review language activation distributions revealed by fmri in post ... · reference and to determine...

Epilepsy Research 66 (2005) 1–12

Review

Language activation distributions revealed by fMRI inpost-operative epilepsy patients: Differences

between left- and right-sided resections

W.H. Backesa,∗, K. Deblaereg, K. Vonckg, A.G. Kesselsb, P. Boonc,P. Hofmana, J.T. Wilminka, G. Vingerhoetsh, P.A. Boong,f ,

R. Achteng, J. Vermeulene, A.P. Aldenkampd,f

a Department of Radiology, Maastricht University Hospital, P. Debyelaan 25, NL-6229 HX Maastricht, The Netherlandsb Department of Epidemiology and Medical Technological Assessment, Maastricht University Hospital,

P. Debyelaan 25, NL-6229 HX Maastricht, The Netherlandsc Department of Clinical Psychology, Maastricht University Hospital, P. Debyelaan 25, NL-6229 HX Maastricht, The Netherlands

d Department of Neurology, Maastricht University Hospital, P. Debyelaan 25, NL-6229 HX Maastricht, The Netherlandse S.E.I.N., Heemstede, The Netherlands

f Epilepsy Center Kempenhaeghe, Heeze, The Netherlandsg

y

cidxt

arice

st-t

Department of Radiology, Ghent University Hospital and Ghent University, Ghent, Belgiumh Department of Neuropsychology, Ghent University Hospital and Ghent University, Ghent, Belgium

Received 1 June 2005; received in revised form 12 June 2005; accepted 22 June 2005Available online 22 August 2005

Abstract

Objective: To reveal differences of cerebral activation related to language functions in post-operative temporal lobe epileps(TLE) patients.Methods: Right (RTL) and left temporal lobe (LTL) resected patients, and healthy controls were studied using functional magnetiresonance imaging (fMRI). Only patients with complete left-hemispheric language dominance according to the intracarotamytal procedure (IAP) were included. Language-related activations were evoked by performing word generation and tereading language tasks. Activation lateralization and temporo-frontal distribution effects were analysed.Results: For word generation, only LTL patients showed reduced left lateralized activation compared to controls, due todecrease in activation in the left prefrontal cortex and an increase in the right prefrontal cortex. For reading, the left-hemisphelateralization in RTL patients increased because of enhanced activity in the left prefrontal cortex, whereas for LTL patients thactivation became bilaterally distributed over the temporal lobes. Lateralization results between pre-operative IAP and pooperative fMRI were highly discordant. Significant temporo-frontal distribution changes manifested from the reading but nofrom the word generation task.

∗ Corresponding author. Tel.: +31 43 3876948; fax: +31 43 3876909.E-mail address: [email protected] (W.H. Backes).

0920-1211/$ – see front matter © 2005 Elsevier B.V. All rights reserved.doi:10.1016/j.eplepsyres.2005.06.007

2 W.H. Backes et al. / Epilepsy Research 66 (2005) 1–12

Conclusion: The cerebral language representation in post-operative LTL epilepsy patients is more bi-hemispherically lat-eralized than in controls and RTL patients. Post-operative temporo-frontal and interhemispheric redistribution effects,involving contralateral homologuous brain areas, are suggested to contribute to the cerebral reorganisation of languagefunction.© 2005 Elsevier B.V. All rights reserved.

Keywords: Epilepsy; Anterior temporal lobectomy; fMRI; Language

Contents

1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22. Materials and methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1. Subjects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32.2. Language paradigms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.3. MR protocol. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.4. Data analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.4.1. Pre-processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.4.2. Activation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.4.3. Lateralization index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.4.4. Regions of interest. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.4.5. Distribution differences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3. Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63.1. Word generation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63.2. Reading. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81111

;

Acknowledgment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Introduction post-operatively (Rausch et al., 2003; Lee et al., 2002Bell and Davies, 1998). When a hippocampal scle-

Anterior temporal lobectomy (ATL) is a well-established surgical treatment for pharmaco-resistanttiopecrmnlaAfif

rosis is diagnosed less post-operative difficulties arefound, presumably caused by reorganization of func-

sueuitsrain

forerican-0%)

undn ined byy lefts that

emporal lobe epilepsy. The surgical procedure resultsn either a significant reduction in seizure frequencyr complete seizure remission in about 70% of theatients. The resection involves parts of the unilat-ral mesial temporal cortex as well as lateral neo-ortical structures of the anterior temporal lobe. Theesection may disturb or damage cortex eloquent foremory and language functioning. Despite extensiveon-invasive and invasive pre-operative hemispheric

anguage and memory testing, subtle problems mayrise in post-operative language functioning after leftTL (Hermann et al., 1994; Davies et al., 1998; Lang-tt and Rausch, 1996; Clusmann et al., 2002). Morerequently difficulties in verbal memory are found

tion during the transformation of healthy brain tisinto a sclerosis. Consequently, affected neural circmay already pre-operatively reorganize to distant bregions.

