BRAIN IMAGING NEUROREPORT

0959-4965 & Lippincott Williams & Wilkins Vol 12 No 13 17 September 2001 2785

Language representation in a patient with adominant right hemisphere: fMRI evidence

for an intrahemispheric reorganisation

Mohamed Seghier,1,2 FrancËois Lazeyras,1 Shahan Momjian,3 Jean-Marie Annoni,4 Nicolas de Tribolet3

and Asaid Khateb4,5,CA

Departments of 1Radiology and 3Neurosurgery, and 4Neuropsychology Unit and 5Functional Brain Mapping Laboratory,Department of Neurology, Geneva University Hospital, 24, rue Micheli-du-Crest, CH-1211 Geneva 14; 2Plurifaculty Program of

Cognitive Neuroscience, University of Geneva, Switzerland

CA,4Corresponding Author and Address

Received 16 May 2001; accepted 9 July 2001

Studies have suggested that congenital left hemispheric (LH)frontal arteriovenous malformations (AVMs) are associatedwith an early transfer of language to right hemisphere (RH)frontal regions. The question remains whether such anatomo-functional reorganisation is due to RH compensatory abilitiesor to a general principle of lateral shift. In this study, we usedfMRI language paradigms to investigate the case of a patientpresenting aphasic symptoms following an haemorrhage due toa right frontal AVM. Prior to surgery, fMRI showed thatlanguage processing was con®ned to the RH, suggesting thatlanguage had not shifted during childhood from this congeni-

tally dominant RH to the LH. After surgery, the patientpresented severe aphasia that recovered to presurgical levelwithin 70 days. At this time, fMRI showed that language taskswere still not associated with activations in the LH. Theseresults suggest that the principles of early cerebral reorganisa-tion after congenital lesions may differ in the RH and the LH.In addition, they support the idea that ef®cient restoration oflanguage is achieved if a suf®ciently large neuronal network ispreserved around the lesion. NeuroReport 12:2785±2790 &2001 Lippincott Williams & Wilkins.

Key words: AVM; Frontal cortex; Functional magnetic resonance imaging; Functional recovery; Phonologic and semantic tasks; Right hemisphere

INTRODUCTIONPrior to brain surgery, the question of the functionalintegrity of certain brain regions is of major importance toprevent major post-surgical cognitive de®cits, particularlywhen atypical functional organisation is suspected. In lefthemisphere (LH) damaged patients, a crucial questionconcerns the hemispheric dominance and the preciselocalisation of language areas. For epilepsy surgery forexample, the invasive Wada test is carried out when thelanguage-dominant hemisphere is not clearly determined[1]. fMRI paradigms, using various linguistic tasks includ-ing verbal ¯uency, rhyme detection and semantic judg-ment, have been proposed as non-invasive methods toassess the language-dominant hemisphere [2,3] and toprecisely localise the different brain areas involved indifferent language components [4]. fMRI studies have alsobeen conducted in aphasic patients to assess the neuro-logical basis of language recovery in single patients [5].Such studies could also provide evidence for both intra-hemispheric and inter-hemispheric reorganisation inchronic neurovascular diseases such as cerebral arterio-venous malformations (AVM) [6,7]. The interest in AVM isdue to its congenital nature, existing prior to languagedevelopment. Some studies reported that patients withAVMs affecting the left perisylvian regions recruited more

intensively the RH into language processing than strokepatients whose left cortical language networks were da-maged in adulthood [8]. Other reports suggested that RHparticipation occurs more particularly in left frontal AVMs[7] while more posterior lesions induce an intrahemi-spheric posterior to anterior reorganisation [6]. However,these suggestions rely on cases where only classical lan-guage organisation was studied. It would be interesting toverify whethr such inter-frontal reorganisation afterchronic lesions is a more general process and can takeplace also in atypical language lateralisation.

