Early Aphasia Rehabilitation Is Associated With Functional Reactivation of the Left Inferior Frontal Gyrus
A Pilot Study
Background and Purpose—Early poststroke aphasia rehabilitation effects and their functional MRI (fMRI) correlates were investigated in a pilot, controlled longitudinal study.
Methods—Twelve patients with mild/moderate aphasia (8 Broca, 3 anomic, and 1 Wernicke) were randomly assigned to daily language rehabilitation for 2 weeks (starting 2.2 [mean] days poststroke) or no rehabilitation. The Aachen Aphasia Test and fMRI recorded during an auditory comprehension task were performed at 3 time intervals: mean 2.2 (T1), 16.2 (T2), and 190 (T3) days poststroke.
Results—Groups did not differ in terms of age, education, aphasia severity, lesions volume, baseline fMRI activations, and in task performance during fMRI across examinations. Rehabilitated patients significantly improved in naming and written language tasks (P<0.05) compared with no rehabilitation group both at T2 and T3. Functional activity at T1 was reduced in language-related cortical areas (right and left inferior frontal gyrus and middle temporal gyrus, right inferior parietal lobule and superior temporal gyrus) in patients compared with controls. T2 and T3 follow-ups revealed a cortical activation increase, with significantly greater activation in the left hemisphere areas in rehabilitated patients at T2 and T3, and a time×treatment effect at T2 in the left inferior Broca area after rehabilitation. Left inferior frontal gyrus activation at T2 significantly correlated with naming improvement.
Conclusions—Early poststroke aphasia treatment is useful, has durable effects, and may lead to early enhanced recruitment of brain areas, particularly the left inferior frontal gyrus, which persists in the chronic phase.
An important issue for language rehabilitation after stroke is the relationship between the effects of therapeutic interventions and the functional changes observed in brain language-related areas. Most studies focused on the spontaneous recovery of language,1–7 but only a few, often single case descriptions, focused on patients submitted to language rehabilitation.8–13 All of these were conducted on chronic patients.
The rate of complete spontaneous (ie, without language therapy) aphasia recovery poststroke has been estimated at ≈33% in the first month, 43% after 4 months, and 50% 12 months later.14 However, the concept of spontaneous recovery of language may not be appropriate (as human beings are extensively exposed to language). The term language therapy commonly relates to formal interventions provided by speech therapists aimed at improving different language abilities with specific exercises.
Evidence supports the use of language rehabilitation in stroke, particularly if intensively administered for short courses in chronic aphasics.15,16 In a pilot study, Godecke et al17 found that daily aphasia therapy in early stroke, that is, mean 3 days poststroke (dps), resulted in improved communication outcomes in moderate to severe aphasia.
Brain plasticity induced by language training has not been extensively investigated to date. Several studies, conducted in chronic patients, suggested that functional correlates of aphasia recovery could be either the activation of predominantly left hemisphere (LH) language-related areas18–22 or of the homologous right hemisphere (RH) areas,2,23–25 but often the authors do not clarify whether patients were formally treated and to what extent, or not. Saur et al4 suggest that in the postacute stages of recovery (2 weeks poststroke), the LH areas would be reduced in activation, whereas after months they would increase their activation, reflecting the language improvement. Patients examined in the study were reportedly submitted to language therapy, but details on its duration and type were not provided.
A maladaptive role of the RH language-related areas in aphasia recovery has also been hypothesized; Breier et al11 and Heiss and Thiel26 suggested that the RH could be implicated when large LH lesions are present or when aphasia recovery is incomplete. Finally, brain modulation with transcranial direct current stimulation or transcranial magnetic stimulation also supports this hypothesis, because anodal inhibition over the RH has been reported to improve naming in chronic aphasics.6,27 The recruitment of RH may reflect maladaptive functional reorganization, brought about by the presence of the lesion itself, or a transcallosal disinhibition, unrelated to the level of recovery.
Furthermore, some studies clearly addressing the functional correlates of language rehabilitation, particularly anomia, performed on chronic patients found LH,10 bilateral,9,28 or RH29 activation. It is worth noting that functional MRI (fMRI) studies using different tasks8,19,20,24,30,31 also found variable activation patterns within the language-related areas (see Thompson and den Ouden31 for a review), suggesting that the whole network could be consistently activated by different language modalities.
In conclusion, language improvement was consistently found to be associated with modifications at a neural level, but the effect of rehabilitation on brain plasticity, particularly in the acute phase, is unclear and the way in which training modulates functional recovery remains to be determined.
