Role of the Nondominant Hemisphere and Undamaged Area During Word Repetition in Poststroke Aphasics
A PET Activation Study
Background and Purpose Although the resting regional cerebral blood flow (rCBF) in aphasic patients has been thoroughly investigated with positron emission tomography (PET) and single-photon emission CT, and PET studies in normal subjects have elucidated the functional localization of language processing, little is known about the activation pattern of language processing in aphasic patients.
Methods We measured the changes in rCBF during a repetition task (hearing a single word and repeating it aloud) and the resting state using the H215O PET activation technique in 6 normal subjects (mean±SD age, 58.3±8.1 years) and 16 aphasic patients: 10 fluent aphasics (age, 60.3±12.5 years) and 6 nonfluent aphasics (age, 50.5±8.3 years).
Results In normal subjects, the posteroinferofrontal area (PIF) including Broca’s area, the posterosuperotemporal area (PST) including Wernicke’s area, the rolandic areas, and a few other areas were activated with left side dominance by the repetition task. In the resting state, the rCBF in the left PIF and the left posterotemporal area was reduced in both fluent and nonfluent aphasics. In aphasic patients, the magnitude of activation in the right PIF and PST by the repetition task was greater than in normal subjects. The increase in rCBF during the repetition task in the left PIF correlated with the Western Aphasia Battery score of spontaneous speech in the nonfluent aphasics with a left inferofrontal lesion.
Conclusions This study shows the importance in aphasic patients of the mirror regions of the left PIF and PST in the nondominant (right) hemisphere for performing the word repetition task. The results also show the importance for nonfluent aphasic patients of the recruitment of the undamaged PIF for spontaneous speech.
Aphasia frequently occurs in patients with infarcts and often ruins their social quality of life. Investigations of language and aphasia have been conducted with lesion analysis, focusing on the relationship between the damaged brain localization and resulting aphasic symptoms. Those studies on language disorders due to stroke provided the basis for the classic 19th century model of language localization. Since then, analysis of living patients has been performed by morphological examination with x-ray CT1 and MRI. However, the symptoms of aphasia are variable and were not always predicted by the morphological features observed by x-ray CT or MRI.2 In contrast, PET and single-photon emission CT have revealed the functional suppression occurring in morphologically intact regions as a remote effect (eg, diaschisis) and related it to the aphasic symptoms.2 3 While these investigations have been performed with patients in the resting state, different approaches have been taken to investigate the working brain during psychological tasks. Findings in the dichotic listening test suggested the importance of the right hemisphere in recovered aphasic patients.4
The brain PET activation technique allows repeated measurement of rCBF under different conditions and has been used to reveal the regional cortical neural activity engaged in language processing in the living human brain. Although a large number of studies have been performed on normal subjects to investigate the functional anatomy involved in various aspects of language processing,5 6 7 8 9 10 11 few have been conducted on aphasic patients with lesions in language-relevant areas.12 13 Heiss and colleagues12 reported that the persisting structure in the left hemisphere played an important role in the recovery from aphasia. However, it is not clear how the undamaged area and the nondominant hemisphere compensate for lost function and how this is related to aphasic symptoms in mildly or moderately aphasic patients.
To investigate the difference in the activation pattern of rCBF between normal and poststroke aphasics, we used PET activation techniques to measure the changes in rCBF associated with simple language processing in normal subjects and aphasic patients.
Subjects and Methods
We selected a total of 16 aphasic Japanese patients who had suffered a single cerebral infarction in the left hemisphere and had recovered to the extent that they could at least repeat words. The patients were studied at least 1 month after the onset of the stroke. Each subject underwent the WAB within 10 days of the PET study. The type of aphasia was determined at 1 month after the stroke onset and at the time of the PET study by two neurologists and two speech therapists. The aphasic patients included 12 men and 4 women (mean±SD age, 56.6±11.8 years), 10 fluent (10 men; age, 60.3±12.5 years) and 6 nonfluent (2 men and 4 women; age, 50.5±8.3 years), who were all strongly right-handed, as confirmed by the Edinburgh Handedness Inventory (lateral quotient ≥89).14 The clinical profile of the patients is shown in Table 1⇓. MRIs were taken of each patient to obtain morphological information. The volume of infarction was estimated by the number of pixels with signal intensity abnormality on the T2-weighted MRI with the use of a computerized user-driven tracing routine.15
The control subjects were 6 normal subjects (5 men and 1 woman; age, 58.3±8.1 years) without known neurological disorders. They were all strongly right-handed, as confirmed by the Edinburgh Handedness Inventory (lateral quotient ≥89).14 MRIs were taken of each subject to confirm the absence of silent lesions and to provide an anatomic reference for the data analysis.
