The protocol for this systematic review (SR) was registered in PROSPERO (registration number CRD420250652194). The review was conducted in accordance with the PRISMA statement (Moher et al., 2009).

A systematic search was conducted across multiple databases from their inception to February 28, 2025. The databases included China National Knowledge Infrastructure (CNKI), SinoMed (China Biology Medicine Disc), the Chinese Scientific Journal Database (VIP), Wanfang Data, PubMed, the Cochrane Library, Embase, and Web of Science. The main search terms are listed in Table 2.

The retrieved articles were first imported into NoteExpress software (version 4.1.0) to remove duplicates. Two independent researchers (Qin and Zhuo) then screened titles and abstracts based on the eligibility criteria. Full texts of potentially eligible studies were subsequently reviewed to determine final inclusion. Any discrepancies were resolved through consensus.

Data extraction was carried out by Yang using a standardized template in Microsoft Excel 2019 to record the following information: general study characteristics (first author, year of publication, study design), participant details (sample size, sex, age, time since stroke, treatment duration), intervention information, control group details, and outcome measures. The Standards for Reporting Interventions in Clinical Trials of Acupuncture (STRICTA) were applied to ensure comprehensive reporting of acupuncture details (e.g., needle depth, retention time). Discrepancies were resolved through discussion among all researchers.

Two researchers (Qin and Li) assessed the methodological quality of RCTs using the Cochrane Risk of Bias 2.0 tool (RoB 2). The RoB results were classified as high risk, low risk, or some concerns. Disagreements were resolved through consensus.

Due to methodological heterogeneity among the included studies, a descriptive analysis was conducted. The synthesis aimed to provide neuroimaging evidence supporting the clinical efficacy of acupuncture for PISCI.

A comprehensive search across multiple databases initially retrieved 1,582 records. After removing duplicates using reference management software, 1,115 records remained. Titles, abstracts, and keywords were then screened against the eligibility criteria, yielding 54 potentially relevant studies. Following full-text assessment, 8 studies (Li, 2023; Su, 2016; Wang et al., 2014, 2021; Wang, 2021; Xiao and Yu, 2024; Yu et al., 2021; Zhang et al., 2020) were included for analysis. The PRISMA flowchart of the search and selection process is shown in Figure 1.

Table 3 summarizes characteristics of the 8 included studies, published between 2014 and 2024. These studies comprised 5 RCTs and 3 non-RCTs, all conducted in China.

The risk of bias assessment for the included RCTs is presented in Figure 2. Of the five RCTs, two were judged to have a high risk of bias, and three were rated as having some concerns according to the RoB 2 tool. Regarding the randomization process, all studies presented certain issues due to inadequate reporting of random sequence generation or allocation concealment. With respect to deviations from the intended interventions, none of the studies blinded acupuncturists, which is expected given the nature of the intervention. Three studies reported participant blinding and were rated as low risk; the remaining two raised some concerns due to unclear blinding descriptions. Risk of bias due to missing outcome data was generally low across all studies. Measurement of the outcome was rated as low risk in all studies. In terms of the selection of the reported result, potential bias may exist due to the lack of complete study protocols or trial registration information.

A total of 397 participants were included across the eight studies, comprising 367 patients with PISCI and 30 healthy controls. Sample sizes ranged from 25 to 80. Two studies compared PISCI patients with healthy controls, while six enrolled only PISCI patients. The age of PISCI patients ranged from 45 to 80 years. All studies reported gender; in total, there were 237 males and 150 females, and only one study had a male proportion below 50%.

