Stroke is a serious neurological disease. According to the 2022 release of the China Stroke Report, the incidence rate of stroke in China is 2022.0 per 100,000, with an annual incidence of 276.7 per 100,000 and a mortality rate of 153.9 per 100,000 [1]. Poststroke cognitive impairment (PSCI) is defined as a sustained cognitive deficit for at least 6 months after a stroke [2], which affects about one-third of patients with stroke [3]. PSCI significantly elevates mortality risk [4], impedes the recovery of motor and speech functions and activities of daily living [5], and substantially increases the caregiver burden and socioeconomic costs [6]. Therefore, focusing on the rehabilitation of patients with PSCI is of crucial clinical importance.

Repetitive transcranial magnetic stimulation (rTMS) is a safe, efficient, and well-tolerated noninvasive neuromodulation technique [7]. It induces cortical excitability changes through electromagnetic induction [8]. For PSCI, 2 primary therapeutic models exist: the interhemispheric inhibition model and the compensation model. The interhemispheric inhibition model posits that stroke disrupts callosal balance, suppressing the affected hemisphere while hyperactivating the unaffected hemisphere [9]. rTMS counteracts this by upregulating the affected hemisphere excitability (via high-frequency stimulation) or downregulating the unaffected hemisphere activity (via low-frequency stimulation) [10]. The compensation model proposes neuroplastic reorganization around lesions or in contralesional regions to compensate for deficits [11]. These models are integrated in the bimodal balance-recovery model, which recommends selecting rTMS strategies based on the structural preservation in the affected hemisphere [12].

Studies indicate that rTMS can modulate neural activity in specific brain regions [13,14] and improve PSCI behavioral indicators [15-23]. Numerous meta-analyses have consistently supported the effectiveness of rTMS intervention for improving cognitive function in patients with stroke [15-23]. Multiple meta-analyses consistently demonstrate rTMS-induced cognitive improvement in patients after stroke across diverse assessment tools (the Montreal Cognitive Assessment [MoCA], the Mini-Mental State Examination [MMSE], the Loewenstein Occupational Therapy Cognitive Assessment, and the Repeatable Battery for the Assessment of Neuropsychological Status) [16,17]. Significant global cognitive benefits have been observed with both high-frequency rTMS (HF-rTMS) and low-frequency rTMS (LF-rTMS) protocols targeting the dorsolateral prefrontal cortex (DLPFC) [16,17]. Enhanced executive function and working memory (although not memory or attention) have also been observed when rTMS is combined with cognitive training, with optimal effects from HF-rTMS over the left DLPFC or LF-rTMS over the right DLPFC [18]. The superior efficacy of HF-rTMS compared with LF-rTMS for DLPFC stimulation in improving cognition and functional independence (as measured by the Modified Barthel Index) has also been reported [19]. Although these studies have confirmed the behavioral efficacy of rTMS in improving PSCI, the potential neuroimaging mechanisms underlying this behavioral effect remain unclear.

Resting-state functional magnetic resonance imaging (rs-fMRI) provides a noninvasive tool for mapping rTMS-induced neural changes in PSCI [24,25]. Evidence indicates that rTMS can modulate the excitability and functional connectivity (FC) in cognition-related networks (eg, enhanced DLPFC-frontotemporal coupling [26], and increased amplitude of low-frequency fluctuation [ALFF] in the medial prefrontal cortex [27]). Concurrently, cognitive scores (MoCA, MMSE, and Modified Barthel Index) and neural activity have also improved [26-31]. However, systematic evidence quantifying the consistent neuroimaging effects of rTMS remains limited, which hinders the clarification of its mechanistic pathways and clinical optimization. Existing findings are predominantly derived from small-sample, single-center randomized controlled trials (RCTs), resulting in insufficient statistical power and generalizability [27,28]. Although narrative reviews acknowledge the potential of functional magnetic resonance imaging (fMRI) for elucidating rTMS mechanisms, they rely on the qualitative synthesis of methodologically heterogeneous studies and cannot integrate spatial patterns of neural change [32]. Crucially, no quantitative meta-analysis has yet synthesized coordinate-based neuroimaging data to identify robust, spatially convergent neural targets of rTMS in PSCI.

