NeuroImage 17, 174 –183 (2002) doi:10.1006/nimg.2002.1238 Neural Substrates of Spoken Language Rehabilitation in an Aphasic Patient: An fMRI Study A. Léger, J-F. Démonet, S. Ruff, B. Aithamon, B. Touyeras, M. Puel, K. Boulanouar, and D. Cardebat INSERM U 455 and Department of Neurology, CHU Purpan, Toulouse, France Received October 23, 2002 Little is known about the neural counterparts of speech therapy in aphasic patients. An fMRI experiment was performed before and after a specific and intensive speech output therapy in RC, a patient with long-lasting speech output deficit following a leftsided ischemic lesion. Overt picture naming and picture/word rhyming were used as activation tasks in RC and 6 control subjects. The naming task concerned the output lexicon deficit to be rehabilitated while rhyming referred to preserved levels of processing and was used to control for repetition effect. The speech therapy program improved naming performance. By comparison to the pattern observed before therapy, the naming task after therapy induced a pattern of activation close to that observed in control subjects, involving left-sided language areas surrounding the lesion. Speech therapy effect was associated with activations in Broca’s area and the left supra-marginal gyrus, which might reflect a therapyinduced phonological compensatory strategy for naming. © 2002 Elsevier Science (USA) INTRODUCTION Aphasia following stroke generally evolves spontaneously toward some degree of recovery, despite persistent brain damage. Several neuroimaging studies sought to evidence neural patterns related to spontaneous recovery from aphasia. A crucial issue regards whether language improvement is sustained by the left hemisphere zones spared by the lesion or by recruitment of homologous right hemisphere regions. Indeed, some activation studies showed that spared perilesional regions of the left hemisphere were the main substrate of recovery mechanisms (Heiss et al., 1999; Warburton et al., 1999) whereas an involvement of the right hemisphere has been interpreted as a compensatory shift of function to homologous right-sided territories by other authors (Buckner et al., 1996; Cardebat et al., 1994; Ohyama et al., 1996; Thulborn et al., 1999; Weiller et al., 1995). Such a discrepancy across studies is probably related to the heterogeneity of patients in 1053-8119/02 $35.00 © 2002 Elsevier Science (USA) All rights reserved. terms of lesion localization, cognitive deficits, and activation tasks. Remediation by speech therapy can help spontaneous recovery of language in aphasic patients (Robey, 1994; Holland et al., 1996), even at the chronic stage (Elman and Bernstein-Ellis, 1999; Katz and Wertz, 1997) especially when an intensive training program is used (Pulvermuller et al., 2001). Contradictory results came from the very few studies of therapy-induced neurofunctional changes in aphasic patients. For instance left-sided activations were reported by Belin et al. (1996) and Small et al. (1998) in patients who benefited from Melodic Intonation Therapy or a phonological training of reading aloud, respectively. On the other hand, Musso et al. (1999) demonstrated a correlation between increased activity in the right temporal cortex and comprehension scores, in Wernicke-type patients undergoing a brief and intensive training of comprehension between scanning sessions. Similarly, a therapy program devoted to sentence processing was associated with changes in the right hemisphere during a sentence-picture matching task, in a patient described by Thompson (2000). On the whole, these few results suggested that remediation might elicit activation in the right hemisphere whereas the left hemisphere would be recruited when speech output was required. Although group studies, when lesion and neuropsychological profile are controlled, offer a good context for evaluating the physiological changes associated with the language improvement induced by therapy, single case studies should be interesting as well, as both remediation program and activation tasks can be specifically set up. In the present study, we report the case of an aphasic patient presenting a massive speech output deficit in whom we conducted a specific language therapy devoted to output lexicon rehabilitation. An fMRI experiment including two lexical activation tasks was performed before and after therapy. In order to assess the specificity of therapy-induced brain reorganization, one of these tasks concerned preserved levels of processing while the other focused on the speech 174 SPEECH THERAPY AND fMRI IN AN APHASIC PATIENT output deficit to be rehabilitated. fMRI data from the patient were compared to those from 6 healthy volunteers who performed the same functional neuroimaging experiment. We hypothesized that activation in left perilesional areas would parallel performance improvement in speech output after intensive language therapy. CASE HISTORY RC, a highly educated 42-year-old right-handed man, has suffered from a left middle cerebral artery infarct on April 1998, after a spontaneous dissection of the left internal carotid. Initially, he presented a right hemiplegia, which disappeared in a few weeks, and a severe mixed aphasia that persisted 6 months later. The present study was carried out 2 years poststroke. Structural MRI (see Fig. 1), performed in 2000, disclosed a left hemispheric lesion in the superficial territory of the middle cerebral artery. According to the atlases from Talairach and Tournoux (1988) and Duvernoy (1992), the lesion involved the posterior half of the insular cortex, and the posterior two-thirds