In accordance with STRICTA, acupuncture details are summarized in Table 4. All studies reported the theoretical rationale for acupoint selection. The number of needles used per session ranged from 1 to 22 across studies. Among all studies, GV20 (6/8, 75%) was the most frequently used acupoint, followed by EX-HN1 (5/8, 62.5%), and both GV24 and LR3 (4/8, 50% each) (Figure 3). Acupuncture types included scalp acupuncture (1/8, 12.5%), abdominal acupuncture (1/8, 12.5%), and body acupuncture (6/8, 75%). Manual acupuncture predominated (7/8, 87.5%), while electro acupuncture was used in one study (1/8, 12.5%). Needle insertion depth was reported in six studies (15-50 mm) and a specific needling response (e.g., “Deqi”) was described in five studies (5/8, 62.5%). Except for one study, all reported total treatment sessions, ranging from 10 to 72 over 2 to 12 weeks. Practitioner qualifications were reported in three studies (3/8, 37.5%).

Four types of comparisons were identified across the eight studies: acupuncture vs. sham acupuncture (1/8, 12.5%); acupuncture plus drug vs. drug alone (2/8, 25.0%); acupuncture vs. healthy volunteers (2/8, 25.0%); and acupuncture plus cognitive training vs. cognitive training alone (3/8, 37.5%).

PISCI was assessed across multiple domains. All studies used MoCA to evaluate cognitive function, with most (62.5%) also employing MMSE. Additional cognitive measures and assessments of neurological, motor, and daily living functions were used in individual studies (detailed in Table 3).

All included studies were conducted in China, reflecting the widespread clinical use of acupuncture in that context. However, the lack of studies from other countries may limit generalizability to populations with different cultural backgrounds and acceptance of acupuncture. The included studies comprised five RCTs and three non-RCTs. Quality assessment revealed that two RCTs were at high risk of bias, primarily due to inadequately reported randomization process and lack of allocation concealment; the remaining three were judged to have “some concerns,” largely related to deviations from intended interventions and incomplete outcome reporting. These methodological weaknesses represent critical threats to internal validity and indicate that findings should be interpreted with caution. Future research should adhere closely to methodological standards outlined in the Cochrane Handbook for Systematic Reviews of Interventions to improve rigor and transparency, thereby enhancing the overall credibility of the evidence.

All included studies had small sample sizes (fewer than 80 participants), which fundamentally limits statistical power and reproducibility. Prior research has shown that (Klapwijk et al., 2021) small neuroimaging samples may overestimate effect sizes and reduce reproducibility. Without predefined analysis plans, especially for ROI analyses, the risk of false positives further increases (Gentili et al., 2021). Cross-validation may mitigate this risk but cannot fully compensate for inadequate sample sizes (Goltermann et al., 2023). Additionally, all studies recruited participants from tertiary care centers, potentially limiting generalizability to community settings. Future studies should increase sample sizes and standardize data-analysis pipelines. Regarding participant characteristics, seven of the eight studies included more men than women, consistent with the higher global incidence of ischemic stroke in males, especially under the age of 75 (de Miguel-Yanes et al., 2021). Sex may influence neural repair and cognitive recovery after stroke (Madsen and Guo, 2020; Zinman et al., 2023); thus, future trials should explicitly include sex as a biological variable.

Cross-study replication of findings strengthens the credibility of the evidence. Among the five rs-fMRI studies, several brain regions showed consistent changes: the para hippocampal gyrus, fusiform gyrus, and lingual gyrus demonstrated enhanced activity across independent different studies. In contrast, changes in other regions were reported in single studies only, representing preliminary observations requiring validation. However, due to the limited number of included studies and methodological heterogeneity, quantitative meta-analysis was not feasible, which constrained our ability to estimate pooled effect sizes and assess publication bias. Notably, the predominance of positive findings among small-sample studies suggests potential publication bias, although formal statistical testing was not possible due to insufficient study numbers.

The included studies varied widely in acupuncture methods, acupoint selections, and treatment protocols. Seven studies used manual acupuncture, one used electro acupuncture, and only five reported practitioner qualifications. Variability in techniques, needle retention time, and stimulation intensity could influence treatment effects and reproducibility (Yang et al., 2013). Several studies did not fully comply with STRICTA, with incomplete reporting of details such as deqi sensations or acupoint rationale; only five studies described patient responses to acupuncture. Evidence suggests that (Li et al., 2015a; Shi et al., 2016) deqi is associated with activation of regions such as the insula and cingulate gyrus, potentially underlying central effects of acupuncture. Standardization and consistent adherence to STRICTA are therefore critical for improving comparability and reliability.

