Abnormalities of language networks in temporal lobe epilepsy

Abnormalities of language networks in temporal lobe epilepsy

H.W. Robert Powell,a Geoff J.M. Parker,b Daniel C. Alexander,c Mark R. Symms,a

Philip A. Boulby,a Claudia A.M. Wheeler-Kingshott,d Gareth J. Barker,e

Matthias J. Koepp,a and John S. Duncana,!

aDepartment of Clinical and Experimental Epilepsy, Institute of Neurology, University College London, Queen Square, London, WC1N 3BG, UKbImaging Science and Biomedical Engineering, University of Manchester, Oxford Road, Manchester, UKcDepartment of Computer Science, University College London, Gower Street, London, UKdNMR Research Unit, Institute of Neurology, University College London, London, UKeKing’s College London, Institute of Psychiatry, Department of Clinical Neuroscience, Centre for Neuroimaging Sciences, London, UK

Received 30 October 2006; revised 24 January 2007; accepted 10 February 2007Available online 28 February 2007

Functional magnetic resonance imaging (fMRI) in patients withtemporal lobe epilepsy (TLE) has demonstrated reorganisation oflanguage functions with greater involvement of the non-dominanthemisphere. The structural brain connections supporting this atypicallanguage dominance have not previously been identified. Weperformed fMRI of language functions and imaging of white matterconnections using MR tractography in 14 patients with unilateral TLEand hippocampal sclerosis and 10 controls. Verb generation andreading comprehension paradigms were used to define functionalregions which were used to generate starting regions for tractography.Controls and right TLE patients had a left-lateralised pattern of bothlanguage-related activations and the associated structural connections.Left TLE patients showed more symmetrical language activations,along with reduced left hemisphere and increased right hemispherestructural connections. Subjects with more lateralised functionalactivation had also more highly lateralised connecting pathways. Weprovide evidence for structural reorganisation of white matter tractsthat reflects the altered functional language lateralisation in left TLEpatients. The combination of fMRI and tractography offers apromising tool for studying the reorganisation of language functionsin many neurological conditions and may prove useful in predictinglanguage deficits following temporal lobe surgery.© 2007 Elsevier Inc. All rights reserved.

Introduction

The relationship between brain structure and function isfundamental to understanding both cognitive neuroscience andclinical neurology. There is good reason to expect a strongcorrespondence between regional brain function and the underlyingneuronal architecture; however, studying this relationship iscomplicated by the intricacy of the brain’s connections, and thelack of techniques for exploring connectivity in vivo. Recent

advances in neuroimaging now allow such relationships to bestudied non-invasively, with functional magnetic resonance (fMRI)providing the location of functional regions and diffusion-weightedimaging allowing exploration of the anatomical connectivity ofthese regions.

Temporal lobe epilepsy (TLE) due to hippocampal sclerosis(HS) is a common cause of medically refractory seizures but has agood outcome following anterior temporal lobe resection (ATLR).TLE may be associated with disrupted lateralisation and localisa-tion of language. fMRI studies have demonstrated a higher in-cidence of atypical language dominance in preoperative left TLEpatients (Adcock et al., 2003; Thivard et al., 2005), suggesting thatdominant hemisphere lesions can lead to functional reorganisationof language to contralateral cortical regions. The recruitment ofright hemisphere regions not activated in control subjects wasinvolved in maintaining reading skills in patients following leftATLR (Noppeney et al., 2005).

Verbal fluency fMRI tasks show relative shifts in activation tothe right hemisphere in patients with left temporal lobe seizures(Adcock et al., 2003; Lehericy et al., 2000; Voets et al., 2006;Woermann et al., 2003), probably due to epileptic activity affectingnormal left hemisphere language organisation (Janszky et al.,2003). Greater right hemisphere reorganisation has been demon-strated for receptive, temporal lobe language functions than forproductive, frontal lobe language functions (Thivard et al., 2005).

Language function may also be affected by subtle abnormalitiesremote from the hippocampus. Evidence from a number of imagingmodalities including quantitative MRI and MR volumetry (Moranet al., 2001; Sisodiya et al., 1997), MR-spectroscopy (Stefan et al.,2001), voxel based morphometry (Keller et al., 2002) and positronemission tomography (Hammers et al., 2001) suggests theexistence of more widespread abnormalities in TLE associatedwith HS than demonstrated by conventional MRI. The strongconnectivity between the inferior frontal and medial temporal lobe(MTL) may make frontal lobe functions particularly sensitive toMTL pathology (Powell et al., 2004), while electrophysiological

www.elsevier.com/locate/ynimgNeuroImage 36 (2007) 209–221

! Corresponding author. Fax: +44 207 8373941.E-mail address: [email protected] (J.S. Duncan).Available online on ScienceDirect (www.sciencedirect.com).

1053-8119/$ - see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.neuroimage.2007.02.028

disruption of extratemporal regions may lead to cognitive impair-ments, in particular on tests of executive function (Hermann andSeidenberg, 1995). Knowing whether any structural changesexisted within the white matter pathways involved in languagefunction in TLE patients would be clinically useful as well asproviding important insights into the mechanisms underlyingfunctional plasticity within cerebral networks.

Diffusion tensor imaging (DTI) is an MRI technique thatevaluates brain structure by measurement of water molecules’ three-dimensional diffusion through tissue. The diffusion tensor and othermodels of diffusion allow the directionality of the motion (aniso-tropy) and the principal orientation(s) of diffusion for each voxel inthe brain to be calculated (Pierpaoli et al., 1996; Pierpaoli and Bas-ser, 1996). These provide the basis for MR tractography, a tech-nique that non-invasively investigates the trajectories of cerebralwhite matter tracts (Conturo et al., 1999; Mori et al., 1999). In thisstudy we used a combination of fMRI and tractography todemonstrate the structure of the white matter tracts connectingfunctionally defined language regions in patients with TLE andunilateral HS and hypothesised that the language connections in leftTLE patients would reflect the pattern of functional activation, withreduced ipsilateral and increased contralateral connections. We alsolooked for correlations between structure and function, andhypothesised that those subjects with more lateralised languagefunction would have more lateralised structural connections.

Materials and methods

Subjects

We studied 14 patients (median age 35.5; range 25–47 years; 7female; 13 right-handed) with medically refractory TLE undergoingpre-surgical evaluation at the National Hospital for Neurology andNeurosurgery (NHNN), London, UK. Structural MRI at 1.5 Tshowed unilateral left HS in 7 patients and right HS in 7 onqualitative and quantitative criteria (Woermann et al., 1998). Video-EEG had confirmed seizures arising from the ipsilateral medialtemporal lobe in all 14. Post-operative histology confirmed HS in allcases. All patients were on antiepileptic medication and all werefluent English speakers. Patients’ details are given in Table 1. Aspreviously reported (Powell et al., 2006), we also studied 10 right-handed native English-speaking healthy volunteers (median age29.5; range 23–50 years; 6 female) with no history of neurological orpsychiatric disease. Handedness was determined using the Edin-burgh Hand Preference Inventory. Language dominance wasassessed using fMRI. The study was approved by the NHNN andthe Institute of Neurology Joint Research Ethics Committee andinformed written consent was obtained from all subjects.