This coordinate-based meta-analysis (CBMA) aims to (1) quantitatively synthesize evidence regarding the effects of rTMS on cognitive function in patients with PSCI, (2) identify consistent patterns of rTMS-induced neural changes using voxel-wise meta-analysis of fMRI data, (3) explore the effects of intervention parameters—including stimulation frequency, target location, and intensity—as potential moderators of efficacy, and (4) explore whether rTMS sham stimulation intervention has a placebo effect. This study will pioneer the application of CBMA in rTMS neuromodulation research for PSCI, with the goal of advancing mechanistic understanding and supporting evidence-based optimization of stimulation protocols.

The certainty of the evidence will be evaluated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework [45]. This systematic approach assigns 1 of 4 grades, “high,” “moderate,” “low,” or “very low,” to each outcome, considering limitations in design and implementation, indirectness of evidence, unexplained heterogeneity, imprecision of results, and high probability of reporting bias [46]. Two independent researchers (XX and YR) will use the “GRADEprofiler” software to assess the evidence and then import the findings into the Review Manager software (version 5.4; The Nordic Cochrane Centre, The Cochrane Collaboration). In case of discrepancies between the researchers’ assessments, a third researcher (NZ) will arbitrate.

The GRADE framework was originally designed to assess the quality of evidence for clinical outcomes (eg, mortality, symptom severity, and functional status) and is not directly applicable to neuroimaging-derived biomarkers such as ALFF, ReHo, and FC, which serve as mechanistic or surrogate end points. Therefore, in this study, the GRADE tool will be applied exclusively to behavioral measures (eg, MoCA and MMSE) for evaluating the certainty of evidence.

The study aims to investigate how rTMS improves cognitive function in patients with PSCI at the neuroimaging level. We plan to integrate relevant studies and introduce CBMA to quantitatively synthesize findings from individual neuroimaging studies [47]. In this study, we will apply SDM-PSI, which is a new algorithm for CBMA used for analyzing rs-fMRI imaging data [48,49]. First, we will use CBMA to explore patterns of neuroimaging biomarker changes in patients with PSCI following rTMS intervention. Second, we will use CBMA maps to compare differences in brain region changes before and after interventions and between groups. Third, we will use meta-regression analysis methods to investigate correlations between changes in neuroimaging indicators in different brain regions and behavioral and demographic variables. The importance of this study is that it will provide intuitive and visual imaging evidence for the efficacy of rTMS and may aid in clinical decision-making regarding PSCI rehabilitation programs through CBMA.

A potential limitation of this study is that studies using rTMS for the treatment of animals were excluded from our meta-analysis. In addition, our study only reports imaging metrics such as ALFF, fALFF, ReHo, and FC. Other imaging metrics from rs-fMRI, such as effective connectivity and small-world properties, were not considered. Imaging metrics that reflect white matter and gray matter structure, such as diffusion tensor imaging and voxel-based morphometry, were also excluded. Moreover, the field of rs-fMRI still faces challenges. First, temporal variability includes susceptibility to physiological noise (eg, respiratory and cardiac artifacts) and state-dependent fluctuations [50]. Second, interpretive ambiguity arises from difficulty in distinguishing neural from vascular contributions to the BOLD signal [51]. Third, analytical heterogeneity results from issues with cross-study comparability arising from inconsistent preprocessing pipelines and connectivity metrics [52]. These limitations need to be addressed in future related studies.

Our protocol, based on rs-fMRI findings, will summarize the beneficial role of neuroimaging metric changes after rTMS intervention in cognitive-related brain regions compared with the baseline level and the control conditions. Moreover, this meta-analysis will investigate the relationship between neuroimaging changes and behavioral outcomes. In summary, our study may aid future clinical decision-making concerning PSCI rehabilitation programs and provide evidence-based medical insights into the neuroimaging mechanisms of rTMS treatment for PSCI.