of the superior temporal gyrus (T1) sparing the Heschl’s gyrus. The lesion spread to the parietal opercule and the inferior part of the supra-marginal gyrus. Cortical atrophy was observed in the middle part of the precentral gyrus. A general neuropsychological assessment revealed no buccofacial apraxia, agnosic, or apraxic problems. Auditory verbal short-term memory, tested by pointing to written items because of the repetition deficit, was severely impaired (digit span: 3 forward and 2 backward). Visual and long-term memories were spared. The language assessment evidenced the following pattern. RC’s spontaneous speech was effortful and displayed severe phonemic distortions leading to unintelligible fragments with conduites d’approche, impoverished use of morphological and syntactic structures, and some word-finding difficulties. These symptoms were evidenced by poor performance on word and pseudo-word repetition, reading aloud, and oral-naming tasks. All errors consisted in phonemic paraphasias that coexisted with an effortful and hesitating speech output. The patient seemed to seek for the correct articulatory gestures to be combined to achieve oral production tasks. By comparison to the massive deficit observed in actual speech output tasks, performance was normal on comprehension tasks involving semantic-lexical processing or access to the output lexicon without vocalization. For example, he scored 18/24 in a task in which he was asked to match two pictures of homonymous items among distracters (e.g., “renne” -reindeerand “reine” -queen-). Although he made occasional errors, his performance on a picture/word rhyming task 175 was remarkably close to the normal range. A similar pattern was found in writing tasks. In summary, RC presented a severe expressive aphasia reflected by his impaired spoken and written language production with a relative sparing of semanticlexical processing and access to the phonological lexicon. Together with impaired phonemic representations, severe auditory verbal short-term memory impairment could contribute to RC’s disability in producing the correct syllable sequence involved in a word. The patient was proposed a customized speech therapy program lasting 6 weeks with 6 sessions per week, 1-h per day, which focused on speech output processes. The method was based on visual memory, spared in RC, in order to teach him how to combine and produce the phonemes to be articulated during word production. The patient was trained to memorize “by heart” drawings showing the articulatory gestures associated with the syllables of the 30 words that constituted the material of the fMRI experiment. Various oral production tasks such as repetition, reading aloud, or picture naming were used for training. It should be noted that rhyming tasks were not worked out during therapy. MATERIALS AND METHODS Subjects RC underwent two fMRI sessions, the first one before the beginning of therapy (Session 1: S1), and the second one at the end of the therapeutic program (Session 2: S2). Six healthy right-handed volunteers (5 men and 1 woman, mean age 52.2 years), matched for education level, were recruited as control subjects. Control subjects had no history of neurological or psychiatric illness. fMRI experiment was performed only once for the control subjects. All control subjects and RC gave informed consent to participate in the study and the local ethics committee approved the study. Stimuli, Experimental Design, and fMRI Procedure Stimuli were (i) a set of 60 black and white line drawings of familiar objects or animals taken from the Snodgrass and Vanderwart corpus (1980) and (ii) a set of 60 French written frequent words typed in 40-point Geneva font. All the words were nouns that ranged in length from 5 to 8 letters. These stimuli were included under two activation conditions. The first one was an overt picture-naming task including the 60 pictures, half of them being used in the therapy program as training material. Subjects were instructed to name the pictures overtly but to avoid head movements while whispering responses. In the second condition, the same stimuli were used in a picture/written noun rhyming task. Subjects were FIG. 1. An illustration of the anatomy of RC’s cerebral infarct. In the top row, RC’s brain visualized in a three-dimensional rendering of the cortical surface obtained from structural MRI data. RC’s lesion is shown in red. In the lower rows, MRI axial contiguous slices (thickness ⫽ 1.2 mm) parallel to the bicommissural plane (slice 0) in RC (left hemisphere shown on right) indicate the depth of the lesion. RC’s lesion involves, in the left hemisphere, the posterior half part of the insula, the posterior two-thirds of the superior temporal gyrus, the parietal opercule, and the inferior part of the supra-marginal gyrus and spares the Heschl’s gyrus. On slice 30 and above, a limited atrophy is in the precentral gyrus. 