Typically, the left hemisphere is dominantlanguage function, while complete right-hemisphdominance is rare (about 1–2%) (8). Bilateral lguage representation is found far less (about 2than left-hemispheric dominance (Loring et al., 1990).Remarkably, bilateral language representation is foto be more prevalent in patients with epilepsy thahealthy people. This, again, is assumed to be causneuronal reorganization as a consequence of earlhemisphere epilepsies and lesions. Evidence exist

W.H. Backes et al. / Epilepsy Research 66 (2005) 1–12 3

functional reorganization enables the right hemisphereto mediate language functions in response to left brainpathology early in life (Helmstaedter et al., 1997).

Functional MRI is an imaging technique that pro-vides unique and highly reproducible (Rutten et al.,2002a,b) means to localize the brain systems involvedin processing language functions. For diagnostic lan-guage and memory evaluation in patients with epilepsy,a number of research centres (Baxendale, 2002) hasrecently correlated the outcomes of fMRI and the IAP(i.e. the Wada test), which is considered the gold stan-dard for language lateralization. Until now, extremelygood correlations have been published between fMRIand the IAP, especially for epilepsy patients with a typi-cal (i.e. left dominant) language representation (Binderet al., 1996; Yetkin et al., 1998; Lehericy et al., 2000;Rutten et al., 2002a,b; Spreer et al., 2002; Adcocket al., 2003; Woermann et al., 2003).

The number of imaging studies dealing with post-lesional plasticity of the brain is, however, small andlimited to case studies.Hertz-Pannier et al. (2002)recently reported in a serial fMRI study an interhemi-spheric transfer of brain activation in response to lan-guage demands in a 9-year-old child after left hemi-spherotomy. The authors explained the plasticity effectby the assumption of a pre-existing bilateral languagen oni cti-v ithl ,2

lan-g with

epilepsy after ATL. The hypothesis is that when nocerebral language reorganizations occur the activationpatterns should be independent of the resection side. Totest this an fMRI study was set out that compared thebrain activation patterns between patient groups whohad a left- and right-sided resection against controls.By using word generation and reading as neuropsy-chological language tasks brain systems were studiedthat respond to expressive and receptive language func-tioning, respectively.

2. Materials and methods

2.1. Subjects

Characteristics of the patients and the healthy con-trols are summarized inTable 1. Nine healthy controls,12 RTL and 9 LTL resected epilepsy patients wereincluded in the study protocol. All patients and controlswere right-handed. The patients underwent a presur-gical language and memory evaluation with the IAP(Wada and Rasmussen, 1960). During the IAP amobar-bital, a short acting anaesthetic, is injected in one of theinternal carotid arteries producing a transient anaesthe-sia of the territories of the middle and anterior cerebrala s aret . Thel am-i ntsw ncew enta ith

TC

C

MAAATSP )P )RP

V be; RT t; VIQ:v pocam

etwork that is inhibited by preferential left activatin the normal healthy situation. Shifts in brain aation patterns are also found in TLE patients wesions at or near the language areas (Adcock et al.003; Springer et al., 1999).

This study aimed at detecting reorganizations inuage activation patterns in a group of patients

able 1haracteristics of epilepsy patients and controls

haracteristic Controls

ale/female 3/9ge at onset epilepsy (year)ge at surgery (year)ge at imaging (year) 26± 6ime between surgery and fMRI (year)eizure frequency (month−1)resurgical PIQresurgical VIQesection volume (cm3)athology

alues are mean± 1S.D., median (range). LTL: left temporal loerbal intelligence quotient; HS: hippocampal sclerosis; HT hip

rteries of the ipsilateral hemisphere. Several testhen performed to evaluate language lateralizationanguage part of this test comprised overt object nng, picture description, and story recall. Only patieith complete left-hemispheric language dominaere included in this study. All patients underwpartial resection of the anterior temporal lobe w

RTL LTL

5/7 3/612.4± 7.6, 12 (4–31) 10.7± 6.8, 8.5 (3–22)

30± 8, 31 (18–43) 34± 9, 33 (23–48)34± 7, 35 (23–46) 37± 8, 36.5 (26–52)

4.5± 2.0, 5 (1–8) 3.8± 3.4, 3.5 (1–11)3.5± 1.2, 4 (1–4) 4.2± 2.4, 4 (1–4)

112± 15, 116 (88–134) 109± 13, 110 (82–126105± 15, 108 (82–133) 103± 17, 111 (71–122

25.5± 10.1, 27 (8.9–42) 16.9± 5.5, 17 (9.9–24)7 HS, 1 HT, 4 OP 7 HS, 1 HT, 1 OP

L: right temporal lobe; PIQ: performance intelligence quotienpal tumor; OP: other pathology (e.g. gliosis, cyst or atrophy).


varying margins, including the amygdalo-hippocampalcomplex, and resection of lesions located in the neocor-tex when present. Postoperatively, no patients showedclinical decline in cognitive language and memoryfunctioning. Resection volumes were derived from MRimages. The right temporal lobe resection volume wason average 50% larger (p = 0.01) than in the left tem-poral lobe due to the lack of clear restrictions to spareeloquent cortex. The medical ethical committee of theMaastricht University Hospital approved the proto-col of the project, to which all subjects gave writteninformed consent.