MR data acquisition

MRI studies were performed on a 1.5 T General Electric SignaHorizon scanner. Standard imaging gradients with a maximumstrength of 22 mTm!1 and slew rate 120 Tm!1 s!1 were used. Alldata were acquired using a standard quadrature birdcage head coilfor both RF transmission and reception.

Functional MRI

For the language tasks, gradient-echo echo-planar T2*-weighted images were acquired, providing blood oxygenation

level dependent (BOLD) contrast. Each volume comprised 17contiguous 4.6 mm axial slices, with a 22 cm field of view and96!96 acquisition matrix, reconstructed as a 128!128 matrixgiving an in-plane voxel size of 1.7 mm!1.7 mm. TE was 40 msand TR was 4.5 s. The field of view was positioned to maximisecoverage of the frontal and temporal lobes. A single volume EPIwas acquired with similar parameters and equal sensitivity togeometric distortions but a longer TR (Boulby et al., 2004). Thisallowed whole-brain coverage and was used as an anatomicalreference and to aid spatial normalisation.

Each subject performed a verb generation and a reading com-prehension task. Both tasks used a blocked experimental designwith 30 s task blocks alternating with 30 s of rest over 5 1/2 min.During the verb generation task concrete, nouns were visuallypresented every 3 s in blocks of 10 nouns. Subjects were instructedto covertly generate verbs from the nouns during the task blockand to silently repeat the nouns during the rest block. Thisparadigms were used to identify anterior language regions in theinferior and middle frontal gyri (Liegeois et al., 2004; Woermannet al., 2003).

During the reading comprehension activation condition, sub-jects silently read 9-word sentences covering a range of differentsyntactic structures and semantic content. No explicit task wasrequired during scanning in order to avoid inducing additionalexecutive processes. In the baseline condition, subjects attentivelyviewed 9-word sentences after all the letters were transformed intofalse fonts. This baseline controlled for visual input but notlexical, semantic or syntactic content. The paradigm thereforemaximised our chances of seeing reading-related activation at anylevel of the reading system. Blocks of six sentences wereinterleaved with blocks of six false font sentences. Sentences andfalse fonts were presented one word at a time at a fixed rate (wordduration 500 ms, sentence duration 5000 ms; block length: 30 s).This serial presentation mode was used to control for visual input,eye movements and to equate the subjects’ reading pace. Thisparadigm was designed to identify posterior language regions inthe superior and middle temporal gyri (Gaillard, 2004; Noppeneyet al., 2005).

The data were analysed with statistical parametric mapping(using SPM2 software from the Wellcome Department of ImagingNeuroscience, London; http://www.fil.ion.ucl.ac.uk/spm). Scansfrom each subject were realigned using the first as a reference,spatially normalised (Friston et al., 1995) into standard space(using the whole brain EPI) (Talairach and Tournoux, 1988),resampled to 3!3!3 mm3 voxels and spatially smoothed with aGaussian kernel of 10 mm FWHM. The time-series in each voxelwas highpass filtered with a cutoff of 1/128 Hz. A two-levelrandom effects analysis was employed.

At the first level, condition-specific effects for each subjectwere estimated according to the general linear model (Friston et al.,1995). Regressors of interest were formed for each task by creatingboxcar functions of task against rest. Parameter estimates for theseregressors were then calculated for each voxel. Two contrastimages were produced for each subject, corresponding to the maineffects of verb generation and reading comprehension against thecontrol conditions. All these images were used for the second-levelanalysis.

At the second level of the random effects analysis, eachsubject’s contrast image was entered into a one-sample T-test toexamine the main effect of verb generation and reading compre-hension within each group. For each task, we derived lateralisation

210 H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

Table 1Clinical and demographic details for all patients

# Age/gender

Handedness Epilepsyonset (years)

Seizure typesand frequency(per month)

MRI andpathologicaldiagnosis

Clinicaland EEG

VIQ PIQ Naming Fluency Right HV (cm3) Left HV (cm3) Hippocampalvolume ratio(%)

AEDs(mg/day)

Languagedominance

1 25/F Right 17 CPS 8 Left HS Left TLE 82 80 7 33 2.729 1.351 50 TPR 150 LeftSGTC 0.5 LTG 200

2 37/M Left 1 SPS 12 Left HS Left TLE 94 93 19 34 3.3 1.88 57 VPA 800 LeftCPS 4 CBZ 800

LVT 20003 33/M Right 1 SPS 4 Left HS Left TLE 86 88 13 21 2.83 2.02 71 PMD 500 Left

CPS 4 CBZ 1200CLB 10TPR 175LVT 4000

4 32/F Right 22 SPS 30 Left HS Left TLE 88 111 13 28 2.359 1.789 76 OXC 1050 LeftCPS 5 TPR 200SGTC 0.25

5 28/M Right 3 CPS 1 Left HS Left TLE 76 81 4 25 3.077 1.417 46 LVT 3000 LeftLTG 600

6 34/M Right 21 CPS 5 Left HS Left TLE 67 76 6 14 2.235 1.896 85 LVT 2000 LeftVPA 2000

7 37/F Right 1 SPS 12 Left HS,left fusiform gyrusganglioglioma

Left TLE 70 94 10 24 2.8 2.385 85 CBZ 1000 LeftCPS 8 CLB 10SGTC 1

8 47/M Right 13 CPS 1 Right HS Right TLE 73 86 7 21 1.038 2.769 37 LVT 500 LeftPHT 300CBZ 800

9 44/F Right 14 CPS 4 Right HS Right TLE 95 104 22 36 2.606 3.034 86 LVT 750 LeftPHT 400CLB 10

10 44/M Right 10 CPS 4 Right HS Right TLE 117 118 22 51 2.775 3.047 91 CBZ 800 LeftLVT 1000

11 46/F Right 8 CPS 5 Right HS Right TLE 88 84 10 20 1.756 2.695 65 TPR 600 LeftPMD 1000OXC 2400

12 32/F Right 19 CPS 2 Right HS Right TLE 87 110 15 26 2.299 2.562 90 CBZ 1600 LeftLVT 1000

13 29/M Right 9 CPS 30 Right HS Right TLE 93 114 20 27 2.273 2.978 76 VPA 2400 LeftSGTC 4

14 41/F Right 14 CPS 4 Right HS Right TLE 90 88 13 20 1.628 2.302 71 TGB 15 Left

M=male; F=female; SPS=simple partial seizure; CPS=complex partial seizure; SGTC=secondary generalised tonic–clonic seizure; HS=hippocampal sclerosis; EEG=electroencephalogram; TLE=temporal lobeepilepsy; VIQ=verbal intelligence quotient; PIQ=performance intelligence quotient; HV=hippocampal volume; AED=antiepileptic drug; TPR=topiramate; LTG=lamotrigine; VPA=sodium valproate;CBZ=carbamazepine; LVT=levetiracetam; PMD=primidone; CLB=clobazam; OXC=oxcarbazepine; GBP=gabapentin; PHT=phenytoin; TGB=tiagabine. Hippocampal volume ratio is the ratio of ipsilateral(sclerotic) to contralateral hippocampus.