176 SPEECH THERAPY AND fMRI IN AN APHASIC PATIENT 177 FIG. 2. A rendering showing the regions activated for (1) the Naming task and (2) the Rhyming task of (a) the control group, (b) RC before the speech therapy, and (c) RC after the speech therapy. The activated areas are projected onto a template of a standard MNI brain for the control group (all areas shown were significant at P ⬍ 0.05, uncorrected for multiple comparisons and k extent ⬎ 50) and for RC, onto a template of RC’s anatomical MRI scan (all areas shown were significant at P ⬍ 0.05, corrected for multiple comparisons and k extent ⬎ 50). asked to say, “Yes” or “No” if the written word displayed below the picture rhymed (e.g., picture of a chicken with the word “children”) or not (e.g., picture of knife with the word “hat”) with the name of the depicted object. Rhyming (15 items) and nonrhyming (15 items) trials were presented randomly. The rhyming task that included the same lexical items as the naming task was purposefully selected as control task for two reasons. First it shared many cognitive components with naming insofar as subjects had to retrieve the phonological form of the object, but without planning and producing the syllabic series involved in spoken words. Second, and most importantly, the patient performed at normal level on the rhyming task whereas performance on naming was known to be poor before therapy. The fMRI procedure alternated Naming or Rhyming with rest periods in a block design with 4 runs (Naming task, run 1 and run 3, and Rhyming task, run 2 and run 4). The duration of each block was 30 s and 1 run consisted in a succession of 12 blocks alternating activation with rest (during which a gray screen was pre- sented). In the activation conditions, 5 stimuli per block were centrally delivered via special goggles (Resonance Tech., Northridge, CA) during 4 s with an intersequence interval (gray screen) of 2 s. Responses were transmitted thanks to a microphone and recorded by the examiner. Imaging MRI was performed on a 1.5-T scanner (Siemens Vision, Erlangen, Germany) equipped for echo-planar imaging (EPI). A 3D high-resolution T 1-weighted data set of the whole brain (3D MPRAGE; 3D magnetization prepared rapid acquisition gradient echo) was acquired for each subject (128 slices, TR ⫽ 15 ms, TE ⫽ 7 ms, flip angle ⫽ 12°, FOV ⫽ 30 cm, matrix ⫽ 256 ⫻ 256, voxel size ⫽ 1.17 ⫻ 1.17 ⫻ 1 mm 3). After sagittal localization images, 10 contiguous, 5-mm-thick, axial anatomic images were obtained parallel to the intercommissural plane (from z ⫽ ⫺10 mm to z ⫽ ⫹35 mm). For functional MR imaging studies, blood oxygen level-dependent (BOLD) imaging was performed using 178 LÉGER ET AL. a T 2*-weighted single-shot EPI sequence (TE ⫽ 64 ms, flip angle ⫽ 90°, FOV ⫽ 22 cm, 128 ⫻ 128 matrix, TR ⫽ 2.95 s, 5-mm slice thickness). Each scanning run (6 min each, 6 blocks of activation, and 6 blocks of rest) thus comprises 120 image volumes (10 volumes per block of activation and 10 volumes per block of rest, except for the first block of rest which was discarded to allow for T1 stabilization and dissipation of gradientinduced auditory cortical activation). Functional MR Image Postprocessing Image analysis was carried out on a SPARC workstation (Sun Microsystems, Surrey, UK) using interactive image display software (Analyze, Biodynamics Research Unit, Mayo Clinic, Rochester, MN), Matlab (Math Works Inc., Natick, MA), and SPM99 software (Wellcome Department of Cognitive Neurology, London, UK). EPI images were normalized into Talairach’s space using affine transformations (translations and zooms in x and y axes), realigned, and smoothed using a Gaussian filter (FWHM 6-6-6 mm). The statistical analysis involved the following steps: (i) Individual analyses were performed on each of the 6 control subjects and RC for both sessions, using a hemodynamic response function modeled by a bimodal curve. “Main Contrasts” (Activation minus rest) for Naming and Rhyming tasks were calculated in the control group on the one hand and in RC for each session on the other hand. For the control group, analyses were performed using a “random-effect” model (Holmes and Friston, 1998). (ii) To explore common activations in control subjects and RC for the Main Contrasts, group analyses that involved the control subjects and RC were performed by using one-sample t tests and the randomeffect model. These analyses concerned, on the one hand, RC before therapy (S1) and RC after therapy (S2), on the other hand. The threshold was set up at P ⬍ 0.05 for peak height, uncorrected for multiple comparisons, with spatial extent k ⬎ 50. (iii) Similarities and differences between sessions 1 and 2 in RC were studied by conjunction and interaction analyses using Main Contrasts (activation minus rest) for Naming and Rhyming tasks. Areas activated by RC in Session 2 but not in Session 1 were identified by compound contrasts as follows [(Activation-Rest) RCS2 ⫺ (Activation-Rest) RCS1] ⬎ 0. An inclusive mask (with [Activation-Rest] RCS1 ⬎ 0) was used to avoid the selection of significant voxels due to deactivations (threshold mask P ⬍ 0.5). The same procedure was used to obtain the inverse contrast. Statistical threshold for activated clusters was set at P ⬍ 0.05 (corrected for multiple comparisons) for peak height and k ⬎ 50 for cluster extent, unless otherwise specified. Speech-therapy effects were assessed by contrasting Naming at the two sessions ([(Activation NamingRest) RCS2 ⫺ (Activation Naming-Rest) RCS1] ⬎ 0), with an inclusive mask (P ⫽ 0.5) corresponding to a conjunction of [Activation Naming-Rest] RCS1 ⬎ 0 and [(Activation Rhymingg⫺ Rest) RCS2 ⫺ (Activation Rhyming-Rest) RCS1] ⬎ 0. This contrast was thresholded with P ⬍ 0.05 (corrected for multiple comparisons) for peak height and k ⬎ 15 for the cluster extent. RESULTS Behavioral Results The mean accuracy in the 6 control subjects was 94.6% for the Naming task and 94.5% for the Rhyming task. These results suggest that the level of difficulty for the subjects was the same in both tasks. As expected, RC’s performance for the Rhyming task did not differ significantly from the normal subjects (RC’s hit rate was 91.7% at S1 and 95% at S2). On the Naming task, at Session 1, RC scored identically (6 correct/30) for trained and untrained items; however, the patient gave tentative