2.2. Language paradigms

To test two language tasks were used: covert wordgeneration and covert text reading. During word gen-eration, the subjects had to covertly generate as manywords as possible beginning with a presented letter (U,N, K, A, E, and P). The word generation task was con-ducted in a block-design time course in which one letterwas presented per epoch and the baseline task com-prised the viewing of a fixation cross. The reading taskwas performed in the same sequence in which the read-ing of a meaningful text was contrasted with reading oftext containing pronounceable nonsense words. Eachl mes.V ns-p temu pre-v .,2 -s ssibleg text.A ntlyw

2

teme RI as es rum.I ngle9t vol-u icalr e, a

3D T1-weighted fast-field echo pulse sequence wasconducted, with parameters TR 11 ms, TE 3.5 ms, flipangle 11◦, matrix 256× 256× 150, and voxel size1 mm× 1 mm× 1 mm. The images of the controls havebeen acquired in similar way as described elsewhere(Deblaere et al., 2002), but were analysed differentlyto be consistent with the patient data.

2.4. Data analysis

2.4.1. Pre-processingImage processing and statistical analysis were per-

formed using the software package SPM99 of theWellcome Department of Cognitive Neurology (Lon-don, UK). Data pre-processing amounted to 3D motioncorrection, spatial normalisation to Montreal Neuro-logical Institute (MNI) T2* image template, and spatialsmoothing with an 8 mm Gaussian kernel. The anatom-ical images were also spatially co-registered to thecorresponding MNI T1 image template. The imagesof the resected patients were spatially normalised tothe template images by using affine transformations(i.e. rigid body, global scaling and skewing) only. Useof additional local deformations would erroneouslystretch the cortex near the resected zone. Using affinetransformations only, in the contrary to local deforma-t e inr ere2

2were

o allyfi theM nif-i inedf ualv thep elsa taln ents(

2the

r t ins ighth sing

anguage block took 32 s and was conducted six tiisual stimuli were projected with a beamer on a traarency screen and synchronised with the MRI syssing custom-made software. The paradigm wasiously tested in healthy volunteers (Deblaere et al002; Aldenkamp et al., 2003). After the fMRI sesions subjects were asked to repeat as many as poenerated words and to tell the contents of thell subjects performed both language tasks sufficieell.

.3. MR protocol

Images were acquired on a 1.5 T MRI sysquipped with a standard receiver head coil. For fMingle-shot multi-slice echo-planar T2* -weighted pulsequence was applied covering the entire cerebmaging parameters were TR 4 s, TE 50 ms, flip a0◦, matrix 64× 64, pixel size 3.5 mm× 3.5 mm, slice

hickness 3.5 mm, 34 contiguous axial slices perme, and 96 volumes per paradigm. For anatomeference and to determine the resection volum

ions, preserved the shape of the resected volumelation to the temporal lobe. Final voxel sizes wmm× 2 mm× 2 mm.

.4.2. ActivationThe amplitudes of the hemodynamic responses

btained for each voxel in each subject by numerictting a model hemodynamic response function toRI signal time courses. Maps representing the sig

cance of the brain activation amplitudes were obtaor each subject. Statistical inferences of individoxels, in the regions of interest, were subjected toroblem of multiple comparisons of many brain voxnd the resultingp-values were corrected for the toumber of spatially independent resolution elemresels) (Holmes and Friston, 1997).

.4.3. Lateralization indexThe lateralization index (LI) was defined as

elative difference between the activation extenpecified regions of interest in the left and the remisphere. LI was computed for each subject u


the formula LI = (VL − VR)/(VL + VR), whereVL andVR represent the extent (i.e. number of voxels) of brainactivation above the statistical threshold in the leftand right hemispheres, respectively. LI values rangebetween−1 and 1, which represents complete rightand left-hemispheric lateralization, respectively. Notethat by using the LI, one corrects (i.e. normalizes)for inter-individual variations in the amount of brainactivation. By displaying LI as a function of thethreshold one moreover is able to index the level ofactivation. The choice of the optimal threshold valuemay be limited by upper and lower boundary statisticaleffects. The threshold value should on the one handbe high enough to exclude a relatively large numberof voxels false-positively accepted for activation.On the other hand, the threshold value should notbe too high in order to reduce the number of voxelsfalse-negatively rejected for activation and to reducethe weight of the relatively high T2* responses (i.e.

high Z-values) of draining veins. To this end, forevery individual the LI was calculated for a range ofthreshold values to find deviations from predefinedthreshold value. For statistical comparison betweengroups the threshold was set toZ > 3.8 for the groupaveraged LI, which moreover corresponds top < 0.05when corrected for multiple (voxel) comparisons.