211H.W.R.P

owell

etal.

/NeuroIm

age36

(2007)209–221

indices for each subject’s mean fMRI activation (LI=L!R), overspherical regions of interest (10 mm radius) centred on eachgroup’s peak frontal and temporal activations. A very low SPMthreshold was employed (P<0.5 uncorrected) in order to includethe contributions of all voxels within the regions of interest. Inaddition, we performed a one-way ANOVA to identify brainregions demonstrating greater or less activation in patientscompared with controls.

Maps of the main effect of each task across all subjects(controls and patients combined) were generated, thresholded atP<0.001 (uncorrected) and transformed into each individual’snative space. This uncorrected threshold was adopted on the basisof a strong a priori hypothesis as to the areas of expectedactivation. These reverse-normalised group maps were then usedto define volumes of interest (VOIs) for initiating probabilisticfibre tracking. A total of four VOIs were defined for each subject,one each in the left and right inferior frontal gyri (IFG) and leftand right superior temporal gyri (STG), by drawing over selectedareas of fMRI activation (see results for details) on consecutivebrain slices, using MRIcro (http://www.psychology.nottingham.ac.uk). Each VOI was of identical volume, comprising 125 voxels.

Diffusion tensor imaging

The DTI acquisition sequence was a single-shot spin-echoecho planar imaging (EPI) sequence, cardiac gated (triggeringoccurring every QRS complex) (Wheeler-Kingshott et al., 2002),with TE=95 ms. The acquisition matrix (96!96, 128!128reconstructed), field of view (22 cm!22 cm) and in-plane reso-lution on reconstruction (1.7 mm!1.7 mm) were identical to thefMRI data. Acquisitions of 60 contiguous 2.3 mm-thickness axialslices were obtained, covering the whole brain, with diffusionsensitising gradients applied in each of 54 non-colinear directions(maximum b value of 1148 s mm!2 (!=34 ms, "= 40 ms, usingfull gradient strength of 22 mTm!1)) along with 6 non-diffusion-weighted (b=0) scans. The DTI acquisition time for a total of3600 images was approximately 25 min (depending on the heartrate).

The diffusion tensor eigenvalues #1, #2, #3 and eigenvectors $1,$2, $3 were calculated, and fractional anisotropy (FA) maps weregenerated (Pierpaoli et al., 1996; Pierpaoli and Basser, 1996). Weused the method of Parker and Alexander (Parker et al., 2003;Parker and Alexander, 2003) to reduce fibre orientation ambi-guities in voxels containing fibre crossings. Voxels in which thesingle tensor fitted the data poorly were identified using thespherical–harmonic voxel-classification algorithm of Alexanderet al. (2002). In these voxels a mixture of two Gaussian probabilitydensities was fitted and the principal diffusion directions of the twodiffusion tensors provided estimates of the orientations of thecrossing fibres (Tuch et al., 2002). In all other voxels a singletensor model was fitted.

We used the Probabilistic Index of Connectivity (PICo)algorithm extended to cope with crossing fibres (Parker et al.,2003; Parker and Alexander, 2003) to track from the functionallydefined VOIs. This algorithm adapts the commonly used stream-line approach to exploit the uncertainty due to noise in one ormore fibre orientations defined for each voxel. This uncertainty isdefined using probability density functions (PDFs) constructedusing simulations of the effect of realistic data noise on fibredirections obtained from the mixture model (Parker andAlexander, 2003). The streamline process is repeated using Monte

Carlo methods to generate maps of connection probability orconfidence of connection from the chosen start regions.

Each output connectivity map was spatially normalised usingSPM and binary masks at a threshold connection probability of0.05 were constructed. Previously we demonstrated that increasingthe probability threshold had no effect on the lateralisation of tractvolume and no effect on mean FA apart from at very lowthresholds (Powell et al., 2006). The chosen threshold strikes abalance between losing non-specific low probability connectionswhile retaining the main body of the pathways. The binary maskswere averaged across each group, to produce variability (orcommonality) maps indicating the degree of spatial variability andoverlap of the identified connections (Parker et al., 2005). A voxelcommonality value C of 1.0 indicates that each individual had aconnection identified in this voxel while C of 0.0 indicates thatnone of them did (Parker et al., 2005). Normalised tract volumeswere calculated for the left and right tracts of every subject (Toosyet al., 2004) and paired t-tests were used to compare these withingroups. Unpaired t-tests were used to compare the patient andcontrol groups. Two repeated measures ANOVAs were used to testfor interactions between group and hemisphere, one comparingcontrols and the left TLE group and the other comparing left andright TLE groups.

To calculate the mean FA of the connected volume in nativespace for the left and right tracts, the binary masks defined abovewere multiplied with that subject’s whole brain FA image. Themean FA value is the mean intensity value of the voxels within theresulting masked region. We used a paired t-test to compare leftand right-sided mean FA within each group. Unpaired t-tests wereused to compare the patient and control groups and repeatedmeasures ANOVAs were used to test for interactions betweengroup and hemisphere as described above.

Correlations between structure and function

We tested for correlations between the lateralisation of fMRIactivation and of the lateralisation of the derived connections.For the fMRI we defined 10 mm radius spheres based on thecoordinates of the peak magnitude of activation in the leftfrontal and bilateral temporal lobe, along with a homotopicvolume in the right frontal lobe, and calculated left minus rightactivation for each subject. For the tractography we calculatedlateralisation indices LI to assess lateralisation of mean FAbetween hemispheres;

LI ! "Mean FAleft #Mean FAright$="Mean FAleft %Mean FAright$

We calculated the Spearman correlation coefficient to testfor correlations between the two values. We also tested forcorrelations between the LI of the tracts and the duration ofepilepsy.

Regression analysis

Finally, patients’ fMRI contrast images were entered into aSPM2 simple regression model with the mean FA of the left-sidedtracts as a covariate. This allowed us to look at a voxel level forregions showing correlation between fMRI activation and the meanFA of the connections of language-related areas (Powell et al.,2006). This was done for both left and right TLE groups and theresulting SPM maps were thresholded at P<0.001.