responses to any item. At Session 2, a statistically significant improvement of performance was observed on naming for both trained (19/30) (␹ 2 corrected test ⫽ 9.87, P ⬍ 0.05) and untrained items (15/30) (␹ 2 corrected test ⫽ 4.8, P ⬍ 0.05). Correct responses at Session 2 included all the correct responses produced at Session 1. Improved naming scores were also noted for items not belonging to the therapy protocol, like those from the DO80 battery (Deloche et al., 1997). Before the therapy and fMRI protocol, naming impairment was noted as stable (13 correct/80 on DO80 in March 1999 and 15 correct/80 in January 2000). By contrast, after the therapy program, RC’s hit rate increased significantly (33 correct/80, ␹ 2 ⫽ 25.64, P ⬍ 0.05). In sum improvement of naming performance was not observed for the trained items only and a general positive effect, even for external material, was found. fMRI contrasts combined therefore data acquired for both trained and untrained naming stimuli. fMRI Data Naming Versus Rest and Rhyming Versus Rest Control group (see Fig. 2). Activation patterns for naming included the thalamus bilaterally. Areas activated unilaterally were the left inferior frontal gyrus (BAs 44 and 46), the left insular cortex, the left junction of the middle temporal gyrus, and the middle occipital gyrus (BA 19), the left inferior temporal gyrus (BA 37), the left posterior superior temporal gyrus (BA 42, 22), the left visual association areas (BA 19), and the right inferior frontal gyrus (BA 44). 179 SPEECH THERAPY AND fMRI IN AN APHASIC PATIENT TABLE 1 Common Activations for Control Group and RC in (a) Naming and (b) Rhyming Tasks: Stereotaxic Coordinates, Z Values, and Corresponding Brodmann Areas (BAs) for Regions Significantly Activated at Session 1 (S1) and Session 2 (S2) S1 Regions BA (k) x S2 y z Z BA (k) x y z Z ⫺42 ⫺50 50 ⫺50 ⫺46 ⫺14 12 0 6 ⫺24 ⫺56 ⫺20 25 5 10 10 ⫺5 15 3.76 3.01 3.40 3.41 3.13 3.90 (a). Common activations for control group and RC in naming Left inferior frontal Left insula Right insula Left superior temporal Left middle occipital Left thalamus Right thalamus BA 44 BA 42 BA19 (528) ⫺42 10 25 3.51 (94) (105) 32 ⫺50 10 ⫺22 10 10 3.30 3.50 (160) (494) ⫺14 28 ⫺20 18 15 10 4.10 3.40 BA 44 BA 42 BA19 (490) (290) (204) (111) (91) (189) (b). Common activations for control group and RC in Rhyming Left inferior frontal Right inferior frontal Right insula Left supra-marginalis Left inferior parietal Left middle occipital Left inferior occipital Right lingual Left thalamus Right thalamus BA 44 BA 45 BA 18 (512) (212) ⫺42 ⫺42 24 14 10 25 3.70 3.38 (299) 40 18 15 3.05 (64) ⫺38 ⫺72 ⫺5 2.52 BA 44 (426) ⫺46 14 25 3.98 BA 46 (206) (106) (111) (61) (54) (65) (142) (98) (60) 42 36 ⫺44 ⫺54 ⫺30 ⫺42 32 ⫺4 2 28 10 ⫺42 ⫺48 ⫺76 ⫺68 ⫺80 ⫺22 ⫺34 20 0 30 25 5 0 ⫺5 10 5 2.93 2.67 2.89 2.27 2.66 3.08 3.71 3.12 3.69 BA 40 BA 40 BA 19 BA 18 BA 18 Note. Coordinates are given in order x, y, z according to the atlas of Talairach and Tournoux (1988). The Z score is in italics and the voxel extent k is in bracket. The Z scores presented in this table are the mean Z scores of a cluster. Thresholds were P ⬍ 0.05 for peak height (uncorrected for multiple comparisons) and k ⬎ 50 for spatial extent. For Rhyming, activations were found bilaterally in the frontal opercula (BA 44). Areas activated unilaterally were the left supra-marginal gyrus (BA 40), left inferior temporal gyrus (BA 37, 19), left hippocampus, left thalamus, the right insular cortex, and the right visual association areas (BA 18). RC at Session 1 (see Fig. 2). For Naming, activations were located in the left inferior frontal gyrus (BA 44, 45), left cingular cortex (BA 24), the right superior temporal gyrus (BA 22), right supra-marginal gyrus (BA 40), and left and right visual association areas (BA 18, 19). For Rhyming, activations showed a bilateral pattern including the frontal opercula (BA 44, 45, 46), the precentral gyri (BA 6), and the visual association cortices (BA 18, 19) in addition to the left inferior temporal cortex (BA 37), the right angular gyrus (BA 39), and the right insular cortex. RC at Session 2 (see Fig. 2). For Naming, main contrast showed activations in the insular cortex, the superior temporal area (BA 22), and the supra-marginal gyrus (BA 40) and in the association visual cortex (BA 18) bilaterally. Activations restricted to the left hemisphere were found in Broca’s area (BA 44) and the thalamus. For Rhyming, activations were located in the left frontal operculum (BA 44), left inferior temporal gyrus (BA 37), and left supra-marginal gyrus (BA 40) and in the right insular cortex, the right angular gyrus (BA 39), and right association visual areas (BA 18). Common Activations for Control Group and RC before Therapy Naming (see Table 1). Common activations for Control group and RC S1 were found in the left frontal operculum (BA 44), the spared portion of the left superior temporal gyrus (BA 42), the right insular cortex, and the thalamus bilaterally. Rhyming (see Table 1). Common activations for Control group and RC s1 were found in left inferior frontal lobe (BA 44 – 45), left association visual area (BA 18), and right insular cortex. Common Activations for Control Group and RC after Therapy Naming (see Table 1). Areas that were activated by Control subjects and RC S2 revealed significant activations in the left thalamus, the left insular cortex, the left frontal operculum (BA 44), the spared portion of 180 LÉGER ET AL. TABLE 2 Comparison between RC after Therapy (RC S2) and RC before Therapy (RC S1): Stereotaxic Coordinates, Z scores, and Corresponding Brodmann Areas (BAs) for