The changed functional status of the brain was involved in patients with poststroke aphasia: Coordinate‐based (activation likelihood estimation) meta‐analysis

Abstract Background and Purpose Although many functional magnetic resonance imaging (fMRI) studies have investigated the language architecture and neurobiological mechanism underlying poststroke aphasia (PSA), the pathophysiological mechanisms of PSA still remain poorly understood. In addition to a limited number of subjects (<20) tested with different methodologies and stimuli, inconsistent reports of the brain regions involved have been a major factor. Thus, we conducted a meta‐analysis of 12 peer‐reviewed studies of abnormal brain activation regions in PSA patients at rest using activation likelihood estimation (ALE). Results A meta‐analysis was performed based on 24 experiments with 497 total participants in 12 studies to establish the ALE of regional activation in PSA. Through experiments with PSA patients and healthy controls, we found that hypoactivation in PSA converged on the left superior frontal gyrus and the left parietal postcentral gyrus, whereas there was hyperactivation in the right cerebellar anterior lobe, left fusiform gyrus, left superior parietal lobule, and right subgyral hippocampus. Conclusion Our study verified that dominant and nondominant language networks play roles in the recovery of language function.


TA B L E 1 Socio-demographic and clinical characteristics of the subjects included in the selected studies
of their language function. The remainders are left with persistent and disabling impairment of communication (Geranmayeh et al., 2014).
With the advent of neuroimaging, many studies have investigated the abnormal regional activity and functional connectivity (FC) of PSA (Guo et al., 2019;Mayorova et al., 2019;Sandberg, 2017;Yang et al., 2016;Zhu et al., 2014). However, the results of these studies are inconsistent because of small samples (<20 patients) and variable analysis methods. For example, Yang et al. (2016)  To overcome this problem, we performed an activated likelihood estimation (ALE) meta-analysis to study the convergence of the activated brain regions in PSA patients and HCs using rs-fMRI. The ALE approach can be used to integrate the reported coordinates of activation maxima from published studies to identify regions of significant converging activation (Warren et al., 2009). This approach has two advantages: We can take whole pictures of brain activation differences between PSA patients and healthy controls, and we can study relatively larger samples. This study is expected to contribute to further understanding of the pathology of PSA.

| Literature search and selection criteria
In May 2020, the relevant studies were identified through a systematic online database search for peer-reviewed articles on Web of Science, PubMed, and Elsevier. The searches were conducted with the keywords "stroke" or "apoplexy" or "cerebralvascular accident" or "cerebral infarction" or "cerebral haemorrhage", "aphasia", "resting-state functional magnetic resonance imaging" or "rs-fMRI". First, 568 articles were obtained from the initial screening.
Twenty-seven duplicated articles were removed because they were duplicated. Then, 541 potential studies were further assessed according to the following criteria: (a) The study included PSA patients, (b) the study used rs-fMRI, (c) the coordinates in the selected studies were in Montreal Neurological Institute (MNI) (Collins et al., 1994;Evans et al., 1993) space or Talairach space (Talairach, 1988), (d) the studies were written in English, (e) the studies were research articles, (f) the reports contain comparisons between the PSA group and health controls, and the PSA patients did not receive an intervention, and (g) quantitative whole-brain analyses were performed.
Reports with ROI analyses were excluded. Finally, a total of 12 relevant rs-fMRI studies were included in the final meta-analysis, which included 270 healthy participants and 227 stroke survivors. Table 1 illustrates the socio-demographic and clinical characteristics of the subjects included in the selected studies. The retrieval process of each selected study is shown in Figure 1, and the methodological characteristics are shown in Table 2.

| Activation likelihood estimation meta-analysis
An ALE meta-analysis was performed using GingerALE software (Eickhoffet al., 2012;Turkeltaub et al., 2012) using the most recent users' manual (http://brain map.org/ale/manual.pdf). In the ALE meta-analysis of single datasets, we connected the coordinates from the rs-fMRI studies showing "HC > PSA" and "HC < PSA." The ALE map was computed with p < .05 FEW cluster-corrected thresholds with 1,000 permutations (Collins et al., 1994). According to the manual, this threshold is the most conservative and appropriate level.
Resulting ALE maps were visualized by Mango software 4.0.1 (The University of Texas, Austin TX, USA) with the Colin brain template in MNI space (http://brain map.org/ale/).  Table 3 and Figure 2).