In this study, we used fMRI paradigms to determinehemispheric dominance and cortical organisation for lan-guage during two activation tasks in controls and in apatient presenting aphasic symptoms following an haemor-rhage caused by a right frontal opercular AVM. Our aimwas to verify whether only inter- or intra-hemisphericlanguage reorganisation could be observed in this patientduring a phonological and a semantic judgement task bothprior and after brain surgery.

MATERIALS AND METHODSSubjects: Six healthy control men and one patient partici-pated in this study. Control subjects were right-handed(mean age 29� 11 years), native speakers of French and

had normal vision. Both controls and the patient gave theirinformed consent prior to their participation in this study.The patient, SDA, was a 36-year-old Portuguese woman,¯uent French speaking and ambidextrous. She presented atour institution with headache and word-®nding dif®cul-ties. The neurological examination was otherwise normal.Detailed language examination mainly showed mild nam-ing impairment in spontaneous speech. On standardisedtesting, her performance was moderately impaired in theFrench version of Boston naming test (BNT) [9] andshowed some semantic paraphasias. Repetition of sen-tences, but not of single words, was impaired. Verbalsemantic and phonological ¯uency tests (referred to ascontrolled oral word association [9]) both showed signi®-cantly reduced performance. Oral and written comprehen-sion were normal for words and sentences.

Brain CT and MRI scans revealed a 3 cm wide rightfronto-opercular haemorrhage extending into the rightBrodmann area 44/45 with mild surrounding oedema andmass effect. An abnormal vein was seen on the MRI scannear the haematoma and a cerebral angiogram demon-strated a super®cial right fronto-opercular AVM. The niduswas also 3 cm in diameter. These radiological examinationsshowed no abnormalities in the LH.

The patient underwent a ®rst fMRI session before thecomplete surgical excision of the AVM. The immediatepost-operative neurological examination revealed a severenon-¯uent aphasia. In addition, the patient showed apraxiaof speech. Spontaneous speech was limited to isolatedwords with phonological and semantic paraphasias. Rep-etition was impaired for isolated words and non-words(e.g. `taje' instead of `kafe'). Object naming was severelyimpaired with phonological paraphasias and neologisms(e.g. `kla' instead of `kle'). The BNT and the verbal ¯uencytasks were not testable. Oral comprehension tested byeither a name-to-image or simple sentence-to-image asso-ciation task showed . 50% erroneous responses, attesting asevere impairment. Reading and writing to dictation werebetter preserved despite the presence of paragraphias onverbs and function words (e.g. `joue' instead of `je' and`vois' instead of `vai'). Finally, the patient showed an ideo-motor and a buccofacial apraxia.

Follow-up CT and MRI (with diffusion sequences) 2and 4 days, respectively, after surgery demonstrated alocal vasogenic oedema related to the resorption of thehaematoma and to the operative approach and also asmall cytotoxic right temporal anterior oedema markingischemia. Control angiogram revealed the complete resec-tion of the AVM. The aphasic symptomatology had anongoing regression and the patient underwent the secondfMRI about 70 days after surgery. At this time, apraxia ofspeech had nearly disappeared, but the patient stillpresented some word-®nding dif®culties. Particularly,naming performances were within normal limits on theBoston Diagnostic Aphasia Examination (BDAE [9]), andreturned to the pre-surgical level at the BNT. Similarly, inverbal semantic and phonological ¯uency as well asrepetition and oral comprehension, the patient showedscores comparable with pre-operative testing. Finally,written language was characterised by the decrease of thenumber of paragraphias, although syntactic errors werestill observed.

Paradigm and stimuli: For all experiments, a block para-digm was applied, which alternated between control andactivation sequences. The stimuli, concrete imaginablenouns of high frequency, were presented to the subjectsvia a video projector, a front-projection screen and asystem of mirrors fastened to the head coil. The visual ®eldspanned by this setup was �15 3 118. In the ®rst phono-logic task, subjects had to give a response with the righthand if the two words presented simultaneously on thescreen rhymed. In the second semantic task, they had todecide whether the two words belonged to the samesemantic category [10]. In the control condition, wordswere replaced by Greek letter strings and subjects had todecide whether the two strings of the pair were visuallyidentical or not. In both activation and control conditions,the response ratio was 2:1. Stimulus pairs were presentedon the screen for 600 ms at 0.5 Hz, and the two conditions(24 s each) were repeated ®ve times, providing a totalduration of 4 min per stimulation run.