2.4.4. Regions of interestRather than using voxel-by-voxel comparisons (as

automated, e.g. in SPM) we decided to employ regionof interest analysis. The rationale for this choice isthat for many voxels no brain tissue is present inpostoperative patients. Performing a voxel-by-voxelanalysis would therefore not make much sense for vox-els in the temporal lobe as the resected brain regionsvary in size and shape between patients. Quantitativeanalysis, using digital brain masks (Maldjian et al.,2003) (Fig. 1), was restricted to regions of interest

Ffc

ig. 1. T1 weighted images of a volunteer with color-coded brain masrontal cortex (red), the anterior temporal lobe (yellow), the posterioolor-coding of the pertaining Brodmann areas (BA) is depicted.

k comprising the inferior and middle frontal cortex (orange), the superiorr temporal lobe (blue), and the temporo-parietal junction (green). The


that are known to be essential for language processing(Ojemann et al., 1989) and that have been previouslyidentified to be activated in controls for the same lan-guage tasks (Deblaere et al., 2002). These regions com-prise the inferior and middle frontal cortex (Brodmannareas, BA 44–47), the superior frontal cortex (BA 6, 9,and 10), the anterior temporal lobe (BA 38), the poste-rior and inferior temporal lobe (BA 20 and 21), and thetemporo-parietal junction (BA 39 and 40). LI values ofthe language mask were classified to represent bilateralactivation when−0.2 < LI < 0.2.

2.4.5. Distribution differences2.4.5.1. Lateralization. Relative changes of brainactivation lateralization were quantified by calcu-lating the differences of LI between RTL patientsand controls, and between LTL and controlsusing the expressions�LIRTL = LIRTL − LIN and�LILTL = LILTL − LIN, respectively. Here the lateral-ization index for healthy controls, RTL patients, andLTL patients is denoted as LIN, LIRTL, and LILTL ,respectively. Dominance of the left hemisphere, as isexpected for language tasks in controls, is character-ized by a positive LIN. A relative shift of activationfrom the epileptic to the non-epileptic hemisphere isapp

ro-f om-p tedv andf erea

3. Results

3.1. Word generation

The control group as well as the RTL and LTLgroups show a strong positive LI value for the definedlanguage regions (Table 2). Fig. 2a shows that theLTL group has a reduced LI value compared to thecontrol group for a broad range of threshold values.The asymmetry in the distribution between the hemi-spheres is mainly mediated by the strong left later-alized activation in the prefrontal cortex (example inFig. 4).

The largest drop (20%) of activation percentage forthe LTL patient group was in the left frontal cortex(near Broca’s region, BA 44–47) (Fig. 2b). The dis-tributed activation percentage increased (about 11%predominantly in the BA 6, 9, 10 and 20–22) in the righthemisphere in the LTL patient group. The number ofpatients, moreover, showing an atypical language later-alization was the highest in LTL patient group (exam-ple in Fig. 4). None of the correlations between LIand the demographic parameters gender, age at onset,age at imaging, seizure frequency, interval betweensurgery and imaging, and resection volume weres

w nft -hNn d(

TG orrecte nts,a , LI≥ 0

T RTL

W 53 (0.0:2:1

R 78 (0.01:1:0

V empora LIw

ssociated with significantly positive�LIRTL for RTLatients and a significantly negative�LILTL for LTLatients.

Temporo-frontal distribution. Possible temporontal distribution changes were investigated by caring the difference in the percentage of activaoxels between the temporal (BA 20–22, 39, 40)rontal lobe (BA 44–47, 6, 9, 10) for each hemisphnd each task.

able 2roup averaged lateralization indices (thresholdp < 0.05 orZ > 3.8, cnd the number of patients classified as right (R, LI≤ −0.2), left (L

ask Quantity LIN LI

ord generation LI 0.66 (0.09)** 0.L:B:R 8:1:0 9

eading LI 0.59 (0.12)** 0.L:B:R 8:1:0 1

alues are mean± 1 S.E.M. RTL: right temporal lobe; LTL: left tith respect to controls are listed.∗ Statistical significance level:p < 0.03.