Regarding acupoint selection, GV20 and EX-HN1 were most frequently used; both are located at the vertex. Unlike distal points commonly used for Alzheimer's disease (AD) or mild cognitive impairment (MCI) (Yin et al., 2023a,b), these acupoints directly stimulate cortical regions beneath the skull. Neuroimaging studies indicate that (Deng et al., 2015; Kong et al., 2024; Wei et al., 2021) these regions correspond to prefrontal and parietal cortices—areas related to executive function, attention, and working memory. Animal research further suggests that (Zheng et al., 2020) GV20 may promote neuroplasticity and neurotrophic factor expression, potentially improving cognitive function.

For outcome measurement, the most commonly used cognitive assessment tools were MoCA and MMSE. Numerous studies indicate that MoCA is more sensitive than MMSE for detecting MCI, making it suitable for identifying early post-stroke cognitive deficits (Dong et al., 2014; Lestari et al., 2017; Pendlebury et al., 2012). MMSE is more commonly used for dementia screening and for assessing overall cognitive function (Kawada, 2018; Lyrakos et al., 2014). Additional tools included WMS-RC, TMT, AVLT, and DS. WMS-RC is widely used in China for assessing verbal and visual memory (Ding et al., 2009). TMT is sensitive to attentional and executive dysfunction and, when combined with MoCA, can improve screening accuracy for post-stroke vascular cognitive impairment (VCI) (Dong et al., 2014; Kodama et al., 2022). AVLT assesses auditory-verbal memory (Xin-Xian, 2012), while the DS evaluates working memory and correlates with daily functional status (Fitri et al., 2020; Leung et al., 2011). Overall, although MoCA and MMSE remain the primary screening tools in PISCI, they have limitations for domain-specific deficits. Future studies should incorporate more targeted assessments (e.g., AVLT, DS, TMT) and consider emerging approaches such as digital cognitive tools (Bateh et al., 2025) or virtual reality-integrated cognitive tasks (Gunawan et al., 2024) to enhance comprehensiveness and sensitivity.

In this SR, rs-fMRI and MRS were used to investigate the effects of acupuncture on brain functional activity and neurometabolic levels. It should be noted that all included studies reported within-group pre-post comparisons as their primary outcome. Two studies included healthy controls as normative references, but the primary analysis in all studies focused on within-group pre-post changes in acupuncture-treated PISCI patients. Therefore, the findings below reflect treatment-associated changes from pre- to post-treatment. However, studies varied in design (some lacked parallel control groups while others were randomized controlled trials), which should be considered when interpreting causal implications.

Rs-fMRI, a non-invasive BOLD-based technique, characterizes regional neural activity and large-scale network organization during rest (Zhen et al., 2018) and has been widely applied in studies of cognitive impairment and post-stroke brain functional reorganization (Cha et al., 2015; Golestani et al., 2013). Common rs-fMRI metrics include ALFF/fALFF (amplitude of spontaneous activity) (Wang et al., 2023), ReHo (local synchronization) (Han et al., 2018), and FC (interregional coupling) (Hua et al., 2024). FC analyses are often combined with graph theory to extract the topological features of brain networks. Frequently used graph-based indicators include DC and local efficiency. DC refers to the number of connections a given brain region has with other regions, indicating its hub-like role within the network (Kuhnert et al., 2012). Local efficiency measures the efficiency of information transmission among neighboring nodes, reflecting the integrative capacity of local subnetworks (Sang et al., 2020). Integrating these multilevel metrics provides a comprehensive framework for probing central regulatory mechanisms of acupuncture in PISCI.