Written informed consent was obtained from each normal subject and from each aphasic patient.
Each subject underwent two CBF measurements during each of three different conditions (repetition, resting, and counting). In the resting state, the subject’s ears were left uncovered, and the eyes were fixed on a white dot on the television monitor in front of the PET camera. In the repetition task, the auditory stimulus was a series of words (eg, clock and orange, in the native language) delivered every 2.5 seconds binaurally from earphones, and the subject was instructed to repeat the word aloud. The task performance was checked by a neurologist. In the counting task, the subject was instructed to count the numbers aloud from 1 to 10 repeatedly. The counting task data were used only in the statistical analysis to determine the activated areas in the normal subjects. The results of the counting task will be published in another report.
CBF was measured with an intravenous bolus injection of 1.5 GBq of H215O and the PET autoradiographic method16 on a Headtome IV tomograph. This system simultaneously acquires 14 parallel slices with a center-to-center interslice distance of 6.5 mm. The images were reconstructed with a spatial resolution of 7.0 mm full width at half maximum. Each task began 60 seconds before the start of data accumulation and continued throughout the 120-second data acquisition that started at the injection of H215O. A series of six blood flow scans was performed in random order for each subject, twice for each of the three tasks, with a 15-minute interscan interval. The arterial blood was sampled, and the parametric CBF images were created. The data were analyzed by the image processing system Dr View (Asahi Kasei Co Ltd) on Indigo2/Indy (Nihon Silicon Graphics Inc) computers.
Intersubject Averaging Analysis
The data of the normal subjects were analyzed with a stereotaxic intersubject averaging analysis to determine the activated foci in the normal brain. The PET images of each subject were linearly transformed to the Talairach frame,17 18 which is defined on the line connecting the anterior and posterior commissures. The anterior and posterior commissures were identified on the sagittal MRI, and their locations in the PET coordinates were calculated by matching the PET and MRI. The brain images were reoriented to the anterior and posterior commissures, and the size of the brain was proportionally adjusted to the atlas in each axis. An intersubject statistical analysis was performed pixel by pixel on the CBF values normalized by the gCBF, using a general linear model in the IMSL library (a two-way ANOVA model), as follows:
where α and β are the subject effect and the task effect, respectively; i=subject, j=task, and k=trial; and Error is an independently distributed normal error with unknown variance. The estimate (mean ΔCBF) and the t value of “βrepetition minus βrest” were mapped. The contrast of “repetition minus rest” was considered to be associated with the word repetition process.
An omnibus test was performed at P=.002, which is equivalent to t=3.136 in the t map, to see whether the number of pixels with t>3.136 was significantly larger than the chance level; possible activation foci with a peak above the threshold were described. Next, significant foci in specific locations were determined by the multiple comparison method of Friston et al19 with measured smoothness at a significance level of P=.05. This method uses a stochastic model to describe the t image and, although statistically conservative, allows the demonstration of significant foci by multiple comparison with a Bonferroni correction, taking into account the autocorrelation between nearby pixels.
The PET CBF images were registered to the subject’s MRI three dimensionally with the use of software by Senda.20 The PET images were superimposed on the MRI to interpret them in terms of the subject’s own anatomy and to provide information for ROI placement. For the rCBF images of each subject, circular ROIs of 12 mm diameter were drawn on the language-relevant areas in the same manner as that of Metter and colleagues.21
The subtraction (ΔCBF) images were created by subtracting resting from repetition CBF images, which had been normalized to gCBF=50 mL/min per 100 mL, pixel by pixel in each subject. The subtraction images were superimposed on the MRI. As the bilateral PIF, PST, rolandic area, and supplementary motor area were activated in normal subjects in the intersubject averaging analysis (see “Results”), a circular ROI of 12 mm diameter was drawn on the peak within each of these areas in the subtraction images superimposed on the MRI of each subject, and the magnitude of activation was calculated and analyzed.
For the ROI analysis, ANOVA was conducted to compare the normal subjects with the fluent and nonfluent aphasic patients for each region and for each activation focus. Spearman ranked correlation analysis was used to examine the relationship between the WAB scores and the resting rCBF value or the rCBF increase at the activation foci.