MR acquisition: Experiments were performed on a 1.5 Twhole-body Eclipse system (Marconi Medical Systems,Cleveland, OH) using the standard head coil con®guration(body coil for excitation, head coil for detection). Acquiredmulti-slice volume was positioned on sagital scout images.Following the functional MR scans, anatomical scans wereperformed. A ®rst GRE T1-weighted sequence (TR/TE/Flip� 15 ms/4.47 ms/258, FOV� 250 mm, matrix� 256 3256, slice thickness� 5 mm) was performed to acquire thesame volume as in the functional session. Anatomicalreference images consisted of a 3D GRE T1-weightedsequence (TE� 5 ms, FOV� 250 mm, matrix� 256 3 256,slice thickness� 1.25 mm).

Functional imaging consisted of an EPI GRE sequence(TR/TE/Flip� 2 s/40 ms/808, FOV� 250 mm, matrix�128 3 128, 17 contiguous 5 mm axial slices). The exploredvolume was measured 12 times during each condition ofthe paradigm. Functional scanning was always precededby 10 s of dummy scans to ensure tissue steady-statemagnetization.

Data processing: Data processing relied upon cross-corre-lation analysis [11] after motion correction [12], usingMEDX software (Sensor Systems, Sterling, VA). Data weresmoothed spatially by convolution with a 5 mm±5 mm±10 mm FWHM Gaussian kernel. The cross-correlation, as aZ-value, was calculated pixelwise between a delayed box-car function and the set of measurements, after auto-correlation correction (4 s) [13]. Afterwards, the statisticaldistribution of the Z-values was calculated and a probabil-ity value for each Z-value was attributed. This approachenables one to work with the same probability for eachexperiment and allows more accuracy when addressingthe question of the reproducibility [14]. Only clusters of. 4 pixels showing statistically signi®cant Z-score (atp� 0.01) were considered [15]. Maps were normalized tothe Talairach and Tournoux space.

RESULTSActivation maps for the six subjects showed the implica-tion of left cortical areas, principally the dorsolateralprefrontal cortex (BA 9/46), the middle temporal gyrus

NEUROREPORT M. SEGHIER ET AL.

2786 Vol 12 No 13 17 September 2001

(BA 22/21) and the inferior frontal gyrus (BA 44/45). Otheractivations were found in the superior parietal lobule/superior occipital gyrus (BA 7/19) for the two tasks, theinferior parietal lobule (BA 40) and the precentral gyrus(BA 6/4) particularly during the semantic task (Table 1).All responses were localized in the LH, except in onesubject who showed, mainly during the semantic task,additional RH activations in the inferior frontal gyrus (BA44/45), the superior frontal gyrus (BA 8) and the cuneus/lingual gyrus (BA 18/31).

Figure 1 illustrates the anatomical images of the patientSDA with the superimposition of activated brain regionsobserved before and after surgery. Prior to surgery, activa-tions were essentially observed in the RH. These involvedin both tasks the inferior frontal gyrus (BA 44/45), thedorsolateral prefrontal cortex (BA 9/46) and the angular/superior occipital gyrus (BA 39/19). Smaller clusters ofactivated pixels were also detected during the categorisa-tion task in the dorsal frontal gyrus (BA 6), more preciselyin the supplementary motor area (SMA). Postoperatively,essentially three RH regions were activated: the dorsolat-eral prefrontal cortex (BA 9/46), the dorsal frontal gyrus(SMA, BA 6) for the two tasks, and the angular gyrusextending into the superior occipital gyrus (BA 39/19) forthe categorisation task (Table 1). No activation was foundin the LH. The temporal gyri did not show activationseither preoperatively or postoperatively, although someweak activation was observed in the middle temporalgyrus (BA 21/22) but failed to reach the statistical level ofsigni®cance. The major differences between the two scans

concerned in the rhyme task the post-operative disappear-ance of inferior frontal gyrus and angular/superior occiptalgyrus activation and the appearance of activation in theSMA. In the semantic task, the only difference was absenceof activation, post-operatively, in the inferior frontal gyrus.