212 H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

Results

Demographic data

There were no significant differences between left and rightTLE groups in epilepsy duration (left TLE median (range) 25 years(8–36), right TLE 30 years (13–36)), naming score (left TLE mean(SD) 10.3 (5.2), right TLE 15.6 (6)) or fluency score (left TLEmean 25.6 (6.9), right TLE 28.7 (11.3)). Mean verbal IQ was lowerin the left TLE patients (mean (SD) 80.4 (9.9)) than the right TLE(91.9 (13.2)), although this difference was not statisticallysignificant. There were, as expected, significant differences in

hippocampal volumes between the two groups with left TLEpatients having smaller left hippocampi (left TLE mean (SD)1.82 cm3 (0.35), right TLE 2.77 cm3 (0.28)), and right TLEpatients having smaller right hippocampi (left TLE 2.76 cm3

(0.37), right TLE 2.05 cm3 (0.61).

fMRI results

Verb generation was associated with activation in the left IFGin all three groups, with suprathreshold activation in the right IFGin the left TLE group only (Figs. 1A, C and E, Table 2). Readingcomprehension was associated with activation in bilateral STG,

Fig. 1. fMRI results: Main effects of the two language paradigms for each group. Significant regions (threshold P<0.001 unless stated otherwise) aresuperimposed onto the normalisedmean EPI image from all 10 subjects. The left of the brain is displayed on the left of the image. (A) Controls, verb generation, leftinferior frontal gyrus activation. (B) Controls, reading comprehension, bilateral superior temporal gyrus activation, adjacent to the superior temporal sulcus. (C)Left TLE, verb generation, left inferior frontal gyrus activation, along with smaller activation in the right inferior frontal gyrus. (D) Left TLE, readingcomprehension, left superior temporal gyrus activation*. (E) Right TLE, verb generation, left inferior frontal gyrus activation. (F) Right TLE, readingcomprehension, bilateral superior temporal gyrus activation, adjacent to the superior temporal sulcus. (G) Main effects of the three groups combined, verbgeneration. The regions of activation shown in the coronal slices were used to define VOIs for initiating fibre tracking. (H) Main effects of the three groupscombined, reading comprehension. The regions of activation shown in the coronal slices were used to define VOIs for initiating fibre tracking. (I) Lateralisationindices (LIs) (±SE) for each group for frontal lobe activation. An LI of 1 would indicate a fully left lateralised pattern of activationwhile an LI of !1would indicatea fully right lateralised pattern of activation. The LIs for verb generation were significantly more left lateralised in controls and right TLE patients than left TLEpatients (P=0.001). (J) LIs for each group for temporal lobe activation. No significant differences were seen between groups. * figure threshold P<0.01.

213H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

adjacent to the superior temporal sulci, in both controls and rightTLE patients (Figs. 1B and F, Table 2) but left TLE patientsshowed only an area of subthreshold activation in the posteriorleft temporal lobe (Fig. 1D). Frontal lobe LIs for verb generationwere significantly more left lateralised in controls and right TLEpatients than left TLE patients (F=9.530, P=0.001) (Fig. 1I).There was no significant difference in temporal lobe LIs betweenthe three groups for reading comprehension (F=1.092, P=0.35)(Fig. 1J). Four VOIs were defined for initiating fibre trackingbased on the areas of peak activation for all subjects combined(Figs. 1G–H). One was placed in the left IFG and correspondedto the peak IFG activation seen for verb generation. As nosignificant activation was seen in this region on the right for allsubjects combined, a homotopic VOI of identical size wasmanually defined using MRIcro. Two further VOIs correspondedto the areas of bilateral STG activation seen during the readingcomprehension task.

On direct comparison between groups, no brain regionsdemonstrated significantly greater activation for verb generationin either left or right TLE groups compared with controls. LeftTLE patients demonstrated greater activation in a single area inthe right superior frontal gyrus, and right TLE patientsdemonstrated greater activation in the right middle frontal gyrus(Table 2); however, both of these were distant from the definedstart regions of interest. There were no significant differencesbetween left and right TLE patients. For reading, no brain regionsdemonstrated either greater or less activation in patients comparedwith controls and there were no significant differences betweenleft and right TLE groups. Right TLE patients demonstratedgreater activation than left TLE patients in the right middlefrontal gyrus (Table 2), posterior to the defined start region ofinterest.

Tractography results

IFG connectionsThe group variability maps for the volumes of connection from

the frontal lobe VOIs are shown in Fig. 2. As reported previouslyin the control group (Powell et al., 2006), greater connections(particularly fronto-temporal connections via the superior long-itudinal fasciculus (SLF)) were seen in the left hemisphere than theright. Both connection volumes (left 48.5 cm3, right 41 cm3;P<0.05) and mean FA (left 0.317, right 0.308; P<0.05) weregreater on the left than the right.

For the left TLE group, visual assessment of the pattern ofconnections demonstrated consistent bilateral connections extend-ing posteriorly from the IFG with greater temporal lobe andsupramarginal gyrus connections on the right than on the left.Paired T-tests revealed both greater volume of connections (left36.1 cm3, right 44.9 cm3; P=0.018) and mean FA (left 0.301, right0.316; P=0.05) on the right than the left.

For the right TLE group, the pattern of connections is similar tocontrols, with greater fronto-temporal connections via the SLF onthe left than the right. Mean FA was greater on the left than theright (left 0.307, right 0.295; P=0.032) but there was no significantdifference in the volume of connections (left 46.6 cm3, right40.4 cm3; P=0.15).

STG connectionsFig. 3 shows the group variability maps for the volumes of

connection from the temporal lobe VOIs. For the controls, greaterconnections were seen in the left hemisphere than the right, inparticular along the SLF. Both connection volumes (left 44.8 cm3,right 32.3 cm3; P<0.001) and mean FA (left 0.313, right 0.303;P=0.05) were greater on the left than the right.

Table 2Activation peaks for fMRI effects of interest

Group Contrast MNI coordinates Z score Region

Controls Verb generation !44 2 24 5.56* Left inferior frontal gyrus!40 20 0 5.52* Left inferior frontal gyrus

Reading comprehension !44 !56 12 4.99* Left posterior superior temporal gyrus!52 !26 !6 4.63 Left superior temporal gyrus48 !26 !4 3.53 Right superior temporal gyrus

Left TLE Verb generation !40 10 12 4.60 Left inferior frontal gyrus38 6 18 3.18 Right inferior frontal gyrus

Reading comprehension !62 !46 !2 2.78** Left superior temporal gyrusRight TLE Verb generation !42 14 22 5.09* Left inferior frontal gyrus

Reading comprehension ! 46 28 !2 5.04* Left inferior frontal gyrus!52 !52 6 4.90 Left superior temporal gyrus!56 !30 !12 4.04 Left superior temporal gyrus54!28 !12 3.60 Right superior temporal gyrus

All subjects Verb generation !44 10 20 7.47* Left inferior frontal gyrusReading comprehension !50 !34 !6 5.18* Left superior temporal gyrus

52 !26 !6 5.41* Right superior temporal gyrusGroup comparisons Left TLE>controls 50 22 40 3.76 Right superior frontal gyrus

Right TLE>controls 42 12 38 4.03 Right middle frontal gyrusRight TLE> left TLE 46 !8 14 3.77 Right middle frontal gyrus

Left TLE Regression: verb generation !50 !2 12 3.81 Left inferior frontal gyrusRegression: reading !44 !52 22 3.87 Left supramarginal gyrus

Right TLE Regression: verb generation !42 2 36 3.47 Left middle frontal gyrus

Activation peaks for all fMRI effects of interest. For each effect, the Montreal Neurological Institute (MNI) coordinate, Z score and anatomical location aregiven. Z scores not marked as * or ** are significant at a threshold of P<0.001 (uncorrected).* P<0.05 corrected for multiple comparisons.** Sub-threshold activation only.