∗∗ Statistical significance level:p � 0.001.

ignificant.A significantly positive LIN (p � 0.001) was found,

hich indicates clear left-hemispheric lateralizatioor controls. The�LILTL was significantly nega-ive, which implies a redistribution towards the rightemispheric activation in the LTL group (p < 0.03).o significant increases in LIRTL were found. Alsoo temporo-frontal distribution changes were founTable 3).

d) for controls, right and left temporal lobe resected epilepsy patie.2) or bilateral B (−0.2 < LI < 0.2) representation

LILTL �LIRTL �LILTL

9) 0.31 (0.17) −0.13 (0.16) −0.35 (0.17)*

5:2:2

8) 0.09 (0.27) 0.19 (0.21) −0.50 (0.23)*

3:3:3

l lobe; N: control; LI: lateralization index. Also the differences in


Fig. 2. The lateralization index (LI) as function of the statistical threshold valueZ for the controls (solid), the right temporal lobe (RTL) (dashed)and left temporal lobe (LTL) (dotted) patient group for the word generation task (a). The relative distribution of activated brain voxels (expressedin percentages) in each of the selected language-related regions (Brodmann areas) of the three groups (b). The error bars represent the standarderror of the means.

Table 3Group differences in percentage of activated voxels between thefrontal and temporal lobe (thresholdp < 0.05 orZ > 3.8, corrected)for controls, right and left temporal lobe resected epilepsy patients

Task Hemisphere Controls RTL LTL

Wordgeneration

L 43 ± 9 47 ± 12 30± 12

R 10 ± 3 13 ± 3 8 ± 5

Reading L −71 ± 12 −18 ± 14* −20 ± 7**

R −18 ± 7 4 ± 3** −32 ± 11

Values are mean± 1 S.E.M. RTL: right temporal lobe; LTL: lefttemporal lobe. Significant changes in temporo-frontal activation areindicated by one or more asterisks.

∗ Statistical significance level:p < 0.01.∗∗ Statistical significance level:p < 0.005.

3.2. Reading

The control group shows a positive LI for theselected language regions (Fig. 3a andTable 2). TheLTL group exhibit a LI (Z > 3.8) that is much lowerthan for the control and RTL group and approximatelyequal to zero for a broad range of thresholds, imply-ing bilateral language representation. The LI curve ofthe RTL group lies higher than for controls. The spa-tial distribution among classical language areas showsan asymmetric pattern, which is mainly mediated bythe strong left temporal lobe (BA 44–47 and 6, 9, 10)activation for the control and RTL group. The numberof patients showing an atypical language lateraliza-tion was the highest in LTL patient group (example inFig. 4). No significant relations were found between the


Fig. 3. The lateralization index as function of the statistical threshold valueZ for the controls (solid), the right temporal lobe (RTL) (dashed)and left temporal lobe (LTL) (dotted) patient group for the text reading task (a). The relative distribution of activated brain voxels (expressedin percentages) in each of the selected language-related regions (Brodmann areas) of the three groups (b). The error bars represent the standarderror of the means.

LI and the demographic parameters age at onset, age atimaging, seizure frequency, interval between surgeryand imaging, and resection volume.

For the LTL patient group the percentage of activa-tion in the left posterior temporal lobe dropped by 16%,while the right posterior temporal lobe increased by16% (Fig. 3b). For the RTL patient group the left pos-terior temporal lobe activation remained unchanged,while the right posterior lobe decreased (about 14%)and the left inferior frontal lobe increased (about 18%)(Fig. 3b).

Temporo-frontal distribution changes were signifi-cant in both hemispheres in the RTL group and in theleft hemisphere (only) of the LTL group (Table 3).

Again a significantly positive LIN (p � 0.001) wasfound, which indicates clear left-hemispheric lateral-ization for controls. The�LILTL was significantly neg-

ative, which implies a redistribution towards the right-hemispheric activation in the LTL group (p < 0.02). Nosignificant increases in the LIRTL were found.

4. Discussion

Language activation assessed by fMRI, using wordgeneration and reading tasks, was compared betweenpatients with a right- or left-sided resection for tempo-ral lobe epilepsy and with healthy controls. Althoughbrain activation in response to language processingappeared in both hemispheres, a clear prevalent role ofthe left hemisphere was found only in controls and RTLpatients. LTL patients, on the contrary, showed reducedleft-hemispheric lateralization for word generation andshowed bilateral activation for reading. Concordance


Fig. 4. The top row shows the (color-coded) activation map of a control subject for the word generation and the reading task superimposedon a T1 weighted image of one control. Clearly visible is the left lateralized prefrontal activation for word generation and the left lateralizaedtemporal activation for the reading task. Also the homologuous areas in the contralateral hemisphere are activated but to a lesser degree (a). Themiddle row show the activation pattern of a patient of the right temporal lobe (RTL) group who yielded left lateralized activation for both tasks.The arrow points at the resected brain part (b). The lower row represents a patient of the left temporal lobe (LTL) group who showed bilateralactivation for both tasks. The arrowhead points at the resected brain part (c). Note that the echo-planar T2* -weighted activation images had anoriginal resolution of 3.5 mm and were co-registrated to the normalised T1-weighted 3D image with a final spatial resolution of 2 mm. Possiblegeometric distortions on the echo-planar images due to field inhomogeneities may contribute to inaccurate co-registration.

with the pre-operative IAP language results is expect-edly high for RTL patients but is surprisingly low forLTL patients. These results strongly suggests that aninterhemispheric reorganisation of language activationhas occurred in LTL patients.