Therefore, the aim of the present pilot study was double-fold. First, we aimed to replicate previous findings on the efficacy of early poststroke aphasia rehabilitation in people with moderate to severe aphasia17 on a group of patients with mild/moderate aphasia. We predicted that daily language rehabilitation leads to the improvement of communicative outcomes. Second, we aimed to investigate the functional correlates of language rehabilitation. We predicted that language therapy should be able to address brain plasticity,32 promoting an early shift from RH to LH language-related areas in rehabilitated patients. Hence, early recruitment of LH perilesional areas would be the correlate of language improvement in early treated patients.
Material and Methods
Inclusion criteria were: (1) first ever acute stroke in the territory of the middle cerebral artery; (2) aphasia with mildly impaired oral comprehension, that is, comprehension sufficient to perform a screening task of the fMRI paradigm; (3) native Italian speakers; (4) right-handedness33; (5) absence of previous history of other neurological or psychiatric diseases; (6) absence of general contraindications to MRI; (7) age <80 years; (8) absence of hearing deficits. Twelve first-ever-stroke patients participated, with 6 randomly assigned to Rehab group and 6 to NRehab group. A random number generator was used to randomize participants. The study was conducted in accordance with the Helsinki Declaration and was approved by the Spedali Civili of Brescia institutional board. Informed oral consent was obtained from each patient and written informed consent by the closest relative. See the online-only Data Supplement for additional information.
All patients had a CT scan and a NIHSS (National Institute of Health Stroke Scale) evaluation performed in the first 24 hours poststroke. Details of lesion volume analysis are available in the online-only Data Supplement.
Patients who entered the study were assessed 3 times, both with neuropsychological evaluation and with fMRI. Examinations were conducted ≈2 dps (T1: mean [SD], 2.2 [1.3] dps), 2 weeks poststroke (T2: 16.2±1.3 dps), and 6 months later (T3: 190±25.5 dps). Neuropsychological assessment and fMRI were separated by ≤24 hours. Ten sex- and age-matched healthy subjects were also examined with the same fMRI paradigm only once.
The Aachen Aphasia Test (AAT)34 was used in all 3 examinations, given the moderate severity language impairment of our patients and the high test–retest reliability of the AAT. The test includes subtests of repetition, naming, reading, writing, oral, and written comprehension; a 50-item version of the Token test; and a semiquantitative scoring of several aspects of spontaneous speech (communicative ability, articulation and prosody, automatic speech, semantic, phonemic, and syntactic structure). Neuropsychological assessment was performed by a blinded speech therapist. We separately considered oral and written comprehension in the AAT comprehension subtest to better understand the behavioral data. The number of correct responses was used for all the subtests, except for spontaneous speech, which was scored with a mean of each spontaneous speech aspect’s score.
Severity (based on the Italian AAT version) and subtyping of patients’ aphasia was also obtained, according to the AAT procedure.
For each treated patient, language therapy started 2 dps after the first evaluation and lasted until the subsequent examination, ≈2 weeks poststroke. Treatment consisted of a 1-hour session per day, for 5 days per week. After T2, no patient was further rehabilitated. Given the low severity of their aphasia and the low impact of the disorder on their everyday activities, this was not considered an ethical violation. For details on rehabilitation, see the online-only Data Supplement. NRehab patients did not undergo any language rehabilitation, as was the clinical practice of our hospital in the acute phase of stroke. However, they were all exposed to the natural speech environment of people they were visited, and this could be considered an unstructured language therapy.
An oral comprehension task was used in an event-related fMRI experiment at T1, T2, and T3, similar to the task used in Saur et al.4 This task was chosen because comprehension is thought to be the earliest linguistic-improving ability poststroke and because of the well-studied functional correlates of the comprehension task in the poststroke acute phase. Details of fMRI paradigm are available in the online-only Data Supplement.
Details of fMRI acquisition and preprocessing are available in the online-only Data Supplement. For each subject, analysis was performed by pooling correct and incorrect predictors (intelligible sentences) versus reverse speech and noise [1 1 −1 −1], and relative contrast maps were calculated with a confidence level ≤0.001.
First of all, within-group analysis was performed, using a general linear model approach with statistical threshold of 0.001 (uncorrected), considering contrasts in each examination (T1, T2, T3) for both controls and patients. To understand the difference in activation between Rehab and NRehab, a between-group comparison was performed for each examination (T1Rehab>T1NRehab, T2Rehab>T2NRehab, T3Rehab>T3NRehab), considering significant effects at P<0.001. A cluster size threshold of 160 mm3 was used. Moreover, to study the longitudinal effect of rehabilitation on brain activations, a within-group analysis of contrasts between examinations (T2>T1 and T3>T2) was performed as well for both Rehab and NRehab groups (error level <0.005; uncorrected).
