Application and Research Progress of Brain-Computer Interface for Post-Stroke Psychiatric Disorders: A Narrative Review

Introduction

Stroke remains one of the leading causes of disability and mortality worldwide, with survivors facing a substantial risk of developing neuropsychiatric complications such as depression, anxiety, and cognitive impairment. These post-stroke mental health disorders are highly prevalent, with studies indicating that approximately a third of stroke survivors experience depression or anxiety, and a significant proportion develop cognitive deficits, even years after the acute event [1–3]. The coexistence of these mental health issues with physical impairments not only impedes the rehabilitation process but also profoundly diminishes quality of life and long-term functional outcomes for patients. Notably, cognitive and emotional impairments can persist or even worsen over time, with attention, executive function, memory, and mood disturbances remaining stable or deteriorating in the chronic phase of stroke recovery [3,4]. Furthermore, these psychological sequelae may occur independently of the severity of physical disability, highlighting the need for comprehensive, multidisciplinary approaches to stroke rehabilitation [5–7]. Despite the high burden, traditional interventions such as pharmacotherapy and psychotherapy often yield limited efficacy in this population, and are frequently accompanied by adverse effects, adherence challenges, and underdiagnosis or undertreatment [8,9].

Recent advances in neurotechnology have introduced brain-computer interfaces (BCIs) as a promising frontier for the assessment and intervention of post-stroke mental disorders. BCIs are sophisticated systems that enable direct communication between the brain and external devices, leveraging real-time monitoring, decoding, and modulation of neural activity to facilitate functional restoration and neurorehabilitation [7,10,11]. The theoretical foundation for applying BCI in this domain rests on the link between specific psychiatric symptoms and quantifiable neural markers. For instance, post-stroke depression is frequently associated with frontal alpha asymmetry and disrupted functional connectivity, while cognitive impairment is often reflected in attenuated event-related potentials (ERPs). BCI systems leverage these specific biomarkers to create a closed feedback loop, enabling the direct targeting and modulation of the underlying neurophysiological dysfunctions responsible for psychiatric sequelae. The rapid development of BCI technologies has expanded their clinical applications from motor recovery to the management of cognitive and mental disorders, including those arising after stroke. Emerging research demonstrates that BCIs can be utilized for objective assessment of brain function, early detection of neuropsychiatric complications, and the delivery of targeted neurofeedback or neuromodulation therapies aimed at enhancing neuroplasticity and ameliorating symptoms of depression, anxiety, and cognitive decline [10,12–14]. For example, EEG-based neurofeedback interventions have shown efficacy in improving cognitive domains and modulating affective symptoms in stroke and other neurological conditions, while closed-loop BCI systems offer the potential for individualized, adaptive rehabilitation protocols [12–14]. Importantly, the integration of artificial intelligence (AI), advanced signal processing, and non-invasive brain monitoring technologies further enhances the precision and utility of BCIs in both research and clinical settings [15–17]. Recent studies have particularly highlighted the efficacy of deep learning models and fuzzy expert systems in decoding complex physiological signals and recognizing motor patterns. These technical advancements provide robust frameworks that are increasingly relevant for post-stroke neurorehabilitation [18,19].

Given the limitations of conventional treatments and the complex interplay between neurobiological, psychological, and social factors in post-stroke mental disorders, BCIs represent a promising approach that addresses unmet clinical needs.

However, the implementation of BCI-based interventions in this context is not without challenges. Technical barriers such as signal reliability, user adaptability, and long-term efficacy remain to be fully addressed [7,11,20]. Crucially, a distinction must be made between experimental laboratory systems and clinical-grade devices. While high-density, complex laboratory setups drive theoretical innovation and signal resolution, their applicability is often limited by setup time and cost. In contrast, emerging clinical-grade BCIs prioritize portability, ease of use, and robust signal processing algorithms, representing the necessary bridge for translating BCI technology from controlled experiments to routine bedside rehabilitation. Additionally, ethical considerations regarding patient autonomy, data privacy, and equitable access are increasingly recognized as critical in the deployment of BCI technologies for vulnerable populations [21–23]. Despite these hurdles, the convergence of neuroscience, engineering, and clinical rehabilitation offers new opportunities for innovation.