214 H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

For the left TLE group, extensive and consistent connectionswere seen bilaterally to the superior and middle temporal gyri,extending anteriorly into the temporal lobe. More extensive right-sided connections were seen to the extra-striate visual cortex(Brodmann area 19). Connections to the inferior frontal lobe viathe SLF were demonstrated on the left but not on the right.Overall there was no significant difference in either volume ofconnections (left 41.8 cm3, right 40.2 cm3; P=0.77) or mean FA(left 0.290, right 0.297; P=0.48) between left and right hemi-spheres.

For the right TLE group most connections were relativelysymmetrical, although greater connections along the SLF weredemonstrated on the left. The volume of connections was greater onthe left than the right (left 57.4 cm3, right 43.6 cm3; P=0.006) but

there was no significant difference in mean FA (left 0.305, right0.294; P=0.12).

Comparison with controls

Left TLEIn comparison with controls, left TLE patients demonstrated

less left-sided and greater right-sided connections, most strikinglyfrom the frontal lobe VOIs (Fig. 2). The overall connecting volumeof the left frontal tracts was significantly less than controls(P=0.006) but there was no significant difference in the rightfrontal connections. There was a trend for the patients’ mean FA tobe lower than controls on the left and greater than controls on theright although neither of these differences was significant.

Fig. 2. Group variability maps (at a probability threshold of 0.05) of the connecting paths tracked from the left and right frontal VOIs for each of the 3 groups.Each image shows the maximum intensity of the commonality maps in each plane of view as a brain surface rendering. Radiological convention is applied. Thecolour scale indicates the degree of overlap among subjects (expressed as commonality value); for example, a value of 1 (pure red) represents 100% subjectoverlap (i.e. every subject's identified tract contains the voxel in question). Controls and right TLE patients show a similar pattern of connections with greaterSLF connections to the temporal lobe on the left (arrowed) than the right. In the left TLE group the opposite pattern is seen with greater temporal lobe connectionson the right (dotted arrow).

215H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

Significant two-way interactions between group and hemispherewere demonstrated for both volume of connections (F[1,6]=7.76,P=0.032; Fig. 4A) and mean FA (F[1,6] =14.6, P=0.009; Fig. 4B)with left TLE patients being lower on the left and greater on theright than controls.

The overall connecting volume of the right temporal tracts wassignificantly greater in patients than controls (P=0.031) but therewas no significant difference in left temporal connections (Fig. 4C).Mean FAwas less in patients than controls on the left (P=0.036) butthere was no significant difference on the right (Fig. 4D).

Right TLERight TLE patients demonstrated a similar pattern of results to

controls with greater connections in the left hemisphere. Nosignificant differences from controls were observed in either frontal

volumes of connections (Fig. 4A) or mean FA (Fig. 4B). Fortemporal connections the overall connecting volume of both left(P=0.002; Fig. 4C) and right (P=0.033; Fig. 4C) temporal tractswas significantly greater in patients than in controls. No significantdifferences in mean FA were observed between patients andcontrols (Fig. 4D).

Comparison of Left and Right TLE groups

For the frontal lobe connections right TLE patients demon-strated a significantly lower mean FA on the right than left TLEpatients (P=0.042). There were significant two-way interactionsbetween group and hemisphere for both volume of connections(F[1,6]=16.23, P=0.007; Fig. 4A) and mean FA (F[1,6] =21.79,P=0.003; Fig. 4B) with left TLE patients being lower on the left

Fig. 3. Group variability maps of the connecting paths tracked from the left and right temporal VOIs for each of the 3 groups. Controls and right TLE patientsshow a similar pattern of connections with more substantial SLF connections to the frontal lobe on the left (arrowed) than the right, and relatively symmetricaloccipital lobe connections. In the left TLE group more extensive right-sided connections are seen to the inferior frontal lobe via the inferior fronto-occipitalfasciculus and the uncinate fasciculus (dotted arrow), and to the extra-striate visual cortex (thick arrow). Consistent bilateral connections were seen to the superiorand middle temporal gyri, extending anteriorly into the temporal lobe, in all three groups.

216 H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

and greater on the right. The overall connecting volume of the lefttemporal tracts was significantly less in left than right TLE patients(P=0.018) but there was no significant difference for the righttemporal connections (Fig. 4C). No significant differences wereobserved between the two patient groups in mean FA (Fig. 4D).

Correlations between structure and function

In the left TLE patients, a significant correlationwas seen betweenlateralisation of mean FA of the frontal connections, and functionallateralisation for verb generation in the frontal lobes (Spearmancoefficient=0.857, P=0.014; Fig. 5A). In the right TLE patients, asignificant correlation was seen between lateralisation of mean FA ofthe temporal connections, and functional lateralisation for reading inthe temporal lobes (Spearman coefficient=0.786, P=0.036; Fig.5B). These were characterised by greater structural lateralisation insubjects with greater functional lateralisation. Looking at correlationsbetween duration of epilepsy and tract lateralisation, there was a trendfor longer duration of epilepsy to be associated with less lateralisedtracts (Spearman coefficient=!0.739; P=0.05).

Regression analysis

For the left TLE patients we found a correlation between themean FA of the left frontal connections and fMRI activity for verbgeneration in the left IFG (Fig. 6A, Table 2). For readingcomprehension activation within the left supramarginal gyrus was

significantly correlated with mean FA of the left temporalconnections (Fig. 6B, Table 2). For the right TLE patients wefound a correlation between the mean FA of the left frontalconnections and fMRI activity for verb generation in the leftmiddle temporal gyrus (Fig. 6C, Table 2). No correlation was seenfor the temporal connections.

Discussion

We found evidence for structural reorganisation of white mattertracts underlying altered language lateralisation in patients withunilateral TLE. In patients with left TLE there were reduced lefthemisphere language connections compared to both controls andright TLE patients. This was evident both on visual inspection ofthe pathways delineated, and quantitatively by comparing thevolume and mean FA of the tracts. In contrast, in the righthemisphere there were greater connections in left TLE patients thanboth controls and right TLE patients. This suggests a reduction instructural connections in the pathological left hemisphere and apossible compensatory increase in the unaffected right hemisphere.

For the right TLE patients the overall pattern of connectivitywas similar to controls with greater fronto-temporal connections inthe left hemisphere than the right. Quantitatively the onlysignificant difference from controls was greater mean FA of theleft-sided temporal connections. This suggests a degree of reorga-nisation away from the pathological right hemisphere, althoughgiven that the non-language-dominant hemisphere is affected, the

Fig. 4. (A) Comparison of normalised tract volumes (NV) (±SE) between controls, left and right TLE patients for frontal lobe connections. (B) Comparison ofmean FA (±SE) between controls, left and right TLE patients for frontal lobe connections. (C) Comparison of NV between controls, left and right TLE patientsfor temporal lobe connections. (D) Comparison of mean FA between controls, left and right TLE patients for temporal lobe connections. For the frontal lobeconnections, significant two-way interactions between group and hemisphere are seen for both NVand FA characterised by left TLE patients having lower valueson the left and greater values on the right, compared with both controls and right TLE patients.

217H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

effect was less marked than in left TLE patients. In summary leftTLE had less left-sided and greater right-sided connections thancontrols and right TLE patients had greater left-sided connectionsthan controls.