This is the first imaging study demonstrating hemi-spheric differences in cerebral language activation

between left- and right-resected patients. At first sightboth the left-hemispheric epileptic seizures and theleft temporal lobe resection might have contributed tothe reduced left-hemispheric language lateralization inLTL patients. The data of this study do not allow todisentangle these two effects. Focal epileptic seizuresunderlying lesions with an onset early in life and the


type of epileptic activity in the left hemisphere havealready been reported to be the cause that a high per-centage of LTL patients shows a bilateral language rep-resentation in the IAP (Helmstaedter et al., 1997; Risseet al., 1997; Janszky et al., 2003). This demonstratesthat a slowly occurring reorganisation process mayaffect language organisation in patients with chronicrefractory epilepsy. In the current study all patients,however, have been carefully selected to exhibit com-plete left-hemispheric language dominance accordingto the IAP that was assessed about 6 months beforesurgery. The possibility of (extensive) pre-operativelanguage reorganisation is therefore unlikely. More-over, only a limited number of patients (one LTL andone RTL) had a seizure onset before the age of 5, beingthe critical age to allow early language reorganisa-tion. Adcock et al. (2003)have recently demonstratedthat the LI values for preoperative LTL patients arelower than the RTL patients and controls. The preop-erative LTL patient group, however, included 4 out of12 patients with an atypical (one right and three bilat-eral) language lateralization according to the IAP in thecontrary to the RTL group which included only patientswith a left lateralized IAP result. Reanalysing the studyby Adcock et al. (2003)shows that there are no dif-ferences in terms of LI values between pre-operativeL ft-h IAPw uagel ree-m tionr ne ,b;S anne r-h udya en-t overe tingd MRIr n inp ni-t f thel te allp

u-t tiont ely

due to the fact that the activation for this expressivelanguage task is mainly in the (left) prefrontal lobe. Sig-nificant temporo-frontal changes were however foundfor the reading task, which may be expected as this taskevokes to a large extent the temporal lobe. For RTLpatients the activation was decreased in both temporallobes in comparison to the controls; this was com-pensated by increased (predominantly) left prefrontallobe activation. For the LTL patients the activation inthe right temporal lobe increased without strong rightfrontal activation. Left-hemispheric temporo-frontalchanges occurred in LTL patients. These analysesreveal that damage (i.e. resection) to one cerebral nodein the temporal lobe apparently affects the activation inthe (remote left) frontal lobe.

One potential confounding mechanism that mayhave contributed to the bilateral balancing of activationin LTL patients is the volume loss of possibly activat-ing tissue in the left ATL. This would suggest that ourquantitative results represent a computational artefact.When the activation extent of the left ATL, however,is omitted in the calculation of the LI for the controlsthe LI turns out to be lowered by approximately 0.01for the word generation task and by 0.1 for the readingtask. This means that the resection of activated tissuedoes not fully account for the relative large drop in LIv ulda s int ani-z

thep r ins n int sec-t theL llerr gestd eenc up.T erl dam-a

ringt ptics ime,a velt inga MRI

TL and RTL patients when only patients with a leemispheric language dominance according to theere selected. Because of the pre-operative lang

ateralization criterion and the almost perfect agent between fMRI and IAP language lateraliza

eported in many studies (Binder et al., 1996; Yetkit al., 1998; Lehericy et al., 2000; Rutten et al., 2002apreer et al., 2002; Adcock et al., 2003; Woermt al., 2003) for RTL and LTL patients, the inteemispheric redistributions effects found in this stre more likely to occur due to the surgical interv

ion. Post-operative cerebral changes would morexplain the rather large number of patients exhibiisagreement between IAP and post-operative fesults as found in this study. This means that eveatients after left ATL, who show no obvious cog

ive decline, the existing brain language system oeft hemisphere is affected and reorganized, despire-operative precautions.

No significant effects of temporo-frontal distribion changes were observed for the word generaask in neither of the patient groups. This is most lik

alue for the LTL group. Moreover, volume loss wolso not explain the increase in activation of region

he contralateral hemisphere, which cerebral reorgation indeed does.

Standard ATL procedures, as performed inatients included in this study, are usually smallepatial extent in the language dominant lobe thahe non-dominant lobe. This is reflected by the reion volumes which are on average 50% smaller inTL than in the RTL patient group. Despite the smaesection volume in the left temporal lobe, the stroneviations in hemispheric activation patterns betwontrols and patients occur in the LTL patient grohis notion fits to the idea that during left ATL prop

anguage areas or connections to them might beged, but remain unaffected during right ATL.