Consequently, this narrative review aims to comprehensively synthesize current knowledge on the application of BCIs in this domain, encompassing underlying mechanisms, clinical research findings, and technical challenges. By outlining future directions, the study seeks to bridge the gap between technological innovation and clinical practice. With the global burden of stroke rising, providing such a comprehensive overview is timely and essential for establishing a robust theoretical framework to guide future neurorehabilitation paradigms. This review will provide an overview of the literature on mechanisms, applications, and challenges of the use of BCI for post-stroke psychiatric disorders (PSPD).

Epidemiology and Clinical Characteristics of Post-Stroke Psychiatric Disorders

COMMON TYPES AND EPIDEMIOLOGICAL DATA:

PSPD, particularly depression, anxiety, and cognitive impairment, represent significant complications that affect a substantial proportion of stroke survivors. Epidemiological data indicate that depression is one of the most common mental health issues following stroke, with prevalence rates varying due to differences in diagnostic criteria, assessment timing, and population characteristics, but generally regarded as high. Depression after stroke is multifactorial, impacting emotional, cognitive, and physical functioning, and it often manifests as insomnia, weight changes, and diminished concentration, underscoring its heterogeneous nature and the individualized symptom profiles observed among patients [24]. Anxiety disorders are also frequently observed in the post-stroke population, with studies highlighting a notable prevalence, especially among women, and an association with increased risk for subsequent cardiovascular events. Cognitive impairment, which may range from mild cognitive deficits to full-blown dementia, is another common sequela of stroke, further compounding the burden of psychiatric morbidity and adversely affecting rehabilitation outcomes and quality of life.

The distribution of these disorders is influenced by a range of risk factors. The location and severity of the stroke are critical determinants; lesions affecting the frontal lobe, basal ganglia, or left hemisphere have been particularly implicated in the development of post-stroke depression and cognitive impairment. Greater stroke severity is consistently associated with higher rates of psychiatric complications. Social determinants, such as the level of social support, play a pivotal role in modulating risk; individuals with limited social networks or inadequate support systems are more vulnerable to developing depressive and anxiety symptoms post-stroke. Furthermore, comorbidities such as obesity, diabetes, and cardiovascular disease not only increase the risk of stroke but also independently contribute to the development of psychiatric disorders, pointing to the complex interplay between physical and mental health in this population [25].

In addition to depression, anxiety, and cognitive impairment, post-stroke fatigue (PSF) represents a critical yet often overlooked neuropsychiatric sequela. A recent study highlights that PSF is a multidimensional construct comprising distinct physical and cognitive subtypes, which are highly prevalent and differentially associated with clinical factors such as sleep quality, metabolic comorbidities, and stroke characteristics [26]. Importantly, while PSF frequently co-occurs with depression and anxiety, it possesses unique pathophysiological signatures. Integrating PSF into the BCI rehabilitation framework is therefore essential, as emerging BCI technologies could potentially offer objective metrics to differentiate cognitive fatigue from physical exhaustion and provide targeted neurofeedback to specifically address these fatigue domains.

In summary, the high prevalence and multifactorial etiology of PSPD underscore the need for integrated, multidisciplinary approaches to screening, prevention, and management in stroke survivors.