| Brain regions with lower activation among HCs than among PSA patients
As demonstrated in Table 3 and Figure

| D ISCUSS I ON
In the present study, ALE meta-analysis was used to compare brain activation regions involved in PSA relative to HCs in rs-fMRI. The neuroimaging studies on stroke patients (Kang & Kim, 2008;Tohgi et al., 1995). The SFG was used for spatial analysis of working memory (Boisgueheneuc et al., 2006) and was used to process phonological or vocabulary search tasks (Tohgi et al., 1995;Tonkonogy & Goodglass, 1981). In our results, the reduced activation of the SFG in PSA was consistent with the patients' speech fluency and lexical dysphonia. The parietal postcentral gyrus has been considered an important region of somatosensory (Corkin et al., 1970) and writing abilities (Magrassi et al., 2010). Furthermore, in terms of FC, Gómez, Flores, and Ledesma (2007) found a strong correlation between the left parietal postcentral gyrus and the SFG, which suggests that frontoparietal networks in extensive processing of information during language production.
Converged hyperactivation was observed in the left fusiform gyrus (BA37) and subgyral hippocampus. A similar result has been found in previous rs-fMRI studies that observed the activated regions during the changes to 17 PSA brain functions (Yang et al., 2016). Based on visual picture-naming and auditory-naming tasks to evaluate the cortical structure and dynamics of lexical semantic networks in driving speech production, the authors found that a distinct neuroanatomical substrate in the fusiform gyrus provided access to object semantic information (Forseth et al., 2018)]. Furthermore, the authors found that compared with HCs, PSA patients displayed hyperactivation of the left fusiform gyrus in a resting state, which may be related to the compensatory mechanism of semantic processing. The correlation suggested that the activity in the contralateral parahippocampus may mediate functional recovery of the right homologous language regions (Liégeois et al., 2004). Thus, the fusiform gyrus is very important for aphasia language recovery.
Converged hyperactivation during a resting state in SPA was also observed in the cerebellar right anterior lobe and the SPL.
A previous rs-fMRI study, which found increased activation of the bilateral cerebellum and ipsilesional superior parietal lobe during measuring the dynamic amplitude of low-frequency fluctuations (dALFF) in patients with subacute stroke, reported similar results . The cerebellum was involved in the articulation, visual space, and executive power (Stoodley & Schmahmann, 2010), whereas the parietal lobe was associated with writing and cognitive function (Bzdok et al., 2016;Magrassi et al., 2010). A study that measured brain metabolism in PSA showed diffuse hypometabolism in the ipsilateral parietal area and contralateral cerebellum (Kim et al., 2012). Such results suggested that the cerebellum and parietal activation indicate the patients' potential for further language improvement. Therefore, the right cerebellar anterior lobe and the SPL were closely related to language recovery in PSA.

F I G U R E 1
The literature search and the process of inclusion or exclusion of articles TA B L E 2 All studies entered into the meta-analysis are listed, including year, first author, stereotactic space, neuroimaging, imaging method, and active contrast conditions

| CON CLUS ION
The present study employed a meta-analysis approach to map the convergence of abnormally activated brain areas in PSA. Our ALE meta-analysis revealed that compared with HCs, PSA hypoactivation converged in the SFG and parietal postcentral gyrus, whereas hyperactivation occurred in the right cerebellar anterior lobe, fusiform gyrus, SPL, and subgyral hippocampus. Our study verified that dominant and nondominant language networks play roles in the recovery of language function. This study contributes to a better understanding of brain activation in PSA in the resting state. Moreover, our findings play an important role in understanding the pathophysiological mechanism of aphasia, creating rehabilitation strategies and predicting the effect of rehabilitation.

ACK N OWLED G M ENT
We would like to thank Ms Xuemei Wei and Mingxian Li for screening papers.

D I SCLOS U R E S
There was no conflict of interest to declare.

AUTH O R CO NTR I B UTI O N
Ying Du took part in the whole process of research and was responsible for literature search and essay writing. Yujun LEE was responsible for research design, data process, and data analysis. Chuan He Qian Yong and Zhiyi Cen were responsible for collecting data. Lihan Peng was responsible for data checking. Xin Liu was responsible for choosing literature. Yuqing Chen and Xiaoming Wang were responsible for guiding research design.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.1867.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available