Interhemispheric reorganisations that occur duhe time course of lesion development and epileeizure accumulation may evaluate gradually in tnd at the time of surgery may not be at the le

o be detectable with the IAP. All patients exhibittypical language representation according the f


outcome in our study, do show, to some extent, rem-nant left-hemispheric brain activity. It remains unclearto which extent each activation site, whether left- orright-sided, contributes to the language processing.Moreover, the age at surgery for these patients was 18years or higher, which is considered suboptimal forcomplete shifts of language function across the hemi-spheres. This suggests therefore, that in addition to theuse of homologuous brain regions in the contralateralhemisphere preservation of the neural networks in theresected hemisphere exists. This mechanism has alsopreviously been suggested for patients with languagefunction recovery after stroke (Calvert et al., 2000)and patients with lesions with late onset (Muller et al.,1999). For the patients with left-sided resection in thisstudy, however, it remains unknown whether the devi-ating activation patterns are related to recovery afterimpairment or not.

The fMRI technique is more sensitive to detectlanguage capabilities of the right hemisphere than theIAP. Therefore, a comparison between pre-operativeIAP results and post-operative fMRI results is com-plicated. Forthcoming studies intended to elucidateplasticity effects should feature pre-surgical imagingresults as well. Such a study approach would alsoenable to demonstrate to which extent the righth agep thec nalp urea uter er,b hibitp agef ap s asm ngesi ec thea testr ingl mayb stingb d tos intsa then inala es in

language organization, changes in neuropsychologicalperformance, and patient characteristics as gender,onset of seizures, and seizure frequency, but alsoseizure spread and spiking frequency (Janszky et al.,2003). Moreover, it is noted that we used an observa-tional study design that does not allow to conclude oncausal relations instead of an experimental design.

In conclusion, hemispheric distributions of lan-guage activation obtained from post-operative fMRIexams highly differ between left and right ATL patientgroups. Language activation patterns in patients afterleft ATL are less lateralized to the left hemisphereor are more bi-hemispherically distributed comparedto both controls and RTL patients. Post-operativeepilepsy patients also may have a significantly differ-ent temporo-frontal distribution of activation comparedto controls. The results of this study on postoperativepatients, in combination with recently published resultson pre-operative patients (Adcock et al., 2003), suggestthat resection of the left anterior temporal lobe resultsin interhemispheric redistribution effects for cerebrallanguage organization, involving homologuous brainareas contralateral to the resected side.

A

2 oft ds.

R

A ws,es inwith

A lnessrbitalnd.

B icalper-76.

B ip-gical

B mina-ison

emisphere is pre-operatively involved in langurocessing and avoid complications concerningomparison between different techniques. Additiore-operative fMRI would enable to directly measctivation changes yielding explicit evidence of acedistributions effects in individual patients. Moreovecause LTL patients are much more at risk to exost-operative declines in verbal memory and langu

unctioning it would be interesting to investigateossible relationship between cognitive declineeasured with neuropsychological tests and cha

n brain activation distributions. A limitation of thurrent study is the clinical approach, therebybsence of post-operative neuropsychologicalesults. Although no deficits were noted concernanguage performance after surgery, subtle deficitse detectable by detailed neuropsychological teefore and after surgery. Future work is encouragetudy TLE patients before and on several time-pofter surgery. Ideally one would like to increaseumber of patients included in such a longitudpproach to track correlations between the chang

cknowledgment

This study is supported by grant number 99-0he National Epilepsy Foundation of The Netherlan

eferences

dcock, J.E., Wise, R.G., Oxbury, J.M., Oxbury, S.M., MattheP.M., 2003. Quantitative fMRI assessment of the differenclateralization of language-related brain activation in patientstemporal lobe epilepsy. Neuroimage 18, 423–438.

ldenkamp, A.P., Boon, P.A., Deblaere, K., et al., 2003. Usefuof language and memory testing during intracarotid amobatesting: observations from a fMRI study. Acta Neurol. Sca108, 147–152.

axendale, S., 2002. The role of functional MRI in the presurginvestigation of temporal lobe epilepsy patients: a clinicalspective and review. J. Clin. Exp. Neuropsychol. 24, 664–6

ell, B.D., Davies, K.G., 1998. Anterior temporal lobectomy, hpocampal sclerosis, and memory: recent neuropsycholofindings. Neuropsychol. Rev. 8, 25–41.

inder, J.R., Swanson, S.J., Hammeke, T.A., et al., 1996. Detertion of language dominance using functional MRI: a comparwith the Wada test. Neurology 46, 978–984.


Calvert, G.A., Brammer, M.J., Morris, R.G., Williams, S.C., King,N., Matthews, P.M., 2000. Using fMRI to study recovery fromacquired dysphasia. Brain Lang. 71, 391–399.

Clusmann, H., Schramm, J., Kral, T., et al., 2002. Prognostic factorsand outcome after different types of resection for temporal lobeepilepsy. J. Neurosurg. 97, 1131–1141.

Davies, K.G., Bell, B.D., Bush, A.J., Hermann, B.P., Curtis Dohan,F., Jaap, A.S., 1998. Naming decline after left anterior temporallobectomy correlates with pathological status of resected hip-pocampus. Epilepsia 39, 407–419.