Finally, as postrehabilitation differences were expected to be found, a region of interest (ROI) selection was carried out to perform a single ROI statistical analysis. To perform ROI selection, controls’ and patients’ activation at all examinations (Rehab and NRehab at T1, T2, and T3) were pooled. ROIs were defined besed on the resulting whole group activation: bilateral inferior frontal gyrus (IFG) and bilateral superior temporal gyrus (STG). A 2-way ANOVA on ROI activations using Group (2 levels: Rehab and NRehab) as a between-factor and Time (3 levels: T1, T2, and T3) as a within-factor was performed for each ROI, with a confidence level of 0.05.
Sociodemographical and Clinical Features
Rehab patients included 2 women and 4 men, whereas NRehab patients included 3 women and 3 men. Age, education (years of schooling), and NIHSS were not statistically different between the groups. At baseline, the patients showed Broca aphasia in 8 cases, anomic aphasia in 3, and mild Wernicke aphasia in 1 case. Lesions mainly involved areas in the left frontal lobe: insula, thalamus, internal and external capsula, and the parietotemporal junction. Neither thrombolytic nor clonidine, antipsychotic, or antiepileptic drugs were administered to the included patients. No differences in lesion volume were detected between the 2 groups. The Broca area was damaged by the lesion in 3 patients both in the Rehab and in the NRehab groups. All the included patients completed the study, except 1 participant in the NRehab group who died after T2 because of a cardiac arrest (for details about sociodemographic features, see Table I in the online-only Data Supplement; for details of lesion volume analysis, see Figure I and the Results section in the online-only Data Supplement).
Aphasia Evolution Results
Because the neuropsychological data were not all normally distributed at Shapiro–Wilk test, the data were analyzed by means of nonparametric Mann–Whitney U-tests, with Bonferroni correction. AAT subtest scores for Rehab and NRehab groups at T1, T2, and T3 are reported in Table II in the online-only Data Supplement. At T1 evaluation, groups were not statistically different in all AAT subtests, whereas at T2, Rehab patients obtained a statistically higher score in both naming and written language subtests compared with NRehab patients (mean naming: 110/120, 80/120; P=0.01; mean written language: 84.5/90, 48.7/90; P=0.02). Although all the patients examined singularly or as separate groups improved in all AAT scores at T2, the remaining AAT subtests were not statistically different between groups. The same was observed at T3 (mean naming: 113/120, 98/120; P=0.004; mean written language: 85.5/90, 71/90; P=0.03), confirming a significantly better performance of Rehab patients compared with NRehab patients in written language and naming tests. During the clinical interview at T2, Rehab patients and their relatives appeared satisfied about the improvement in language functions.
Task performance analysis did not reveal differences between groups across examinations (see the online-only Data Supplement). In control subjects, the analysis revealed strong bilateral activation of several brain language-related areas: left and right IFG, anterior part of left and right STG, left inferior parietal lobule (IPL) and precuneus, left supplementary motor area in the medial frontal gyrus, and left middle temporal gyrus. Two days after the stroke, all Rehab and NRehab aphasic patients, as a group, showed a markedly reduced cortical activation both in the RH and in the LH compared with controls. In particular, clusters of activation were found in left and right IFG, left and right middle temporal gyrus, right IPL, and right STG (Table 1 and Figure 1). Based on activations of controls and patients, we selected the following ROIs: bilateral IFG, left middle temporal gyrus, right STG, and left IPL.
Group analysis at T2 follow-up showed different patterns of activation in Rehab and NRehab groups. Rehab patients activated left and right IFG, left and right STG, left IPL/supramarginal gyrus, and left supplementary motor area, whereas in NRehab patients, the main activated area at T2 was only the right IFG. The remaining language-related areas, except left IPL/supramarginal gyrus, showed less cortical activation. At 6-month follow-up (T3), an increase in both hemispheres’ cortical activations was similarly seen in both groups, mainly localized in LH, particularly the left and right IFG, left STG, left and right IPL/supramarginal gyrus, and left supplementary motor area (Figure 2 and Table 2). This pattern was confirmed by the between-group comparison. At T1, no difference between Rehab and NRehab groups was found, as was expected, whereas at T2, left IFG revealed to be predominantly activated by Rehab patients followed, at 6 months poststroke, by a similar recruitment of areas predominantly located in the LH, again with a higher recruitment observed in Rehab patients compared with NRehab patients (Figure 2 and Table 2). The opposite NRehab>Rehab contrast showed no areas with increased activation in the NRehab versus the Rehab group.