We also demonstrated correlations between the structurallateralisation of the identified pathways and the left–rightdifference in functional activation in both left and right TLEpatients, with subjects with more highly lateralised languagefunction having a more lateralised pattern of connections. Thisrelationship between brain structure and function was previouslydemonstrated in controls (Powell et al., 2006). It is interesting thatit is maintained in the presence of pathology, demonstrating howstructural reorganisation underpins functional plasticity in cerebralnetworks, and may be important when planning ATLR. Finally weused a regression analysis to identify regions where voxel-levelfMRI activity correlated with mean FA. These were located indominant hemisphere language-related regions in both left andright TLE groups, similar to those seen in controls (Powell et al.,2006).

Our findings relate to a relatively small sample of patients andwill require confirmation in larger groups. The sample size doesnot allow us to investigate fully the influence of other potentiallyimportant factors, for example age of onset, and duration andseverity of epilepsy on our findings. Nevertheless there was a trendfor longer duration of epilepsy to be associated with less lateralisedtracts, suggesting that early age of onset is associated with greaterplasticity. In addition the small sample did not allow us to assess

the effect of handedness on structural lateralisation. One of the leftTLE patients was left-handed but clearly left-hemisphere-dominantfor language. Excluding this subject from the group analyses hadno effect on the differences between groups or on the correlationbetween structure and function.

We did not carry out performance monitoring or behaviouraltesting outside the scanner to assess the patients’ ability to performthe task adequately. There were, however, no significantdifferences in neuropsychological measures of naming and fluency,or verbal IQ between left and right TLE patients, and previous useof the reading task found no difference between TLE patients andcontrols in their ability to perform the task (Noppeney et al., 2005).Nevertheless there was a trend for left TLE patients to have lowerverbal IQ and language scores, and we cannot exclude differentialperformance as a contributory factor to the differences observed infMRI activation between groups.

The pattern of reorganisation reflects the fMRI results.Functional activation for the expressive language task was moreleft lateralised in controls and right TLE than left TLE patients inwhom overall activation was more bilateral. A number of studieshave shown that left TLE patients have a more symmetrical, lesslateralised pattern of language activation than controls and rightTLE patients (Adcock et al., 2003; Janszky et al., 2003; Thivardet al., 2005; Voets et al., 2006). One recent study has demonstrateda higher degree of atypical language dominance, in both frontaland temporal language areas, in patients with left HS compared topatients with left frontal and lateral temporal lesions (Weber et al.,2006), suggesting that the hippocampus itself may play animportant role in the establishment of language dominance.Fig. 5. Significant correlations between mean FA lateralisation index

(LI=FAleft!FAright /FAleft+FAright) of the pathways identified and func-tional laterality of fMRI activation in the frontal lobes for verb generation inleft TLE patients (A), and in the temporal lobe for reading comprehension inright TLE patients (B).

Fig. 6. Regression analysis between fMRI contrasts and tract mean FA. (A)Correlation between fMRI activation for verb generation and mean FA of theleft frontal connections, left TLE patients. (B) Correlation between fMRIactivation for reading and mean FA of the left temporal connections, leftTLE patients. (C) Correlation between fMRI activation for verb generationand mean FA of the left frontal connections, right TLE patients.

218 H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

A few studies have used tractography to look at the whitematter pathways underlying language function in control subjects,demonstrating a left lateralised pattern of connections, reflectingthe functional dominance of the left hemisphere (Nucifora et al.,2005; Parker et al., 2005; Powell et al., 2006). Our study reportsthese connections in a patient population and differs from work ofother groups in that we use fMRI to define functionally activeregions to use as start points for fibre tracking. This has theadvantage of reducing the observer bias inherent in manual VOIdefinition and allows the study of the neuronal networks con-necting cortical regions subserving particular functions.

A study by Catani et al. used tractography to study lefthemispheric perisylvian language networks (Catani et al., 2005). Inaddition to the direct pathway connecting Broca and Wernicke’sareas, they used a two-VOI approach to demonstrate a second,indirect pathway passing through the inferior parietal cortex. Thisran laterally to the direct pathway and was composed of an anteriorsegment connecting Broca’s area with the inferior parietal lobe anda posterior segment connecting the inferior parietal lobe toWernicke’s area. The authors argued that the existence of thissecond pathway helped to explain the diverse clinical spectrum ofaphasic disconnection syndromes. Our findings are in keeping withthese as we demonstrate a connection to the supramarginal gyrus(Brodmann area 40) in the inferior parietal lobe (Powell et al.,2006), a region implicated in a number of language-related tasks(Hickok and Poeppel, 2000; Wise et al., 2001). The single VOIapproach does not allow us to distinguish whether this is a separateand discrete pathway to the other fronto-temporal connectionsdemonstrated.

We used a single starting region along with a probabilistictractography technique. Tracking from single VOIs for eachfunctional localisation in each individual allowed a globalassessment of the pattern of connectivity without imposing strongprior user knowledge, and provided a measure of confidence to theconnections seen. It may be the case that some pathways whichplay a role in certain aspects of language processing do not directlyconnect Broca and Wernicke’s areas (for example the connectionsbetween Wernicke’s area and visual and auditory cortex) andtherefore would not be seen when the results are constrained byusing two VOIs. Two region approaches also have the disadvan-tage of potential bias due to the a priori assumption that con-nections between the two sites do actually exist. As a result, clearprominence is given to apparent connections between these sites,thus ignoring other potentially interesting connections. Theadvantages of using a two region approach, along with streamlinetractography techniques, are that they reduce the likelihood of falsepositive pathways and allow separate pathways to be studiedindividually or compared between different groups.

In order to minimise operator bias and to ensure consistency inour method, we used the group activation peak for defining startregions for fibre tracking, rather than each individual subject’s ownactivation peak. We realise that this may reduce our sensitivity inidentifying subtle differences in connection patterns betweenindividuals but concluded that it was the most robust and reliablemethod for detecting group level differences between patients andcontrols. Difficulties in co-registering fMRI and DTI can lead toproblems in the use of fMRI to define starting points fortractography. The steps of normalisation to standard space (toobtain the group activation maps) and subsequent reverse normal-isation to native space, along with the differences in susceptibilityand other artefacts between fMRI and DTI images are potential

sources of error when co-registering the two modalities. We tried tolimit the effect of small registration errors by defining relativelylarge VOIs (each consisted of 125 voxels). Spatial smoothing ofthe fMRI scans leads to blurring of activations across neighbouringvoxels, leading to activations which include both grey and whitematter. This provided a relatively unbiased choice of white mattervoxels for tractography seeding, avoiding the necessity to manuallydefine the white matter voxels expected to subserve a particulargrey matter area.