Deblaere, K., Backes, W.H., Hofmann, P., et al., 2002. Developing acomprehensive presurgical functional MRI protocol for patientswith intractable temporal lobe epilepsy: a pilot study. Neurora-diology 44, 667–673.

Helmstaedter, C., Kurthen, M., Linke, D.B., Elger, C.E., 1997. Pat-terns of language dominance in focal left and right hemisphereepilepsies: relation to MRI findings, EEG, sex, and age at onsetof epilepsy. Brain Cogn. 33, 135–155.

Hermann, B.P., Wyler, A.R., Somes, G., Clement, L., 1994. Dys-nomia after left anterior temporal lobectomy without func-tional mapping: frequency and correlates. Neurosurgery 35, 52–57.

Hertz-Pannier, L., Chiron, C., Jambaque, I., et al., 2002. Late plas-ticity for language in a child’s non-dominant hemisphere. A pre-and post-surgery fMRI study. Brain 125, 361–372.

Holmes, A.P., Friston, K.J., 1997. Statistical Parametric Map-ping, Chapter 3: Statistical Models and Experimental Design.Wellcome Department of Cognitive Neurology, http://www.fil.ion.ucl.ac.uk/spm/notes/course.html.

Janszky, J., Jokeit, H., Heinemann, D., Schulz, R., Woermann, F.G.,Ebner, A., 2003. Epileptic activity influences the speech orga-

43–

L after

L oryral83–

L val-ared

Loring, D.W., Meador, K.J., Lee, G.P., et al., 1990. Cerebral languagelateralization: evidence from intracarotid amobarbital testing.Neuropsychologia 28, 831–838.

Maldjian, J.A., Laurienti, P.J., Kraft, R.A., Burdette, J.H., 2003. Anautomated method for neuranatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. Neuroimage 9, 1233–1239.

Muller, R-A., Rothermel, R.D., Behen, M.E., Muzik, O.,Chakraborty, P.K., Chugani, H.T., 1999. Language organizationin patients with early and late left-hemisphere lesion: a PET study.Neuropsychologia 37, 545–557.

Ojemann, G., Ojemann, J., Lettich, E., Berger, M., 1989. Corticallanguage localization in left-dominant hemipshere. J. Neurosurg.71, 316–326.

Rausch, R., Kraemer, S., Pietras, C.J., Le, M., Vickrey, B.G., Passaro,E.A., 2003. Early and late cognitive changes following temporallobe surgery for epilepsy. Neurology 60, 951–959.

Risse, G.L., Gates, J.R., Fangmann, M.C., 1997. A consideration ofbilateral language representation based on the intracarotid amo-barbital procedure. Brain Cogn. 33, 118–132.

Rutten, G-J., Ramsey, N., Van Rijen, P., Alpherts, W., Van Veelen,C., 2002a. fMRI-determined language lateralization in patientswith unilateral or mixed language dominance according to theWada test. Neuroimage 17, 446–460.

Rutten, G.-J., Ramsey, N., Van Rijen, P., Van Veelen, 2002b. Repro-ducibility of fMRI-determined language lateralization in individ-ual subjects. Brain Lang. 80, 421–437.

Spreer, J., Arnold, S., Quiske, A., et al., 2002. Determination ofhemisphere dominance for language: comparison of frontal andtemporal fMRI activation with intracarotid amytal testing. Neu-roradiology 44, 467–474.

Springer, J.A., Binder, J.R., Hammeke, T.A., et al., 1999. LanguageMRI

W amy-surg.

W e lat-psy.

Y MRage

nization in medial temporal lobe epilepsy. Brain 126, 202051.

angfitt, J.T., Rausch, R., 1996. Word-finding deficits persistleft anterotemporal lobectomy. Arch Neurol. 53, 72–76.

ee, T.M.C., Yip, J.T.H., Jones-Gotmannt, M., 2002. Memdeficits after resection from left or right anterior tempolobe in humans: a meta-analytic review. Epilepsia 43, 2291.

ehericy, S., Cohen, L., Bazin, B., et al., 2000. Functional MR euation of temporal and frontal language dominance compwith the Wada test. Neurology 54, 1625–1633.

dominance in normal and epilepsy subjects. A functionalstudy. Brain 122, 2033–2045.

ada, J., Rasmussen, T., 1960. Intracarotid injection of sodiumtal for lateralization of cerebral speech dominance. J. Neuro17, 266–282.

oermann, F.G., Jokeit, H., Luerding, R., et al., 2003. Languageralization by Wada test and fMRI in 100 patients with epileNeurology 61, 699–701.

etkin, F.Z., Swanson, Fischer, M., et al., 1998. Functionalof frontal lobe activation: comparison with the Wada languresults. Am. J. Neuroradiol. 19, 1095–1098.

review language activation distributions revealed by fmri in post ... · reference and to determine...

Documents