–114.

Rothstein, J.D., 2009. Current hypotheses for the underlying biology of amyotrophic lat-eral sclerosis. Ann. Neurol. 65, S3–S9.

Sage, C.A., Peeters, R.R., Gorner, A., Robberecht, W., Sunaert, S., 2007. Quantitative dif-fusion tensor imaging in amyotrophic lateral sclerosis. NeuroImage 15, 486–499.

Sage, C.A., Van Hecke, W., Peeters, R.R., Sijbers, J., Robberecht, W., Parizel, P., Marchal,G., Leemans, A., Sunaert, S., 2009. Quantitative diffusion tensor imaging in amyo-trophic lateral sclerosis: revisited. Hum. Brain Mapp. 30 (11), 3657–3675.

Sato, K., Aoki, S., Iwata, N.K., Masutani, Y., Watadani, T., Nakata, Y., Yoshido, M., Terao,Y., Abe, O., Ohtomo, K., Tsuji, S., 2010. Diffusion tensor tract-specific analysis of theuncinated fasciculus in patients with amyotrophic lateral sclerosis. Neuroradiology52 (8), 729–733.

Senda, J., Ito, M., Wananabe, H., Atsuta, N., Kawai, Y., Katsuno, M., Tanaka, F., Naganawa,S., Fukatsu, H., Sobue, G., 2009. Correlation between pyramidal tract degenerationand widespread white matter involvement in amyotrophic lateral sclerosis: astudy with tractography and diffusion-tensor imaging. Amyotroph. Lateral Scler.10 (5–6), 288–294.

Senda, J., Kato, S., Kaga, T., Ito, M., Atsuta, N., Nakamura, T., Watanabe, H., Tanaka, F.,Naganawa, S., Sobue, G., 2011. Progressive and widespread brain damage in ALS:MRI voxel-based morphometry and diffusion tensor imaging study. Amyotroph.Lateral Scler. 12, 59–69.

Sled, J.G., Zijdenbos, A.P., Evans, A.C., 1998. A nonparametric method for automatic cor-rection of intensity nonuniformity in MRI data. IEEE Trans. Med. Imaging 17,87–97.

Smith, M.C., 1960. Nerve fibre degeneration in the brain in amyotrophic lateral sclero-sis. J. Neurol. Neurosurg. Psychiatry 23, 269–282.

Smith, S.M., Nichols, T.E., 2009. Threshold-free cluster enhancement: addressing prob-lems of smoothing, threshold dependence and localisation in cluster inference.NeuroImage 44 (1), 83–98.

Smith, S.M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T.E., Mackay, C.E.,Watkins, K.E., Ciccarelli, O., Cader, M.Z., Matthews, P.M., Behrens, T.E., 2006.Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data.NeuroImage 31, 1487–1505.

Stanton, B.R., Shinhmar, D., Turner, M.R., Williams, V.C., Williams, S.C., Blain, C.R., Giam-pieto, V.P., Catani, M., Leight, P.N., Anderson, P.M., Simmons, A., 2009. Diffusiontensor imaging in sporadic and familial (D90A SOD1) forms of amyotrophic lateralsclerosis. Arch. Neurol. 66 (1), 109–115.

Tournier, J.D., Calamante, F., Connelly, A., 2007. Robust determination of the fiber ori-entation distribution in diffusion MRI: non-negativity constrained super-resolvedspherical deconvolution. NeuroImage 35 (4), 1459–1472.

Turner, M.R., Modo, M., 2010. Advances in the application of MRI to amyotrophic later-al sclerosis. Expert Opin. Med. Diagn. 4 (6), 483–496.

van der Graaff, M.M., Sage, C.A., Cann, M.W., Akkerman, E.M., Lavini, C., Majoie, C.B.,Nederveen, A.J., Zwinderman, A.H., Vos, F., Brugman, F., van den Berg, L.H., deRijk, C., van Doorn, P.A., Hecke, W.V., Peeters, R.R., Robberecht, W., Sunaert, S., deVisser, M., 2011. Upper and extra-motoneuron involvement in ealry motoneurondisease: a diffusion tensor imaging study. Brain 134 (Pt 4), 1211–1228.

Verstraete, E., van den Heuvel, M.P., Veldink, J.H., Blanken, N., Mandl, R.C., HulshoffPol, H.E., van den Berg, L.H., 2010. Motor network degeneration in amyotrophiclateral sclerosis: a structural and functional connectivity study. PloS One 5 (10),e13664.

Walker, L., Chang, L.C., Koay, C.G., Sharma, N., Cohen, L., Verma, R., Pierpaoli, C., 2011.Effects of physiological noise in population analysis of diffusion tensor MRI data.Neuroimage 2011 Jan 15;54(2):1168–1177. (Electronic publication ahead ofprint 2010 Sep 8).

Witttstock, M., Wolters, A., Benecke, R., 2007. Transcallosal inhibition in amyotrophiclateral sclerosis. Clin. Neurophysiol. 118, 301–307.