Within-group changes in brain activation over time (T2>T1 and T3>T2) are visualized in Figure 3 and reported in Table 3. After a 2-week course of language rehabilitation, an increased activation was found in left IFG (Broca area) and in STG in Rehab patients, whereas NRehab group showed change in activation of right IFG and left IPL. The 6-month follow-up (no language rehabilitation condition for both groups) change in activation was similarly present in LH Broca area, STG, and IPL in both groups. To summarize, left IFG was predominantly recruited 2 weeks after the stroke only in patients submitted to language rehabilitation, whereas after 6 months, all the patients had a prevalent recruitment of brain language-related areas located in the LH.
These results were subsequently confirmed using ANOVA: a treatment×time interaction was found in the left IFG at T2 only (F=10.2; P=0.009), with significantly higher activation in the Rehab group as revealed by post hoc test (P<0.05). No significant difference was found between groups at T1 and T3, neither in other ROIs, indicating left IFG as the crucial site of different brain activation between groups, at T2 only.
Analysis of Correlations
To further corroborate the previous results, a correlation between functional changes after rehabilitation in ROIs and naming and written language changes (T2>T1) was performed. A significant correlation was found between functional imaging changes in the left IFG (T2>T1) only and improvement in naming (r=0.957; P<0.003) in Rehab, and changes in the right IFG only and improvement in naming (r=0.821; P<0.015) in NRehab. See Figure 3 for details.
Two main results arise from this pilot study: first, the clinical efficacy of aphasia rehabilitation has been well demonstrated in acute poststroke patients; second, the possible functional correlate of this treatment-related language improvement is supposed to be the left IFG early activation during the process of aphasia recovery.
Rehabilitation was found to improve naming abilities as well as writing and reading aloud in aphasic patients. The lack of significant difference in comprehension and repetition may, in our opinion, be because of the milder deficits presented by some of the patients, possibly inducing a ceiling effect. The language improvement of Rehab patients, which was evident after 15 days, persists 6 months after stroke. The present results are in line with Godecke et al’s17 findings and may have relevant clinical implications. Although the results need to be confirmed on larger samples of patients harboring aphasic syndromes of diverse severity, they seem to support early daily rehabilitation program of aphasic stroke patients as a valuable intervention. The results also provide at least partial answers to clinicians in deciding whether to wait for spontaneous improvement to occur or provide intensive early aphasia treatment. On the contrary, the present findings are not in line with the negative results of the ACT Now Study,35 in which communication therapy and unstructured social contact were compared in the first 4 months after stroke. This latter study seems to rule out any added benefit of communication therapy. However, important methodological difference could explain the different results: contrary to the present study, the ACT Now Study was not focused on the first-ever-LH-stroke patients, but also included patients with dysarthria, and critical time window for rehabilitation was not considered in the first 2 weeks poststroke.
The second main finding of the present study was that patients submitted to early intensive (2 weeks) poststroke rehabilitation showed a different recruitment of language-related cortical areas compared with the no treatment group, particularly involving the LH. Although all patients showed a dramatic decrease in cortical activations compared with controls 2 dps, Rehab patients showed larger predominantly left-sided areas at the second examination, particularly the left Broca area, compared with NRehab patients, correlating with language behavioral improvement. A further increase in activation in LH at 6-month (T3 to T2) follow-up (when both groups were not submitted to language rehabilitation), with a persistently higher activation in Rehab patients, was also found. In our opinion, the intensive aphasia training protocol seems to promote the early shifting of activation to the LH in the first 2 weeks poststroke, as observed in our patients submitted to early aphasia therapy. On the other hand, NRehab group showed a pattern of activation which is similar to the one found in the Saur et al (2006) patient group, ie, a reduced activation in the L language area in the acute phase, followed by a recruitment of homologue language zone in subacute phase and by a normalization of activation in chronic phase. Importantly, we could reasonably rule out the practice effect of task on fMRI results. In fact, we did not find differences over time in fMRI task performance, especially between T1 and T2, and if practice had influenced the findings, we would have expected decreased activation between T1 and T2,36 which is the opposite of what we actually found.
These findings suggest that a possible correlate of early language poststroke rehabilitation could be the earlier recruitment of the LH, namely the left Broca area, whereas the correlate of spontaneous, and possibly less efficient, language recovery in NRehab patients could be the activation of right IFG, which we found to be related with language improvement in NRehab only.