Intersubject Analysis on Normal Subjects
Fig 1⇓ illustrates the t map images regarding the contrast between repetition and resting, which were derived from the intersubject averaging statistical analysis on the normal subjects and superimposed on the Talairach grid. An omnibus test indicated that the number of pixels above the t threshold (t>3.136) was significantly greater than the chance level (P<.0001). When specific foci were searched for by the method of Friston et al,19 the following areas were observed to have been significantly activated: the bilateral PIF corresponding to BA 44 and 6 including Broca’s area, the bilateral PST (BA 42, 22) including Wernicke’s area, the left posterotemporal area (BA 37), the bilateral rolandic area (BA 3, 4), the supplementary motor area (BA 6), and the right cerebellar hemisphere (Table 2⇓ and Fig 1⇓).
As the clinical profile of the patients shown in Table 1⇑ describes, the aphasia quotient was relatively good (mean±SD, 74.3±12.2) in the aphasic patients in this study. Their task performance was good, and phonemic mistakes were observed in fewer than 4 of the 48 words tested in the repetition task for each patient. No significant negative correlation was observed between any of the WAB scores and the volume of infarction measured in the MRI by Spearman ranked correlation analysis.
The resting rCBF in fluent aphasics was significantly reduced in the left PIF, left posterotemporal, left high frontal, and left parietal areas, as well as in the left hemispheric mean value compared with normal control subjects, while the resting rCBF in nonfluent aphasics was reduced in the left PIF and left high frontal area (Table 3⇑). The resting rCBF in the left posterotemporal area was reduced in nonfluent aphasics, although not significantly so, compared with normal control subjects. In the resting state, a significant correlation was observed between the absolute value of rCBF in the left PST and the comprehension score (rs=.596, P=.021<.05) in the Spearman ranked correlation analysis.
The magnitude of activation by the repetition task in the right PIF and the right PST was greater in both fluent and nonfluent aphasic patients than the corresponding value in normal subjects. Furthermore, the magnitude of activation in the right PIF in nonfluent aphasics was greater than that in fluent aphasics (Table 4⇑).
In contrast, in normal subjects the repetition-induced rCBF change at the activation foci within the language-relevant areas was more pronounced on the left side than on the right side. However, the resting rCBF in normal subjects showed no laterality in those areas.
A typical patient with nonfluent aphasia (patient 15) is shown in Fig 2⇓. The patient is a 44-year-old woman with nonfluent-type aphasia due to cerebral infarction in the left PIF cortex and extending to the white matter. The aphasia quotient in this patient was relatively low (67.2) among the 16 patients. In the repetition task, the rCBF increase in the right PIF was much greater (28.0.% increase) than in normal subjects (mean±SD, 11.8±3.0%).
Spearman ranked correlation analysis revealed that the increase in rCBF in the left PIF correlated with the WAB score of spontaneous speech in aphasic patients as a whole (rs=.648, P=.012<.05) and in the nonfluent group (rs=.928, P=.038<.05) (Fig 3⇓). In the right hemisphere, there was no correlation between any WAB score and the magnitude of activation, except the weak correlation among nonfluent aphasics (n=6) between the increase in rCBF in the right PIF and their repetition score (WAB) (rs=.812, P=.069) by Spearman ranked correlation analysis.
In this study the language-relevant areas (Broca’s and Wernicke’s areas) and their contralateral areas were significantly activated during the repetition task in normal subjects, as revealed by the intersubject averaging analysis. The PIF, PST, and rolandic area (related to the mouth and lips) were activated with a dominance in the left hemisphere during the repetition task. Previous studies also showed a dominance in the left hemisphere regarding speech tasks.5 22
Within the PIF, BA 44 was activated during repetition, but BA 45 was not. The repetition task was designed to detect the language-relevant areas involved in the entire process from sensory (auditory) input to motor (spoken) output. Thus, the classic speech centers (BA 44, 42, and 22) and rolandic areas were activated as expected.
The PST, including Wernicke’s area, was activated with a dominance in the left hemisphere. Wise et al9 reported that the semantic area (Wernicke’s area) was activated even by hearing familiar words without an effort to understand the meaning. Margolin23 suggested that three different routes exist in the information-processing model of speech production, including repetition and conversational speech. The repetition task in the present study might activate the three routes defined in Margolin’s neuropsychological model: the lexical phonological route, the semantic route, and the nonlexical phonological route. When the subjects repeat meaningful words, they unintentionally understand the meaning of the words. Thus, the repetition task stimulates not only the hearing and the articulation processing but also the low-level semantic processing, which might explain the left dominance.