DISCUSSIONThe most signi®cant observation in this report concerns theunexpected implication of the RH in language processingin a patient with a congenital right opercular malformation.In control subjects, the observed pattern of brain activationcon®rms previous ®ndings, in particular by showing theinvolvement in both tasks of the dorsolateral prefrontalcortex, the inferior frontal gyrus and the middle/superiortemporal gyrus in the LH [4,16].

Neuroimaging studies have also suggested that the RHplays a major role in aphasia recovery after stroke in LHlanguage regions [17,18]. In right-handed patients withcongenitally acquired AVM in the LH, other studies usingboth superselective Wada tests and fMRI paradigms havereported a complete or partial RH language lateralisation[6±8], suggesting that language representation had shiftedfrom the classically dominant LH to the RH during theearly development. In our case, we investigated a patientwith a congenital right frontal opercular AVM. The RHdominance for language processing was suggested by theclinical presentation (motor dysphasia) after a haemor-rhage due to the malformation and by the functionalintegrity of the LH as attested by the anatomical pre-surgical CT and MRI scans. A language-dominant RH has

Table 1. Activated brain regions with their coordinates in the Talairach space for the control subjects and patient SDA (pre- and post-surgery) in therhyme and semantic task.

Control subjects

Task Activation sites Brodmann areas (n) Coordinates

x y z

Rhyme dorsolateral prefrontal cortex 9/46 (6/6) ÿ39 22 34middle/superior temporal gyrus 22/21 (5/6) ÿ54 ÿ33 8inferior frontal gyrus 44/45 (4/6) ÿ46 ÿ23 14superior parietal lobule/superior occipital gyrus 7/19 (2/6) ÿ22 ÿ70 40inferior parietal lobule 40 (1/6) ÿ34 ÿ46 40

Semantic dorsolateral prefrontal cortex 9/46 (6/6) ÿ41 19 29inferior frontal gyrus 44/45 (6/6) ÿ44 23 10middle/superior temporal gyrus 22/21 (4/6) ÿ51 ÿ44 12inferior parietal lobule 40 (3/6) ÿ38 ÿ57 38precentral gyrus 6/4 (3/6) ÿ39 ÿ5 41superior parietal lobule/superior occipital gyrus 7/19 (2/6) ÿ26 ÿ78 40

Patient SDA: Pre-operativeRhyme inferior frontal gyrus 44/45 40 24 10

dorsolateral prefrontal cortex 9/46 48 30 26angular/superior occipital gyrus 39/19 34 ÿ66 32

Semantic inferior frontal gyrus 44/45 34 24 6dorsolateral prefrontal cortex 9/46 44 30 28angular/superior occipital gyrus 39/19 34 ÿ70 30dorsal frontal gyrus (SMA) 6 4 12 52

Patient SDA: Post-operativeRhyme dorsolateral prefrontal cortex 9/46 48 24 24

dorsal frontal gyrus (SMA) 6 4 12 48Semantic dorsolateral prefrontal cortex 9/46 50 20 26

angular/superior occipital gyrus 39/19 36 ÿ76 28dorsal frontal gyrus (SMA) 6 4 10 52

fMRI EVIDENCE FOR AN INTRAHEMISPHERIC REORGANISATION NEUROREPORT

Vol 12 No 13 17 September 2001 2787

been observed in �15% of an ambidextrous control popu-lation [19]. Based on previous LH AVM observationswhere language lateralisation had partially or completelyshifted from the LH to the RH [6,8], one could have

expected to observe a bilateral representation of the lan-guage network in our patient involving at least left frontalregions. However, in our study, semantic and phonologicallanguage tasks before surgery only showed activations in

Fig. 1. Activated brain regions (displayed on a sagittal slice and three axial slices) in the phonological (a) and the semantic task (b) before (1) and after(2) the AVM excision. Notice the AVM location in A1 (sagittal slice at left) indicated by a white arrow.