Our results support Saur et al4’s hypothesis of a different temporal dynamics in LH and RH activation after stroke, but most importantly, they point to a specific correlation between early language training and early recruitment of the left Broca area. The frequency of left IFG involvement in the ischemic lesion was identical in the 2 groups, so the possibility that the results may be driven by a sparing of the area seems unlikely. The fall in activation of many language-related areas not involved in the lesion at T1, and their subsequent activation over time, may be otherwise ascribed to their early functional suppression because of diaschysis and the eventual reduction of the ischemic penumbra, associated with a specific exercise-induced brain plasticity. It may be speculated that the more the naming abilities after language training, the higher and earlier would be the recruitment of the left Broca area, possibly causing persistent effects after 6 months.
The hypothesis of an incomplete aphasia amelioration sustained mainly by the right-sided language-related areas, particularly by the right IFG (right Broca jomologue) and, conversely, a better outcome related to activation in left perilesional areas, has been sustained by many authors,7,11,30,37,38 if not all,39 when examining chronic aphasics. Only in some patients, right IFG activation would be essential for residual language function; however, its compensatory potential would be less effective than in patients who recover left IFG function.
Our results support the role of left IFG activation in acute poststroke patients submitted to language rehabilitation and point to the potential role of right IFG as the neural substrate for aphasia spontaneous recovery.
The activation of left IFG has also been described as a reflection of a therapy-induced improvement of naming8,12 in chronic aphasics, whereas right IFG has been described to be activated both by healthy subjects who were taught to name objects in a foreign language40 and in aphasics with left IFG lesions.41 Our study suggests a possible dichotomy between LH and RH IFG activation, respectively, in early rehabilitated and spontaneously recovered aphasics. The differential activation of left IFG reported in patients who received rehabilitation seems to specifically reflect an effect of language therapy in the first 2 weeks poststroke.
Neural correlates of the training-induced neuropsychological improvement have been reported to be different in relation to the linguistic task used,42,43 as well as in relation to the training used both in aphasia12 and in other nonstroke-induced neuropsychological symptoms, that is, the executive functions and information processing training in multiple sclerosis.44
This pilot study, however, has several limitations. First, the study used a small number of patients with relatively low severity of aphasia, probably because of the small lesion size. This small sample size reflects on the difficulty in including homogeneous acute stroke aphasics in an fMRI study. Second, we did not use a control therapy, because the daily social interaction of patients in the stroke unit could, in our opinion, be considered a unstructured language treatment. Third, the inclusion of patients with moderate/mild baseline impairment may have induced a ceiling effect that prevented a higher improvement in their performance over time. Fourth, data were not corrected for multiple comparisons because of the small sample size, causing the statistical power to be low. Finally, we found improved output language tests in our patients, which we tended to associate with brain activation induced by a comprehension task. On this point, we argue that, in conjunction with several studies,31 the same language-related areas have been activated by different tasks (either comprehension or naming). Hence, it is not unlikely that different language tasks activate the same language-related areas. Further studies with different fMRI language tasks, larger samples including patients with more severe aphasia, and using ecological scales on communicative abilities will be useful.
We acknowledge A. Duina and M. Tosetti for technical support.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.113.003192/-/DC1.
- Received August 17, 2013.
- Accepted October 30, 2013.
- © 2013 American Heart Association, Inc.
- Crinion J,
- Price CJ
- Saur D,
- Lange R,
- Baumgaertner A,
- Schraknepper V,
- Willmes K,
- Rijntjes M,
- et al
- Winhuisen L,
- Thiel A,
- Schumacher B,
- Kessler J,
- Rudolf J,
- Haupt WF,
- et al
- van Oers CA,
- Vink M,
- van Zandvoort MJ,
- van der Worp HB,
- de Haan EH,
- Kappelle LJ,
- et al
- Blanken G
- Huber W,
- Poeck K,
- Kerschensteiner M.
- Bhogal SK,
- Teasell R,
- Speechley M
- Fridriksson J,
- Bonilha L,
- Baker JM,
- Moser D,
- Rorden C
- Fridriksson J
- Fernandez B,
- Cardebat D,
- Demonet JF,
- Joseph PA,
- Mazaux JM,
- Barat M,
- et al
- Richter M,
- Miltner WH,
- Straube T
- Baker JM,
- Rorden C,
- Fridriksson J
- Vitali P,
- Abutalebi J,
- Tettamanti M,
- Danna M,
- Ansaldo AI,
- Perani D,
- et al
- Huber W,
- Poeck K,
- Weniger D,
- Willmess K
- Bowen A,
- Hesketh A,
- Patchick E,
- Young A,
- Davies L,
- Vail A,
- et al
- Winhuisen L,
- Thiel A,
- Schumacher B,
- Kessler J,
- Rudolf J,
- Haupt WF,
- et al
- Musso M,
- Weiller C,
- Kiebel S,
- Müller SP,
- Bülau P,
- Rijntjes M