We used the repetition task to investigate language processing in aphasic patients. The hierarchical tasks used in the previous PET activation studies on the functional anatomy of language processing5 6 were designed to detect the localization of the specific language functions in normal subjects and are difficult tasks for patients with language disorders. The language task was simplified in the present study so that the aphasic patients could fully perform the task. This allows a clear interpretation of the magnitude of rCBF change initiated by the task.
The resting rCBF in the left PIF and in the left posterotemporal area was reduced in both fluent and nonfluent aphasics. The area with an rCBF reduction in aphasics due to focal cerebrovascular lesions is generally much broader than the morphological lesions delineated in the x-ray CT or MRI, as exemplified in one of our patients (Fig 2⇑). Metter et al2 reported that the remote metabolic effects within the speech area might explain the symptoms of aphasia. These remote effects are attributed to a neuroanatomical relation known as fasciculus articulate. Hemodynamic disorders may also contribute to the symptom of aphasia.
In the present study the magnitude of activation in the right PIF and right PST induced by the repetition task was greater in both fluent and nonfluent aphasic patients than in normal subjects. The significance of the nondominant hemisphere in language processing has been discussed,24 25 and the role of the right hemisphere in the recovery from aphasia has been described.26 In a previous study’s dichotic listening tests, right hemispheric activation was shown during language processing to a larger extent in the patients who had recovered from aphasia than in nonaphasic patients or normal subjects, suggesting interhemispheric reorganization or interhemispheric transfer of language function in the recovery from aphasia.3 The results of the studies in which evoked potential paradigms were used supported the concept of right hemisphere contribution in aphasic patients.27 A study in which intracarotid amobarbital injection technique was used showed speech arrest after a right-side injection but not after a left-side injection in recovered aphasic patients.28 These reports support the significance of the right PIF and PST in the processing of repetition in fluent and nonfluent aphasics observed in the present study and the possibility of functional redistribution or reorganization from the left hemisphere into the right hemisphere in our patients.
There was a disproportionate representation of sex in aphasic patients and normal subjects in the present study. Recent studies in which functional MRI29 was used in normal subjects have shown a different pattern of hemispheric functional asymmetry under verbal activation conditions in women, who exhibit a greater degree of bilateral activation compared with men, whose activation tends to be unilateral. In the present study’s group of nonfluent aphasics, the magnitude of activation in the right PIF in women was greater than that in men (women [n=4], 25.8±5.20%; men [n=2], 17.3±3.25%), and the magnitude of activation in the right PST in women was not much different from that of men (women [n=4], 20.7±3.91%; men [n=2], 24.3±0.42%). However, we cannot apply a statistical approach to such a limited number of subjects. In the group of fluent aphasics, which included no women, the activation of the right PIF and PST was significantly greater than that of the left PIF and PST, suggesting a functional reorganization of the right hemisphere to compensate for the left hemisphere, rather than a sex-related difference in verbal function. Furthermore, since there was a weak correlation among nonfluent aphasics (n=6) between the right PIF activation and the repetition score (WAB) (rs=.812, P=.069), their greater activation in the right PIF and PST suggests the compensation of the left hemispheric function, even if some effect of the sex-related difference is present.
The role of the undamaged area in the left language-relevant cortex has also been discussed as a contributor to the recovery from aphasia.12 According to the results of the present study, the increase in the rCBF by the repetition task in the left inferofrontal area of the patients with nonfluent aphasia was significantly correlated with their WAB score of spontaneous speech. The left PIF has been considered an area associated with the motor speech function. Zatorre et al10 reported that the discrimination of phonetic structure led to increased activity in BA 44 and 6, suggesting their role in articulatory recording for phonetic perception. Our present results suggest the importance, for the spontaneous speech of aphasic patients, of the recruitment of the undamaged area surrounding the lesion within the PIF. In addition, a significant correlation was observed between the absolute value of the resting rCBF in the left PST and the comprehension score, suggesting the importance of the undamaged structure in the dominant hemisphere.
No significant negative correlation was observed between the volume of infarction in the left hemisphere and any of the WAB scores. This might be due to the lack of patients with large-scale infarcts in this study population. It also suggests that the morphological index is less predictive of the patient’s state than the functional indicators as measured in this study.