Regions Activated Significantly in (a) RC before (RC S1) and after Therapy (RC S2); (b) RC S2 but Not RC S1; (c) RC S1 but Not RC S2 in Naming and Rhyming Tasks Naming Regions activated BA (k) x Rhyming y z Z BA (k) x y z Z BA 46 BA 46 BA 6 BA 6 (116) (65) (216) (328) ⫺36 38 ⫺54 44 24 28 ⫺2 ⫺6 25 20 15 25 7.58 6.10 8.58 8.40 BA 39/19 (99) 28 ⫺60 30 8.82 BA 6 BA 19 BA 18 BA 31 (32) (37) (28) (120) ⫺48 ⫺34 6 20 ⫺2 ⫺80 ⫺70 ⫺76 30 10 0 10 6.74 6.44 7.65 8.72 (a) Common activations for RC S2 and RC S1 Left inferior frontal Right inferior frontal Left precentral Right precentral Right insula Left superior temporal Right superior temporal Right angular/superior Occipital BA 44 (84) ⫺48 10 15 8.24 BA 6 (124) ⫺54 0 15 8.65 BA19 BA 22 BA 22 (209) (60) (409) 52 ⫺32 52 4 ⫺40 ⫺48 5 15 15 8.35 6.33 9.17 (b) RC S2 but not RC S1 Left inferior frontal Left supra-marginalis BA 44 BA 40 (30) (27) ⫺50 ⫺36 8 ⫺46 25 35 6.07 7.80 (c) RC S1 but not RC S1 Left precentral Left middle occipital Right lingual Right cuneus BA 6 (33) ⫺52 0 30 9.38 BA 18 (32) 18 ⫺68 5 5.97 Note. Coordinates are given in order x, y, z according to the atlas of Talairach and Tournoux (1988). The Z score is in italics and the voxel extent k is in bracket. The Z scores presented in this table are the mean Z scores of a cluster. Thresholds were P ⬍ 0.05 for peak height (corrected for multiple comparisons) and k ⬎ 50 for spatial extent for (a) and k ⬎ 15 for (b), (c), and (d). the left superior temporal gyrus (BA 42), and the left association visual cortex (BA 19) as well as the right insular cortex. Rhyming (see Table 1). RC S2 and Control subjects activated in common bilaterally the inferior frontal gyrus (BA 44 – 46), the association visual areas (BA 18 –19), and the thalamus, in addition to the superior part of the left supra-marginal gyrus (BA 40), the left inferior parietal lobule (BA 40), and the right insular cortex. Comparison between RC S2 and RC S1 Naming (see Table 2). Common activations for RC at Sessions 2 and 1 were found in the superior temporal gyrus (BA 22) bilaterally, the left inferior frontal gyrus (BA 44), the left precentral gyrus (BA 6), and the right insular cortex. Regions activated by RC at Session 2 but not at Session 1 were found in the left inferior frontal gyrus (BA 44) and in the superior part of the left supramarginal gyrus (BA 40). Conversely, areas activated by RC at Session 1 but not at Session 2 were located in the upper part of the left precentral gyrus (BA 6), and in the right association visual areas (BA 18). Rhyming (see Table 2). Common activations between RC S1 and RC S2 were found mainly in the inferior frontal cortex (BA 46) and in the precentral gyrus (BA 6) bilaterally, and in the junction between the right angular gyrus and the superior occipital gyrus (BA 39/19). No difference was found in terms of areas significantly activated by RC at S2 but not at S1. The opposite contrast, revealing areas activated by RC at S1 but not at S2, showed activations in the left precentral gyrus (BA 6), and bilaterally in the association visual areas (left BA 19, right BA 18, and right BA 31). Speech-Therapy Effect. As noted under Materials and Methods, speech-therapy effects were assessed by contrasting Naming at the two sessions (NamingS2 ⫺ NamingS1 ⬎ 0), with an inclusive mask (P ⫽ 0.5) corresponding to a conjunction of NamingS1 ⬎ 0 and RhymingS2 ⫺ RhymingS1 ⬎ 0 in order to prevent for effects of deactivations in Naming and account for a possible task repetition. Analysis demonstrated speechtherapy-induced activations in the superior posterior part of the left supra-marginal gyrus (BA 40) (k extent ⫽ 18; coordinates x ⫽ ⫺34, y ⫽ ⫺44, z ⫽ 35; z score ⫽ 6.51) and in the upper part of Broca’s area (BA 44) (k extent ⫽ 28; coordinates: x ⫽ ⫺50, y ⫽ 8, z ⫽ 25; z score ⫽ 6.07) SPEECH THERAPY AND fMRI IN AN APHASIC PATIENT DISCUSSION The purpose of this study was to investigate in an aphasic patient the neurofunctional changes that accompany behavioral modifications after a language therapy, specifically devoted to speech output processing, even 2 years after stroke onset. We hypothesized that speech output improvement would be associated with peri-infarct activity in the left hemisphere. Aphasia Features and Speech Therapy Effects The major speech output deficit with phonemic errors, the severe deficit of auditory verbal working memory, and the sparing of word comprehension are in agreement with conduction aphasia profile (Goodglass, 1992). But, at variance with typical features of conduction aphasia, RC speech output appears unstable, effortful, and hesitating when attempting to repeat words. This speech output deficit could be compatible with a clinical profile of apraxia of speech (Hardcastle, 1987; Kent and Rosenbek, 1983). A recent study by Wise et al. (2001) emphasized the role of the left parieto-temporal junction, deep within the lateral sulcus, in speech production, and the infarction of this region in our patient could account for the speech production deficit we observed. RC’s deficit seemed to concern mainly planning and articulation of the syllable series that constitute spoken words. The speech therapy program performed in this study focused on this deficit and was based on the methods used for remediation of apraxia of speech (Pannbacker, 1988; Wambaugh et al., 1998). This program induced a significant improvement not only for trained items but also for nontrained stimuli, suggesting a general beneficial effect. Moreover, this positive therapeutic effect remained very stable with time, since 1 year after this experiment, RC obtained the same scores in the Naming