We studied a homogenous group of patients, all of whom hadHS. One previous tractography study in patients with TLE andunilateral HS, showed bilateral symmetrical abnormalities and FAreduction of the fornix and cingulum, suggesting that HS isassociated with bilateral limbic system pathology not revealed onconventional neuroimaging (Concha et al., 2005). Given the directconnections between the hippocampus and these structures, theauthors speculated that this may reflect downstream axonalWallerian degeneration (Werring et al., 2000). A more likelyexplanation for our findings is that structural reorganisation occursalong with functional plasticity secondary to a pathological processaffecting the temporal lobe early in life. In patients with left HS thisfunctional reorganisation could lead to reduced left hemisphereactivation and an increased role for the right hemisphere inlanguage function which in turn could affect the neuronalarchitecture underlying language functions. Alternatively morewidespread structural abnormalities may occur following the initialpathological insult that causes HS, which subsequently serve as astructural substrate upon which the abnormal functional activationdevelops. The effect is similar to that observed in people with earlyblindness, in whom tractography of the optic radiations hasrevealed reduced thalamocortical connections compared to normalcontrols (Shimony et al., 2006), and suggests that the status of thestructural components of cerebral networks may be critical inunderstanding the overall changes in function.

We feel that it is unlikely that patients’ lesions will have adirect affect on the tracking results as we have selected ahomogenous patient sample with the same pathological lesion ofhippocampal sclerosis. In particular, bias as a result of falsepositive pathways is unlikely to be a major problem in the patientgroup. If anything we would expect to get more false negatives iftracking gets dispersed in areas of unusually low anisotropy,indicating tract damage.

The lateralisation of language function inevitably leads toproblems in the identification of the right hemisphere homologuesto Broca’s and Wernicke’s areas. Our solution in the frontal lobeswas to manually define right hemisphere VOIs of identical size inareas homotopic to the functionally defined regions. The aim wasto minimise operator bias, although we recognise the limitationsof this approach given that structurally homotopic regions do notnecessarily correspond functionally, and the location of functionalregions may differ between groups. One recent study has indeeddemonstrated that right frontal activation on tasks of verbalfluency was not homologous to that seen in the left frontal lobeand that in a group of patients with left TLE the right frontalactivation shifted in location (Voets et al., 2006). By performingdirect contrasts of activation between each patient group in turnwith controls we demonstrated that there were no significantdifferences of activation within the chosen VOIs, justifying theuse of a single right-sided VOI for all three groups. For thesuperior temporal gyrus, however, we had areas of fMRIactivation from the reading comprehension paradigm in both left

219H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

and right hemispheres which we used for defining left and right-sided VOIs.

Both the quantitative measures used, tract volume and meanFA, reflect the underlying structural organisation of the pathwaysin question. FA is a measure of intravoxel fibre coherence and hasbeen shown to be reduced in a number of diseases including MS(Rovaris et al., 2005) and epilepsy (Eriksson et al., 2001),reflecting axonal degeneration (Concha et al., 2005) or loss ofmyelination. The lower values of FA seen in the non-dominanthemisphere of healthy subjects, and in the patients’ pathologicalhemisphere, are consistent with a reduction in the coherence ofaxons within the white matter pathways, indicating a reduction instructural organisation relative to the dominant and undamagedhemispheres. Tract volume changes are more difficult to interpret,particularly when using probabilistic tractography. One theory isthat tract volume changes are a reflection of the uncertainty of fibreorientation, with increasing uncertainty leading to greater spreadingof tracts. If this was the case, however, we would expect to seereduced FA and increased tract volume in the patients. We see theopposite, however, with reduced FA and smaller volumes suggest-ing that our findings reflect a reduction in real tract volume, due toatrophy.

Selective language deficits have been reported followinglanguage-dominant ATLR in patients with medically refractoryTLE. In one study, nearly 40% of patients demonstrated asignificant postoperative decline on a standard neuropsychologicalnaming test (Davies et al., 1998). Our results suggest thattractography, in combination with fMRI, may be able to predictand minimise such post-operative language deficits. Tractographyalso has broader applications to delineate the structural basis ofcerebral connectivity and how it is affected by disease.

Acknowledgments

This work was supported by the Wellcome Trust (ProgrammeGrant No.067176, HWRP, MRS), the National Society forEpilepsy (MJK, JD) and Action Medical Research (PB).

References

Adcock, J.E., Wise, R.G., Oxbury, J.M., Oxbury, S.M., Matthews, P.M.,2003. Quantitative fMRI assessment of the differences in lateralizationof language-related brain activation in patients with temporal lobeepilepsy. NeuroImage 18, 423–438.

Alexander, D.C., Barker, G.J., Arridge, S.R., 2002. Detection and modelingof non-Gaussian apparent diffusion coefficient profiles in human braindata. Magn. Reson. Med. 48, 331–340.

Boulby, P.A., Symms, M., Barker, G.J., 2004. A simple method for matchingdistortions in functional and structural data. ISMRM 12th ScientificMeeting. Proc. Int. Soc. Magn. Reson., p. 2196.

Catani, M., Jones, D.K., Ffytche, D.H., 2005. Perisylvian language networksof the human brain. Ann. Neurol. 57, 8–16.

Concha, L., Beaulieu, C., Gross, D.W., 2005. Bilateral limbic diffusionabnormalities in unilateral temporal lobe epilepsy. Ann. Neurol. 57,188–196.

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, 10422–10427.

Davies, K.G., Bell, B.D., Bush, A.J., Hermann, B.P., Dohan Jr., F.C., Jaap,A.S., 1998. Naming decline after left anterior temporal lobectomy

correlates with pathological status of resected hippocampus. Epilepsia39, 407–419.

Eriksson, S.H., Rugg-Gunn, F.J., Symms, M.R., Barker, G.J., Duncan, J.S.,2001. Diffusion tensor imaging in patients with epilepsy and malforma-tions of cortical development. Brain 124, 617–626.

Friston, K.J., Holmes, A.P., Worsley, K.J., Poline, J.B., Frith, C.D.,Frackowiak, R.S., 1995. Statistical parametric maps in functionalimaging: a general linear approach. Hum. Brain Mapp. 2, 189–210.

Gaillard, W.D., 2004. Functional MR imaging of language, memory, andsensorimotor cortex. Neuroimaging Clin. N. Am. 14, 471–485.

Hammers, A., Koepp, M.J., Labbe, C., Brooks, D.J., Thom, M., Cunningham,V.J., Duncan, J.S., 2001. Neocortical abnormalities of [11C]-flumazenilPET in mesial temporal lobe epilepsy. Neurology 56, 897–906.

Hermann, B., Seidenberg, M., 1995. Executive system dysfunction intemporal lobe epilepsy: effects of nociferous cortex versus hippocampalpathology. J. Clin. Exp. Neuropsychol. 17, 809–819.

Hickok, G., Poeppel, D., 2000. Towards a functional neuroanatomy ofspeech perception. Trends Cogn. Sci. 4, 131–138.

Janszky, J., Jokeit, H., Heinemann, D., Schulz, R., Woermann, F.G., Ebner,A., 2003. Epileptic activity influences the speech organization in medialtemporal lobe epilepsy. Brain 126, 2043–2051.

Keller, S.S., Mackay, C.E., Barrick, T.R., Wieshmann, U.C., Howard, M.A.,Roberts, N., 2002. Voxel-based morphometric comparison of hippo-campal and extrahippocampal abnormalities in patients with left andright hippocampal atrophy. NeuroImage 16, 23–31.