As for the methodological aspect, determination of the activation foci in individual patients is very difficult because the limited number of measurements hinders a statistically significant outcome. The intersubject analysis method could not be applied to the aphasic patients of the present study because of the large individual variation among the patients in both morphology and functional anatomy. In the present study we placed ROIs on the “hot spots” observed within the anatomic areas that were shown to be activated in the intersubject averaging statistical analysis of normal subjects. This is based on the assumption that those areas involved in a task in normal subjects should play a role, more or less, when an aphasic patient performs the same task. If a totally different region has become involved as a result of functional reorganization, it is difficult to affirm its significance even if we can observe it as a hot spot.
In conclusion, our investigation demonstrates the importance of the mirror regions of the left PIF and PST in the nondominant (right) hemisphere of recovering aphasic patients for performing the word repetition task with functional redistribution or reorganization. It also shows the importance of the recruitment of the undamaged PIF for spontaneous speech in nonfluent aphasic patients. A PET activation study with a simple language processing task is useful to activate the language-relevant regions and can be applied to mild or moderate aphasics to investigate the relationship between clinical symptoms and the rCBF response to language processing.
Selected Abbreviations and Acronyms
|CBF||=||cerebral blood flow|
|gCBF||=||global cerebral blood flow|
|PET||=||positron emission tomography|
|rCBF||=||regional cerebral blood flow|
|ROI||=||region of interest|
|WAB||=||Western Aphasia Battery|
We thank neurologists Dr Yuich Komaba and Dr Masashi Yanagisawa for valuable advice and criticism; speech therapists Midori Mori and Hideko Akiyama for performing the WAB; and Dr Hinako Toyama, Dr Kiichi Ishiwata, Dr Toru Sasaki, Keiichi Oda, and Shin-ichi Ishii for rCBF measurement and data analysis.
Reprint requests to Masashi Ohyama, Second Department of Internal Medicine, Nippon Medical School, 3-5-5, Iidabashi, Chiyoda, Tokyo 102, Japan. E-mail email@example.com.
- Received November 20, 1995.
- Revision received January 29, 1996.
- Accepted February 22, 1996.
- Copyright © 1996 by American Heart Association
Alexander MP, Naeser MA, Palumbo C. Broca’s area aphasia: aphasia after lesions including the frontal operculum. Neurology. 1990;40:353-362.
Karbe H, Herholz K, Szelies B, Pawlik G, Wienhard K, Heiss WD. Regional metabolic correlates of Token test results in cortical and subcortical left hemispheric infarction. Neurology. 1989;39:1083-1088.
Posner MI, Petersen SE, Fox PT, Raichle ME. Localization of cognitive operations in the human brain. Science. 1990;240:1627-1631.
Howard D, Patterson K, Wise R, Brown WD, Friston K, Weiller C, Frackowiak R. The cortical localization of the lexicons: positron emission tomography evidence. Brain. 1992;115:1769-1782.
Wise R, Chollet F, Hadar U, Friston K, Hoffner E, Frackowiak R. Distribution of cortical neural networks involved in word comprehension and word retrieval. Brain. 1991;114:1803-1817.
Zatorre RJ, Evans AC, Meyer E, Gjedde A. Lateralization of phonetic and pitch discrimination in speech processing. Science. 1992;256:846-849.
Demonet JF, Chollet F, Ramsay S, Cardebat D, Nespoulous JL, Wise R, Rascol A, Frackowiak R. The anatomy of phonological and semantic processing in normal subjects. Brain. 1992;115:1753-1768.
Graham GD, Kalvach P, Blamire AM, Brass LM, Fayad PB, Prichard JW. Clinical correlates of proton magnetic resonance spectroscopy findings after acute cerebral infarction. Stroke. 1995;26:225-229.
Herscovitch P, Markham J, Raichle ME. Brain blood flow measured with intravenous H215O, I: theory and error analysis. J Nucl Med. 1983;24:782-789.
Talairach J, Tournoux P. Co-Planar Stereotaxic Atlas of the Human Brain: 3-Dimensional Proportional System: An Approach to Cerebral Imaging. Stuttgart, Germany: Georg Thieme Verlag; 1988.
Senda M. Mapping cortical functions using PET activation technique. In: Amsterdam SM, ed. New Horizons in Neuropsychology. New York, NY: Elsevier Science Publishing Co; 1994:23-34.
Berthier ML, Starkstein SE, Leiguarda R, Ruiz A, Mayberg HS, Wagner H, Price TR, Robinson RG. Transcortical aphasia: importance of the nonspeech dominant hemisphere in language repetition. Brain. 1991;114:1409-1427.