tests. Functional Neuroimaging Results In control subjects the brain regions activated during the picture-naming task involved a large pattern including mainly left frontal, insular, and temporal/occipital areas in addition to the right inferior frontal cortex and to the left and right thalami and are consistent with many prior studies (Sergent et al., 1992; Bookheimer et al., 1995; Kosslyn et al., 1995; Damasio et al., 1996; Martin et al., 1996; Menard et al., 1996; Moore and Price, 1999; Murtha et al., 1996; Murtha et al., 1999). Considering the Rhyming task, there are many similarities between the regions activated by normal subjects in our study, and those described in other fMRI studies using the silent Rhyming task (Kareken et al., 2000; Lurito et al., 2000). These studies have shown bilateral but predominantly left-sided peri-sylvian ac- 181 tivations, particularly in Broca’s area (BA 44 – 45) and supra-marginal gyrus (BA 40). Activations for rhyming in RC are readily comparable to those found in control subjects since the patient’s performance was normal and stable between sessions. Two main findings emerged from these results. Despite close to ceiling performance on both sessions, the activation pattern observed after therapy in RC did not remain stable as a decrease of activity was noted at S2 compared to S1 in frontal areas bilaterally and right temporo-occipital areas. This finding might correspond to a test-retest effect that has been shown to induce diminished activations following practice, in normal subjects and in several experimental circumstances (e.g., Raichle et al., 1994; Carel et al., 2000). In the present study it is difficult to reach a definite conclusion on this matter, as we did not assess testretest effects in the control subjects. Moreover, the rhyming pattern of activation in RC at S2 tended to resemble that observed in normal subjects as the number of regions that are activated in common by RC and control subjects increased. In particular, a common activation was found in the (spared) superior part of the left supra-marginal gyrus. The role of this region for phonological processing in control subjects has been emphasized in several studies (Démonet et al., 1996; Paulesu et al., 1993; Salmon, 1996; Schumacher, 1996). Since RC showed normal performance on this task even at Session 1, one may speculate that the left supra-marginal gyrus, at least its inferior part, was not necessary to this Rhyming task for the patient (Price et al., 1999). Nevertheless improvement of naming after speech therapy might have eased the Rhyming task for RC who could resort to less effortful strategies possibly associated with close-to-normal pattern of activation. For Naming, the neuro-functional results in RC lend support to the already noted hypothesis of a key role for left peri-lesional areas in the mechanisms of recovery from aphasia. Indeed, after therapy the “Naming– rest” contrast revealed that, in addition to frontal and temporal regions activated before therapy, left-sided regions surrounding the lesion became activated, namely the anterior insular cortex, the middle portion of the superior temporal gyrus, the superior part of the supra-marginal gyrus. The left anterior insula has been implicated in speech production and planning both in lesion studies (Dronkers, 1996) and in functional neuro-imaging studies (Wise et al., 1999) and this region might play a role in our experiment as speech therapy in RC enhanced articulatory planning in speech production. The activation found in the superior temporal gyrus, a region adjacent to the posterior part of the damaged tissues, is congruent with previous reports on good recovery associated with sparing of this region (Selnes et al., 1985; Naeser et al., 1987, 1990; Heiss et al., 182 LÉGER ET AL. 1999). Nevertheless, activation in homologous right temporal regions was also observed in RC in both sessions. The role of the right hemisphere in aphasic patients while recovering has long been underlined and discussed. Its impact may vary across time (Knopman et al., 1983) and language functions (Musso et al., 1999). In the present study, the activity in the right temporal cortex does not seem essential for recovery of Naming as it was present on the first session associated with poor performance. However a complete rightto-left functional shift was not observed even after therapeutic performance improvement and this incomplete shift in RC might correspond to an unachieved language recovery (Belin et al., 1996). Among the left-sided set of regions activated in Naming after therapy, the comparison for Naming between the two sessions in RC evidenced the importance of the upper parts of Broca’s area and left supra-marginal gyrus since these regions remained activated even when possible repetition effects were taken into account. By comparison to activation seen before therapy in Broca’s area, activation in this region after therapy spread toward the upper part of pars opercularis which might be associated with sublexical output processing (Paulesu et al., 1997). The left supra-marginal gyrus was found activated for Naming only in RC after therapy and it is noticeable that the parietal region does not belong to the normal neuro-functional pattern for Naming tasks as shown by the present study as well as most of studies analyzed by Murtha et al. (1999). Together with Broca’s area, the left supra-marginal gyrus has long been associated with phonological processing and phonological working memory. More specifically a meta-analysis from Démonet and colleagues (1996) emphasized the role of the