MRS, a non-invasive metabolic technique, quantifies relative concentrations of brain metabolites (e.g., NAA, Cho, and MI) to assess neuronal function, glial activity, and inflammation (Tumati et al., 2013). The included studies demonstrated that acupuncture significantly increased NAA/Cr in the hippocampus and prefrontal cortex and reduced MI/Cr, suggesting improved neuronal function and attenuated inflammatory responses. One study reported a positive correlation between NAA/Cr and cognitive scores, supporting its potential as a neuroimaging biomarker for PISCI (Meng et al., 2016). It should be noted that MRS-derived metabolite ratios may vary depending on both the anatomical region examined and the post-stroke recovery phase. Previous studies included in this review measured metabolite levels from different brain regions, such as the hippocampus, frontal white matter, or peri-infarct cortex, and at various stages of stroke recovery (acute, subacute, or chronic). These regional and temporal differences can independently influence NAA, Cho, and MI concentrations, reflecting distinct metabolic and reparative processes at different stages of neural reorganization. Therefore, while the current synthesis demonstrates a generally consistent pattern, these results should be interpreted as reflecting an overall trend toward metabolic improvement rather than a precise quantitative convergence. Future studies with standardized voxel placement, acquisition parameters, and longitudinal follow-up are needed to clarify how acupuncture modulates regional and temporal metabolic dynamics after stroke.

Notably, most included studies used single-modality imaging, with limited multimodal integration, which may constrain comprehensive mechanistic understanding. Different techniques offer complementary strengths: DTI reflects post-stroke white matter microstructural damage and remodeling (Auriat et al., 2015); fNIRS, with high temporal resolution and portability, enables real-time prefrontal hemodynamic monitoring during acupuncture (Annavarapu et al., 2018); and Electroencephalogram and Magnetoencephalography (EEG/MEG) capture rapid, acupuncture-related electrophysiological activity (Hall et al., 2014). Future research should adopt multimodal strategies to elucidate acupuncture-mediated central mechanisms and provide robust evidence for precise interventions.

The included studies focused on the sustained effects of acupuncture in PISCI patients. One study revealed that acupuncture significantly enhanced hippocampus-prefrontal functional connectivity (including the middle, inferior, and superior frontal gyri). Consistent with this, a multimodal MRI study (Hosseini et al., 2016) identified hippocampal structural damage as closely associated with cognitive dysfunction and as a major predictor of post-stroke dementia. The hippocampus plays a central role in PSCI, with functional damage strongly linked to declines in memory and learning capacity (Delattre et al., 2017; Kliper et al., 2016). The prefrontal cortex supports executive function, attention, and memory, and reduced connectivity in this region has been associated with cognitive decline (Li et al., 2015b). By enhancing connectivity within this pathway, acupuncture may promote neuroplasticity and network reorganization, thereby improving cognitive function after stroke (Ding et al., 2014; Yang et al., 2014).

This review also indicates that acupuncture modulates several cognition-related regions, including the para hippocampal gyrus, fusiform gyrus, inferior occipital gyrus, cingulate gyrus, transverse temporal gyrus, lingual gyrus, and angular gyrus. The para hippocampal gyrus links the hippocampus with other cortical areas and is essential for memory encoding and spatial navigation. Studies have shown that both the hippocampus and para hippocampal gyrus often show concomitant abnormalities in PISCI and AD, which relate to memory impairment (Jung et al., 2020; Lu et al., 2024). The fusiform gyrus, implicated in visual recognition, exhibits abnormalities in PISCI (Cheng et al., 2021), and serves as an early biomarker in AD/MCI (Cai et al., 2015; Whitwell et al., 2007). The lingual and inferior occipital gyri, components of the visual network, may affect visuospatial abilities and information processing after stroke (Smagula et al., 2018). The cingulate gyrus is a hub for attention and executive control and has been associated with impaired cognitive regulation post-stroke (Lang et al., 2023). The transverse temporal gyrus, primarily involved in auditory processing, may signal language-related cognitive impairment when disrupted, particularly in left-hemisphere stroke (Delattre et al., 2017). The angular gyrus contributes to semantic processing, attentional regulation, and working memory; decreased functional connectivity of this region has been linked to lower MoCA scores in PISCI (Jiang et al., 2014). Collectively, these regions support multiple cognitive domains, including memory, executive function, attentional control, and visual perception. Dysfunction of these regions is closely associated with PISCI. The present findings suggest that acupuncture may modulate these regions and support network-level reorganization. Future work should clarify the roles of these brain areas across cognitive domains and identify the principal pathways and targets through which acupuncture exerts its effects.