NEUROREPORT M. SEGHIER ET AL.

2788 Vol 12 No 13 17 September 2001

the patient's RH, particularly in the vicinity of the AVM.These activations could be explained by the preservation ofa suf®ciently large neuronal network around the malfor-mation or the infarct to allow reorganisation [20,21]. Theimplication of the right inferior frontal gyrus (BA 44/45),distant by few millimeters from the haemorrhage (AVM),is thus not surprising, since it corresponds to the homo-topic region of the traditional left Broca area generallyactivated during language tasks [4].

More than 2 months postoperatively and after aphasiarecovery, the results indicated no activation in the LH.However, compared with the results before surgery, thepost-surgery results showed little (or no) activation in theright inferior frontal gyrus. This pattern change mightaccount for the worsening of the dysphasia after surgery,notably with the persisting word-®nding and syntacticdif®culties [22]. Post-surgery activations also showed theinvolvement during the two tasks of the dorsal frontalgyrus which comprises the SMA. This ®nding corroboratesprevious observations showing that SMA is a major com-pensatory region during the early weeks of languagerecovery [20]. In addition, the implication of the angulargyrus with some extension into the superior occipital gyruswas more inferior than in the control subjects. The activa-tion of such extrastriate visual areas by semantic tasks in apatient recovering from acquired dysphasia has been inter-preted in terms of a change in the cognitive strategiesrather than a manifestation of cortical plasticity [18].

The activation of the superior/middle temporal gyri,observed in controls in the rhyme and the semantic task,was not observed in the patient both in pre- and post-surgery scans. Although surprising, the absence of thisactivation is in accordance with previous cases where suchactivation was not found in a patient with a left frontalAVM [7] and in another patient with a left ischemic lesion[18]. It is noteworthy that the aphasic symptoms of thelatter patient were in some regards similar to thosedescribed in our patient, in particular word-®nding andsentence-construction dif®culties. Furthermore, such anabsence of activation in the superior/middle temporal gyriwas observed in one of the six controls tested in this studysuggesting that the involvement of this region may dependon the subjects' strategies to resolve the task demands.

The phonological and the semantic tasks used in thisstudy involved the second language (French) of the patientSDA. Previous studies have, however, established thatnative and secondary languages activated overlappingregions in the frontal lobe [23] and, accordingly, the patternof activation observed here cannot be attributed to the useof the second language. In addition, no pathologicalswitching between the native (Portuguese) and the ac-quired language (French) was manifested by the patient[24].

Our results provide additional ground in the largedebate concerning the mechanisms of recovery from apha-sia following lesions in language zones. While someimaging studies have linked recovery essentially to therecruitment of homologous compensatory RH [8,17,25],others suggest that recovery is mainly achieved by implica-tion of preserved LH language areas [26] which constitutea major determinant for long-term prognosis of aphasia[20,27]. More recently, it has been suggested that recovery

from dysphasia may also depend on processing strategiesadopted by the patients and consequently the recruitmentof both the preserved language areas and on the use ofhomologous regions in the other hemisphere [18]. The caseof SDA presented here indicates that cerebral plasticitywhich has allowed intrahemispheric reorganisation of herlanguage network during the early development, allowedalso language recovery after the haemorrhage by the use ofher RH functionally intact regions. Further investigations,in particular on aphasic patients with RH lesions arenecessary to confront the large literature on aphasic pa-tients with LH lesions.

CONCLUSIONThe main result reported in this case study is the demon-stration prior to surgery of a RH language lateralisation inan ambidextrous patient despite the fact that the congenitalAVM could have forced a shift of language towards leftfrontal regions. The second important result is the absenceof interhemispheric reorganisation to explain recovery afterthe excision of the right frontal AVM. The absence of LHinvolvement in language processing during both the earlychildhood and following lesion excision suggest that inter-hemispheric transfer of language in patients with frontalcerebral arteriovenous malformation is not always thenecessary process for language development and recovery.