supra-marginal gyrus in phonological storage whereas the superior part of Broca’s area was linked to verbal rehearsal for tasks requiring phonological awareness and involving verbal working memory. The speech therapy program used in RC purposefully resorted to enhancement of phonological and articulatory awareness and the specific activations of Broca’ s area and left supra-marginal gyrus are likely to reflect the reinforcement of phonological strategies in RC while Naming. In conclusion this study reflects changes in left-sided cortical activity associated with language output improvement after intensive speech therapy at a late stage. At variance with studies showing enlargement of cortical representations in the sensorimotor cortex in studies of motor recovery (Chollet, 2000), the activation pattern observed during Naming after therapy in RC does not recruit the same network as that observed in control subjects. As already noted, a limitation of this study is the absence of test-retest assessment in control subjects. However, this effect generally consists of decreases of activation whereas the therapy effect in RC is associated with alterations of activation patterns including an increase of activity in some regions. Such altered patterns might correspond to the influence of the sublexical, phonological strategy that has been implemented by a dedicated program of speech therapy. ACKNOWLEDGMENTS We thank Patrice Peran and Sandra Lé for their helpful assistance. This work was supported by grants from the French Cognitique Program-MENRT n° 1A012F (1999), and from the CEE Grant QLK6-CT-1999-02140. REFERENCES Belin, P., Van Eeckhout, P., Zilbovicius, M., Remy, P., Francois, C., Guillaume, S., et al. 1996. Recovery from nonfluent aphasia after melodic intonation therapy: A PET study. Neurology 47: 1504–1511. Bookheimer, S. Y., Zeffiro, T. A., Blaxton, T., Gaillard, W., and Theodore, W. 1995. Regional cerebral blood flow during object naming and word reading. Hum. Brain Mapp 3: 93–106. Buckner, R. L., Bandettini, P. A., O’Craven, K. M., Savoy, R. L., Petersen, S. E., Raichle, M. E., et al. 1996. Detection of cortical activation during averaged single trials of a cognitive task using functional magnetic resonance imaging. Proc. Natl. Acad. Sci. USA 93: 14878 –14883. Cardebat, D., Demonet, J. F., Celsis, P., Puel, M., Viallard, G., and Marc-Vergnes, J. P. 1994. Right temporal compensatory mechanisms in a deep dysphasic patient: A case report with activation study by SPECT. Neuropsychologia 32: 97–103. Carel, C., Loubinoux, I., Boulanuar, K., Manelfe, C., Rascol, O., Celsis, P., and Chollet, F. 2000. Neural substrate for the effects of passive training on sensory motor cortical representation: A study with functional magnetic resonance imaging in healthy subjects. J. Cereb. Blood Flow Metab. 20: 478 – 484. Chollet, F. 2000. Plasticity of the adult human brain. In Brain Mapping: The Systems, pp. 621– 638. Academic Press, San Diego. Damasio, H., Grabowski, T. J., Tranel, D., Hichwa, R. D., and Damasio, A. R. 1996. A neural basis for lexical retrieval. Nature 380: 499 –505. De Renzi, E., and Vignolo, L. 1962. The token test: A sensitive test to detect receptive disturbances in aphasics. Brain 85: 665– 678. Deloche, G., Hannequin, D., Dordain, M., Metz-Lutz, M. N., Kremin, H., Tessier, C., et al. 1997. Diversity of patterns of improvement in confrontation naming rehabilitation: Some tentative hypotheses. J. Commun. Disord. 30: 11–21; quiz 21–22. Démonet, J. F., Fiez, J. A., Paulesu, E., Petersen, S. E., and Zatorre, R. J. 1996. PET Studies of phonological processing: A critical reply to Poeppel. Brain Lang. 55: 352–379. Dronkers, N. F. 1996. A new brain region for coordinating speech articulation. Nature 384: 159 –161. Duvernoy, H. M. 1992. Le cerveau humain. Springer-Verlag, Paris. Elman, R. J., and Bernstein-Ellis, E. 1999. The efficacy of group communication treatment in adults with chronic aphasia. J. Speech Lang. Hear. Res. 42: 411– 419. Goodglass, H. 1992. Diagnostic of conduction aphasia. In Conduction Aphasia (S. E. Kohn, Ed.), pp. 39 – 49. Erlbaum, Hillsdale, NJ. Hardcastle, W. J. 1987. Electropalatographic study of articulation disorders in verbal dyspraxia. In Phonetic Approaches to Speech Production in Aphasia and Related Disorders (J. H. Ryalls, Ed.), College-Hill Press, Boston. Heiss, W. D., Kessler, J., Thiel, A., Ghaemi, M., and Karbe, H. 1999. Differential capacity of left and right hemispheric areas for compensation of poststroke aphasia. Ann. Neurol. 45: 430 – 438. SPEECH THERAPY AND fMRI IN AN APHASIC PATIENT Holland, A. L., Fromm, D. S., DeRuyter, F., and Stein, M. 1996. Treatment efficacy: Aphasia. J. Speech Hear. Res. 39: S27–S36. Holmes, A. P., and Friston, K. J. 1998. Generalisability, random effects and population inference. NeuroImage 7: S754. Kareken, D. A., Lowe, M., Chen, S. H., Lurito, J., and Mathews, V. 2000. Word rhyming as a probe of hemispheric language dominance with functional magnetic resonance imaging. Neuropsychiatry Neuropsychol. Behav. Neurol. 13: 264 –270. Katz, R. C., and Wertz, R. T. 1997. The efficacy of computer-provided reading treatment for chronic aphasic adults. J. Speech Lang. Hear. Res. 40: 493–507. Kent, R. D., and Rosenbek, J. C. 1983. Acoustic patterns of apraxia of speech. J. Speech Hear. Res. 26: 231–249. Knopman, D. S., Rubens, A. B., Selnes, O. A., Klassen, A. C., and Meyer, M. W. 1984. Mechanisms of recovery from aphasia: Evidence from serial xenon 133 cerebral blood flow studies. Ann. Neurol. 