Acupuncture may also modulate several functional brain networks implicated in PISCI, including the DMN, CEN, SN, and VN. The DMN, associated with high-level cognitive processes, often shows reduced connectivity after stroke, correlating with cognitive decline (Yue et al., 2023). The CEN, responsible for executive control and working memory, typically shows hypo-activation post-stroke; acupuncture may help modulate this reduction (Hoffmann, 2020). The SN functions as a regulatory switch between the DMN and CEN and frequently demonstrates impaired efficiency in PISCI; acupuncture may help restore this modulatory role (Wang et al., 2024a). The VN, which supports visual information processing, can also be disrupted after a stroke; acupuncture may aid the recovery of visual perception and related functions (Liu et al., 2017). Recent neuroimaging studies increasingly highlight abnormalities in both static and dynamic connectivity within and between the DMN, CEN, and SN in stroke-related cognitive deficits (Veldsman et al., 2020; Wang et al., 2024b; Yue et al., 2023). These findings suggest that acupuncture may support cognitive recovery by modulating network connectivity, particularly in attention, memory, and executive control. Further research is needed to clarify how acupuncture influences network interactions and contributes to reorganization after stroke.

However, it is important to note that most included studies employed rs-fMRI to examine spontaneous brain activity and functional connectivity, without task-based validation to directly link observed neural changes to specific cognitive processes. As a result, the functional interpretation of regional activation changes remains correlational and inferential. For example, while altered activity in the fusiform or lingual gyrus may be associated with visual processing, without concurrent cognitive task performance, we cannot definitively establish whether these neural changes translate into measurable improvements in visuospatial function. Future studies should incorporate task-based fMRI paradigms that probe memory, attention, executive function, and other cognitive domains to validate the functional relevance of acupuncture-induced neural modulation.

Although this SR provides a preliminary synthesis of neuroimaging evidence on acupuncture for PISCI, it is important to note that multiple factors may modulate acupuncture-induced brain functional changes.

Intervention-related parameters represent a key set of determinants. Evidence suggests that timing may critically influence neuroplasticity and functional recovery. One study reported (Li et al., 2024) that treatment initiated during the acute stage of stroke was associated with a lower risk of disability at 6 months, and another neuroimaging study (Wang et al., 2022b) showed stronger neural responses when acupuncture was applied within the 1^st month after stroke, indicating a time-dependent effect. Moreover, differences in treatment dose, such as duration, frequency, and stimulation intensity, also play an essential role. An fMRI study (Xiang et al., 2021) revealed frequency-dependent alterations in BOLD oscillations linked to analgesia, and an animal experiment demonstrated (Wang et al., 2024b) that moderate-intensity stimulation produced optimal analgesic and immune regulatory effects. Acupoint selection likewise contributes to variability. Although GV20 and EX-HN1 were commonly selected in included studies, different acupoints or combinations can elicit distinct activation patterns. Studies have indicated that (Yuan et al., 2021) stimulation of language-related acupoints induced different cortical activations, and in hypertension (Zhang et al., 2019), combined LR3+KI3 stimulation activated broader frontal, insular, and parietal regions than single-point interventions.