REFERENCES1. Wada J and Rasmusen T. J Neurosurg 17, 266±282 (1960).

2. Binder JR, Swanson SJ, Hammeke TA et al. Neurology 46, 978±984 (1996).

3. LeÂhericy S, Cohen L, Bazin B et al. Neurology 54, 1625±1633 (2000).

4. Fiez JA and Petersen SE. Proc Natl Acad Sci USA 95, 914±921 (1998).

5. Pizzamiglio L, Galati G and Committeri G. Cortex 37, 11±31 (2001).

6. Lazar RM, Marshall RS, Pile-Spellman J et al. Neurology 49, 802±808

(1997).

7. Lazar RM, Marshall RS, Pile-Spellman J et al. Neuropsychologia 38,

1325±1332 (2000).

8. Vikingstad EM, Cao Y, Thomas AJ et al. Neurosurgery 47, 562±570

(2000).

9. Spreen O and Strauss E. A Compendium of Neuropsychological Tests:

Administration, Norms and Commentary. New York: Oxford University

Press; 1998.

10. Khateb A, Annoni JM, Landis T et al. Int J Psychopysiol 32, 215±231

(1999).

11. Bandettini PA, Jesmanowicz A, Wong EC et al. Magn Reson Med 30,

161±173 (1993).

12. Woods RP, Cherry SR and Mazziotta JC. J Comput Assist Tomogr 16,

620±633 (1992).

13. Friston KJ, Holmes AP, Poline JB et al. Neuroimage 2, 45±53 (1995).

14. Moser E, Teichtmeister C and Diemling M. Magn Reson Imag 14,

567±579 (1996).

15. Forman SD, Cohen JD, Fitzgerald M et al. Magn Reson Med 33, 636±647

(1995).

16. Poldrack RA, Wagner AD, Prull MW et al. Neuroimage 10, 15±35 (1999).

17. Mimura M, Kato M, Sano Y et al. Brain 121, 2083±2094 (1998).

18. Calvert GA, Brammer MJ, Morris RG et al. Brain Lang 71, 391±399

(2000).

19. Knecht S, DraÈger B, Deppe M et al. Brain 123, 2512±2518 (2000).

20. Karbe H, Thiel A, Weber-Luxenburger G et al. Brain Lang 64, 215±230

(1998).

21. Cao Y, Vikingstad EM, George KP et al. Stroke 30, 2331±2340 (1999).

22. Embick D, Marantz A, Miyashita Y et al. Proc Natl Acad Sci USA 97,

6150±6154 (2000).

23. Kim KH, Relkin NR, Lee KM et al. Nature 388, 171±174 (1997).

24. Fabbro F, Skrap M and Aglioti S. J Neurol Neurosurg Psychiatry 68,

650±652 (2000).

25. Thulborn KR, Carpenter PA and Just MA. Stroke 30, 749±754 (1999).

fMRI EVIDENCE FOR AN INTRAHEMISPHERIC REORGANISATION NEUROREPORT

Vol 12 No 13 17 September 2001 2789

26. Heiss WD, Kessler J, Thiel A et al. Ann Neurol 45, 430±438 (1999). 27. Miura K, Nakamura Y, Miura F et al. J Neurol 246, 939±942 (1999).

Acknowledgements: We thank F. Henry, I. Zimine and D. Joliat for their technical assistance, F. Bernasconi Pertusio for hercomments on language examination reports, J. Delavelle, A.J. Pegna, E. Mayer, and C.M. Michel for their helpful discussions. Thisresearch was supported by the Swiss National Science Foundation (grant no. 31-61680-00) and the Programme Plurifacultaire

des Neurosciences Cognitives of Geneva University.

NEUROREPORT M. SEGHIER ET AL.

2790 Vol 12 No 13 17 September 2001