15: 530 –535. Kosslyn, S. M., Alpert, N. M., and Thompson, C. K. 1995. Identifying objects at different levels of hierarchy: A positron emission tomography study. Hum. Brain Mapp. 3: 107–132. Kosslyn, S. M., Alpert, N. M., Thompson, W. L., Chabris, C. F., Rauch, S. L., and Anderson, A. K. 1994. Identifying objects seen from different viewpoints. A PET investigation. Brain 117: 1055– 1071. Lincoln, N. B., McGuirk, E., Mulley, G. P., Lendrem, W., Jones, A. C., and Mitchell, J. R. 1984. Effectiveness of speech therapy for aphasic stroke patients. A randomised controlled trial. Lancet 1: 1197– 1200. Lurito, J. T., Kareken, D. A., Lowe, M. J., Chen, S. H., and Mathews, V. P. 2000. Comparison of rhyming and word generation with FMRI. Hum. Brain Mapp. 10: 99 –106. Martin, A., Wiggs, C. L., Ungerleider, L. G., and Haxby, J. V. 1996. Neural correlates of category-specific knowledge. Nature 379: 649 – 652. Menard, M. T., Kosslyn, S. M., Thompson, W. L., Alpert, N. M., and Rauch, S. L. 1996. Encoding words and pictures: A positron emission tomography study. Neuropsychologia 34: 185–194. Moore, C. J., and Price, C. J. 1999. Three distinct ventral occipitotemporal regions for reading and object naming. NeuroImage 10: 181–192. Murtha, S., Chertkow, H., Beauregard, M., Dixon, R., and Evans, A. 1996. Anticipation causes increased blood flow to the anterior cingulate cortex. Hum. Brain Mapp. 4: 103–112. Murtha, S., Chertkow, H., Beauregard, M., and Evans, A. 1999. The neural substrate of picture naming. J. Cogn. Neurosci. 11: 399 – 423. Musso, M., Weiller, C., Kiebel, S., Muller, S. P., Bulav, P., and Rijntjes, M. 1999. Training-induced brain plasticity in aphasia. Brain 122: 1781–1790. Naeser, M. A., Gaddie, A., Palumbo, C. L., and Stiassny-Eder, D. 1990. Late recovery of auditory comprehension in global aphasia. Improved recovery observed with subcortical temporal isthmus lesion vs Wernicke’s cortical area lesion. Arch. Neurol. 47: 425– 432. Naeser, M. A., Helm-Estabrooks, N., Haas, G., Auerbach, S., and Srinivasan, M. 1987. Relationship between lesion extent in ‘Wernicke’s area’ on computed tomographic scan and predicting recovery of comprehension in Wernicke’s aphasia. Arch. Neurol. 44: 73– 82. Ohyama, M., Senda, M., Kitamura, S., Ishii, K., Mishina, M., and Terashi A. 1996. Role of the nondominant hemisphere and undamaged area during word repetition in poststroke aphasics. A PET activation study. Stroke 27: 897–903. 183 Pannbacker, M. 1988. Management strategies for developmental apraxia of speech: A review of literature. J. Commun. Disord. 21: 363–371. Paulesu, E., Frith, C. D., and Frackowiak, R. S. 1993. The neural correlates of the verbal component of working memory. Nature 362: 342–345. Paulesu, E., Goldacre, B., Scifo, P., Cappa, S. F., Gilardi, M. C., Castiglioni, I., Perani, D., and Fazio, F. 1997. Functional heterogeneity of left inferior frontal cortex as revealed by fMRI. Neuroreport 8: 2011–2017. Price, C. J., Mummery, C. J., Moore, C. J., Frakowiak, R. S., and Friston, K. J. 1999. Delineating necessary and sufficient neural systems with functional imaging studies of neuropsychological patients. J. Cogn. Neurosci. 11: 371–382. Pulvermuller, F., Neininger, B., Elbert, T., Mohr, B., Rockstroh, B., Koebbel, P., et al. 2001. Constraint-induced therapy of chronic aphasia after stroke. Stroke 32: 1621–1626. Raichle, M. E., Fiez, J., Videen, T., Macleod, A., Pardo, J., Fox, P., and Petersen, S. 1994. Practice-related changes in human brain functional anatomy during nonmotor learning. Cereb. Cortex 4: 8 –26. Robey, R. R. 1994. The efficacy of treatment for aphasic persons: A meta-analysis. Brain Lang. 47: 582– 608. Salmon, E., Van der Linden, M., Gillette, F., Delfiore, G., Maquet, P., Degueldre, C., Luxen, A., and Franck, G. 1996. Regional brain activity during working memory tasks. Brain 119: 1617–1625. Selnes, O. A., Knopman, D. S., Niccum, N., and Rubens, A. B. 1985. The critical role of Wernicke’s area in sentence repetition. Ann. Neurol. 17: 549 –557. Sergent, J., Ohta, S., and MacDonald, B. 1992. Functional neuroanatomy of face and object processing. A positron emission tomography study. Brain 115: 15–36. Small, S. L., Flores, D. K., and Noll, D. C. 1998. Different neural circuits subserve reading before and after therapy for acquired dyslexia. Brain Lang. 62: 298 –308. Snodgrass, J. G., and Vanderwart, M. 1980. A standardized set of 260 pictures: Norms for name agreement, image agreement, familiarity, and visual complexity. J. Exp. Psychol. [Hum. Learn.] 6: 174 –215. Talaraich, P., and Tournoux, J. 1988. A Stereotactic Coplanar Atlas of the Human Brain. Thieme, Stuttgart. Thompson, C. K. 2000. The neurobiology of language recovery in aphasia. Brain Lang. 71: 245–248. Thulborn, K. R., Carpenter, P. A., and Just, M. A. 1999. Plasticity of language-related brain function during recovery from stroke. Stroke 30: 749 –754. Wambaugh, J. L., Kalinyak-Fliszar, M. M., West, J. E., and Doyle, P. J. 1998. Effects of treatment for sound errors in apraxia of speech and aphasia. J. Speech Lang. Hear. Res. 41: 725–743. Warburton, E., Price, C. J., Swinburn, K., and Wise, R. J. 1999. Mechanisms of recovery from aphasia: Evidence from positron emission tomography studies. J. Neurol. Neurosurg. Psychiatry 66: 155–161. Weiller, C., Isensee, C., Rijntjes, M., Huber, W., Muller, S., Bier, D., et al. 1995. Recovery from Wernicke’s aphasia: A positron emission tomographic study. Ann. Neurol. 37: 723–732. Wise, R. J., Greene, J., Buchel, C., and Scott, S. K. 1999. Brain regions involved in articulation. Lancet 353: 1057–1061. Wise, R. J. S. S., Blank, S. C., Mummery, C. J., Murphy, K., and Warburton, E. A. 2001. Separate neural subsystems within ‘Wernicke’s area’. Brain 124: 83–95.