In addition, stroke-related clinical characteristics significantly affect the therapeutic effects of acupuncture. In acute ischemic stroke (Tsai et al., 2024), greater neurological impairment limited the extent of functional recovery achieved with acupuncture, although anti-inflammatory benefits were still observed. Lesion location is a key factor influencing acupuncture effects. A study of post-stroke dysphagia suggested that (Qiu et al., 2021) efficacy depended on lesion distribution in the brainstem and cortico-subcortical extension region, and another study reported (Zhang et al., 2024) distinct acupuncture-induced activation patterns across infarct locations. Furthermore, ischemia duration determines the degree of neuronal necrosis and the therapeutic window. An animal experiment demonstrated that (Guo et al., 2025) acupuncture administered within 24 h after ischemia significantly reduced infarct volume and improved neurological scores, whereas delayed intervention produced weaker effects.

Overall, the included studies did not sufficiently address these factors, limiting understanding of neural effects. Future research should incorporate intervention parameters and stroke characteristics into study design and adhere to STRICTA and the Consolidated Standards of Reporting Trials (CONSORT) guidelines to ensure transparency and scientific rigor.

A study of neuroimaging in AD (Yin et al., 2023a) reported acupuncture-induced activation in the hippocampus, prefrontal cortex, and parietal regions, along with enhanced integration of the DMN, CEN, and fronto parietal network (FPN). Another review focused on MCI (Yin et al., 2023b) highlighted modulation of the anterior cingulate cortex and insula, involving the DMN and SN. An fMRI meta-analysis (Ma et al., 2022) further revealed increased activation in the thalamic, frontal, and cingulate regions in MCI after acupuncture, suggesting potential benefits for attention and executive function. While broadly consistent, these studies differ in focus. The present review targets PISCI and indicates modulation of visual-related brain regions (lingual gyrus, inferior occipital gyrus, fusiform gyrus), which are less reported in AD or MCI. This discrepancy may reflect the heterogeneity of stroke lesions, which can affect motor, language, and visual systems in addition to memory, complicating cognitive assessment (Wang, 2020). Thus, cognitive impairment after stroke may involve multimodal sensory and attentional deficits beyond memory decline. These findings contribute to a more comprehensive understanding of the neural mechanisms underlying acupuncture in PISCI and provide a comparative framework for evaluating differential effects across cognitive disorders. Future research should include cross-condition comparisons, multimodal imaging, and task-based designs to better characterize the neural substrates of acupuncture.

This review has several limitations. First, most included studies had small sample sizes and heterogeneous designs, potentially affecting the generalizability. Second, key factors such as stroke type, lesion location, recovery phase, concurrent rehabilitation programs, and pharmacotherapy were rarely reported or controlled across studies, which may influence brain reorganization and intervention efficacy. Without systematic control of these variables, the specific contribution of acupuncture may be overstated. Future studies should employ stricter protocols for documenting and adjusting for these clinical factors. Third, variability in imaging protocols, analysis methods, and outcome measures precluded meta-analysis or quantitative integration.

Fourth, only eight studies were included in the final analysis, and several were dissertations, which may limit the representativeness and robustness. Fifth, formal assessment of publication bias was not feasible due to the small number of studies and the absence of pooled quantitative data; nevertheless, the predominance of positive findings suggests that potential publication bias cannot be ruled out. Finally, all included studies were conducted in China, which may limit the generalizability due to potential cultural and methodological factors. Cultural factors such as patient expectations and beliefs about acupuncture, which vary across populations, may influence placebo responses and neural patterns. Acupuncture practice varies internationally in practitioner training, technique standardization, and acupoint selection, potentially affecting neural responses. Cross-country differences in neuroimaging protocols and population characteristics (e.g., stroke etiology, genetic backgrounds) may also impact comparability. While the consistency across multiple independent Chinese studies supports reliability, international multicenter studies with standardized protocols are needed to validate whether these mechanisms generalize across diverse populations and cultural contexts.