Lehericy, S., Cohen, L., Bazin, B., Samson, S., Giacomini, E., Rougetet, R.,Hertz-Pannier, L., Le Bihan, D., Marsault, C., Baulac, M., 2000.Functional MR evaluation of temporal and frontal language dominancecompared with the Wada test. Neurology 54, 1625–1633.

Liegeois, F., Connelly, A., Cross, J.H., Boyd, S.G., Gadian, D.G., Vargha-Khadem, F., Baldeweg, T., 2004. Language reorganization in childrenwith early-onset lesions of the left hemisphere: an fMRI study. Brain127, 1229–1236.

Moran, N.F., Lemieux, L., Kitchen, N.D., Fish, D.R., Shorvon, S.D., 2001.Extrahippocampal temporal lobe atrophy in temporal lobe epilepsy andmesial temporal sclerosis. Brain 124, 167–175.

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, 265–269.

Noppeney, U., Price, C.J., Duncan, J.S., Koepp, M.J., 2005. Reading skillsafter left anterior temporal lobe resection: an fMRI study. Brain 128,1377–1385.

Nucifora, P.G., Verma, R., Melhem, E.R., Gur, R.E., Gur, R.C., 2005.Leftward asymmetry in relative fiber density of the arcuate fasciculus.NeuroReport 16, 791–794.

Parker, G.J., Alexander, D.C., 2003. Probabilistic Monte Carlo basedmapping of cerebral connections utilising whole-brain crossing fibreinformation. Lect. Notes Comput. Sci. 2737, 684–695.

Parker, G.J., Haroon, H.A., Wheeler-Kingshott, C.A., 2003. A frameworkfor a streamline-based probabilistic index of connectivity (PICo) using astructural interpretation of MRI diffusion measurements. J. Magn.Reson. Imaging 18, 242–254.

Parker, G.J., Luzzi, S., Alexander, D.C., Wheeler-Kingshott, C.A.,Ciccarelli, O., Lambon Ralph, M.A., 2005. Lateralization of ventraland dorsal auditory-language pathways in the human brain. NeuroImage24, 656–666.

Pierpaoli, C., Basser, P.J., 1996. Toward a quantitative assessment ofdiffusion anisotropy. Magn. Reson. Med. 36, 893–906.

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

Powell, H.W., Guye, M., Parker, G.J., Symms, M.R., Boulby, P., Koepp,M.J., Barker, G.J., Duncan, J.S., 2004. Noninvasive in vivo demonstra-tion of the connections of the human parahippocampal gyrus. Neuro-Image 22, 740–747.

Powell, H.W., Parker, G.J., Alexander, D.C., Symms, M.R., Boulby, P.A.,Wheeler-Kingshott, C.A., Barker, G.J., Noppeney, U., Koepp, M.J.,

220 H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221

Duncan, J.S., 2006. Hemispheric asymmetries in language-relatedpathways: a combined functional MRI and tractography study. Neuro-Image 32, 388–399.

Rovaris, M., Gass, A., Bammer, R., Hickman, S.J., Ciccarelli, O., Miller,D.H., Filippi, M., 2005. Diffusion MRI in multiple sclerosis. Neurology65, 1526–1532.

Shimony, J.S., Burton, H., Epstein, A.A., McLaren, D.G., Sun, S.W.,Snyder, A.Z., 2006. Diffusion tensor imaging reveals white matterreorganization in early blind humans. Cereb. Cortex 16, 1653–1661.

Sisodiya, S.M., Moran, N., Free, S.L., Kitchen, N.D., Stevens, J.M., Harkness,W.F., Fish, D.R., Shorvon, S.D., 1997. Correlation of widespreadpreoperative magnetic resonance imaging changes with unsuccessfulsurgery for hippocampal sclerosis. Ann. Neurol. 41, 490–496.

Stefan, H., Feichtinger, M., Pauli, E., Schafer, I., Eberhardt, K.W., Kasper,B.S., Hopp, P., Buchfelder, M., Huk, J., Paulus, W., 2001. Magneticresonance spectroscopy and histopathological findings in temporal lobeepilepsy. Epilepsia 42, 41–46.

Talairach, J., Tournoux, P., 1988. Co-Planar Stereotaxic Atlas of the HumanBrain. Georg Thieme Verlag, Stuttgart.

Thivard, L., Hombrouck, J., du Montcel, S.T., Delmaire, C., Cohen, L.,Samson, S., Dupont, S., Chiras, J., Baulac, M., Lehericy, S., 2005.Productive and perceptive language reorganization in temporal lobeepilepsy. NeuroImage 24, 841–851.

Toosy, A.T., Ciccarelli, O., Parker, G.J., Wheeler-Kingshott, C.A., Miller,D.H., Thompson, A.J., 2004. Characterizing function-structure relation-ships in the human visual system with functional MRI and diffusiontensor imaging. NeuroImage 21, 1452–1463.

Tuch, D.S., Reese, T.G., Wiegell, M.R., Makris, N., Belliveau, J.W.,Wedeen, V.J., 2002. High angular resolution diffusion imaging reveals

intravoxel white matter fiber heterogeneity. Magn. Reson. Med. 48,577–582.

Voets, N.L., Adcock, J.E., Flitney, D.E., Behrens, T.E., Hart, Y., Stacey, R.,Carpenter, K., Matthews, P.M., 2006. Distinct right frontal lobe activationin language processing following left hemisphere injury. Brain 129,754–766.

Weber, B., Wellmer, J., Reuber, M., Mormann, F., Weis, S., Urbach, H.,Ruhlmann, J., Elger, C.E., Fernandez, G., 2006. Left hippocampalpathology is associated with atypical language lateralization in patientswith focal epilepsy. Brain 129, 346–351.

Werring, D.J., Toosy, A.T., Clark, C.A., Parker, G.J., Barker, G.J., Miller,D.H., Thompson, A.J., 2000. Diffusion tensor imaging can detectand quantify corticospinal tract degeneration after stroke. J. Neurol.,Neurosurg. Psychiatry 69, 269–272.

Wheeler-Kingshott, C.A., Hickman, S.J., Parker, G.J., Ciccarelli, O.,Symms, M.R., Miller, D.H., Barker, G.J., 2002. Investigating cervicalspinal cord structure using axial diffusion tensor imaging. NeuroImage16, 93–102.

Wise, R.J., Scott, S.K., Blank, S.C., Mummery, C.J., Murphy, K.,Warburton, E.A., 2001. Separate neural subsystems within ‘Wernicke’sarea’. Brain 124, 83–95.

Woermann, F.G., Barker, G.J., Birnie, K.D., Meencke, H.J., Duncan, J.S.,1998. Regional changes in hippocampal T2 relaxation and volume: aquantitative magnetic resonance imaging study of hippocampalsclerosis. J. Neurol., Neurosurg. Psychiatry 65, 656–664.

Woermann, F.G., Jokeit, H., Luerding, R., Freitag, H., Schulz, R., Guertler,S., Okujava, M., Wolf, P., Tuxhorn, I., Ebner, A., 2003. Languagelateralization by Wada test and fMRI in 100 patients with epilepsy.Neurology 61, 699–701.

221H.W.R. Powell et al. / NeuroImage 36 (2007) 209–221