CLINICAL MANIFESTATIONS AND DIAGNOSTIC CRITERIA:

The clinical manifestations of PSPD are diverse, encompassing a spectrum that includes depression, anxiety, cognitive impairment, and delirium. Each of these conditions presents with distinct symptomatology and requires specific diagnostic approaches. For instance, post-stroke delirium can manifest as hyperactive, hypoactive, or mixed subtypes, each with varying predisposing factors and clinical characteristics. The Diagnostic and Statistical Manual of Mental Disorders (DSM-5) criteria are widely used to diagnose delirium, with studies demonstrating that all delirium subtypes after ischemic stroke are associated with poor functional outcomes and increased mortality risk, particularly for hypoactive and mixed types [27]. Cognitive impairment, another common complication, is frequently assessed using neuropsychological batteries and standardized tools such as the Mini-Mental State Examination (MMSE). In a cross-sectional survey of older adults, the DSM-5 criteria were employed to classify patients into none, mild, or major cognitive impairment, revealing a high prevalence of concomitant deficits across multiple cognitive domains, including memory, language, attention, and visuospatial function [28]. Tools like the Hospital Anxiety and Depression Scale are also integral in screening for anxiety and depression, which are prevalent post-stroke and can significantly impact rehabilitation engagement and outcomes. The severity and presence of psychiatric disorders are not only associated with poorer rehabilitation outcomes but also with increased risk of long-term disability and mortality, underscoring the importance of early detection and intervention. Moreover, the integration of advanced neuroimaging and AI-based assessments is facilitating more precise and quantitative evaluations of neuropsychiatric sequelae, providing clinicians with robust decision-making support. Collectively, these diagnostic standards and clinical observations highlight the critical need for comprehensive mental health assessment in post-stroke populations to optimize recovery trajectories and long-term prognosis.

Basic Principles and Types of Brain-Computer Interfaces

BCI WORKING PRINCIPLE:

The working principle of the BCI fundamentally revolves around 3 core components: brain signal acquisition, signal processing and decoding, and external feedback mechanisms. The first step, brain signal acquisition, involves capturing neural activity from the brain using various modalities such as electroencephalography (EEG), electrocorticography (ECoG), or intracortical electrodes, depending on whether the interface is noninvasive or invasive. EEG remains the most commonly used noninvasive technique due to its practicality and safety, allowing for the detection of electrical activity from the scalp with specialized electrodes, though with limitations in spatial resolution compared with invasive methods [29–31]. After acquisition, the raw neural signals are subjected to a series of preprocessing steps – such as filtering and artifact removal – to enhance the signal-to-noise ratio and extract relevant features. Signal processing algorithms, often incorporating advanced techniques from AI and machine learning, are then employed to decode the user’s intentions or mental states from these features [15,32,33]. The decoding process translates the complex, multidimensional brain signals into actionable commands or control outputs, such as cursor movements, robotic limb control, or communication with external devices [34–36]. The final stage, external feedback mechanisms, closes the loop by providing sensory or visual feedback to the user based on the system’s output, enabling real-time interaction and adaptation. This feedback can be visual, auditory, or even tactile, and is crucial for facilitating user learning, improving control accuracy, and fostering neuroplasticity [34,37]. Together, these integrated components allow BCI to establish a direct communication pathway between the brain and external devices, bypassing traditional neuromuscular output channels and offering transformative potential for individuals with neurological impairments [38,39]. The integrated framework of these components is illustrated in Figure 1.

COMMON TYPES OF BCI:

BCI can be categorized into non-invasive, semi-invasive, and invasive types, each with distinct characteristics and suitable applications. Non-invasive BCI, such as approaches based on EEG, are the most widely used types due to their safety, portability, and ease of use. EEG-BCI records brain activity from the scalp using electrodes, making it suitable for a broad range of applications including neurorehabilitation after stroke, cognitive assessment, and assistive communication for individuals with severe motor impairments [29,40,41]. Advantages of these BCI approaches include their non-invasiveness, relatively low cost, and rapid setup; however, they suffer from limited spatial resolution and susceptibility to noise and artifacts, which can constrain the accuracy and reliability of signal decoding [42,43]. Semi-invasive BCI, such as ECoG, involves placing electrodes directly on the cortical surface beneath the skull but above the brain tissue. ECoG-based BCI offers a compromise between invasiveness and signal quality, providing higher spatial and temporal resolution than EEG while reducing some of the risks associated with fully invasive approaches. These systems are particularly valuable in clinical and research settings where more precise neural decoding is required, such as in the restoration of communication or motor functions in patients with severe neurological disorders [44]. Invasive BCI, or intracortical BCI, involves implanting microelectrodes directly into the brain tissue. This approach yields the highest fidelity signals, enabling fine-grained decoding of motor intentions, speech, and even complex finger movements, which is crucial for advanced prosthetic control and communication in patients with paralysis [45–47]. Despite their superior performance, invasive BCI approaches carry significant risks, including infection, tissue reaction, and long-term device degradation, which limit their widespread clinical adoption. The choice among these BCI types depends on the specific clinical scenario, patient needs, and the required balance between invasiveness, signal quality, and safety. For example, non-invasive EEG-BCI is preferred for routine neurorehabilitation and cognitive monitoring, while invasive BCI is reserved for cases where high-resolution control is essential and the potential benefits outweigh the surgical risks [48,49]. For a comprehensive comparison of these signal acquisition methods, including their specific advantages, limitations, and typical clinical applications, please refer to Table 1.

FOUNDATION OF BCI APPLICATIONS IN NEUROPSYCHIATRIC DISORDERS: The theoretical foundation and early research on the application of BCI technology in neuropsychiatric disorders – including depression, anxiety, and cognitive impairment – are grounded in the ability of BCI to monitor, decode, and modulate neural activity relevant to mental health. BCI leverages electrophysiological signals, most commonly EEG, to detect biomarkers associated with psychiatric symptoms, enabling both diagnostic and therapeutic interventions. For example, multimodal BCI capable of recording and stimulating prefrontal neural networks has been shown to capture and modulate patterns linked to cognitive dysfunction and behavioral control, which are central features in disorders such as depression and schizophrenia. These systems can also employ machine learning algorithms to classify neural patterns and adapt interventions in real time, offering a personalized approach to therapy [50]. In anxiety disorders, BCI integrated with advanced AI models, such as convolutional and recurrent neural networks, can enhance the identification of EEG biomarkers and facilitate real-time monitoring and feedback, which is crucial for neurorehabilitation and therapy response prediction [51]. Similarly, in cognitive disorders, BCI has demonstrated the capacity to modulate memory functions through direct electrical stimulation, with closed-loop systems providing responsive neuromodulation based on detected electrophysiological biomarkers [52]. The application of BCI extends to neurofeedback, in which individuals learn to control their brain activity to improve cognitive performance or alleviate psychiatric symptoms, including in patients with disorders of consciousness [53]. Furthermore, wearable neural interfaces and biosensors enable continuous, real-time monitoring of mental states, supporting early detection and intervention in conditions such as depression, anxiety, and dementia [54]. Collectively, these advances underscore the growing potential of BCI approaches as both diagnostic and therapeutic tools for neuropsychiatric disorders, driven by ongoing developments in neural signal processing, machine learning, and adaptive neuromodulation. Early research highlights the promise of BCI in providing objective, individualized, and dynamic approaches to the management of mental health conditions, setting the stage for broader clinical translation and integration into psychiatric care [55–57].

Application of Brain-Computer Interface in Post-Stroke Depression

ASSESSMENT AND MONITORING:

The application of BCI technology in the objective assessment and monitoring of post-stroke depression and emotional states represents a promising frontier in neurorehabilitation. BCI systems, often utilizing EEG, provide a non-invasive and real-time window into brain activity, enabling the quantification of mental states such as stress and mood disturbances. Recent advancements in EEG technology – including increased electrode density, innovative analytical approaches, and multimodal integration – have significantly enhanced the precision and clinical utility of these assessments, thereby broadening their role beyond traditional uses like preoperative evaluation and epilepsy monitoring. For instance, high-density EEG combined with machine learning algorithms has demonstrated the ability to accurately classify mental stress states, as evidenced by studies employing deep learning classifiers such as long short-term memory (LSTM) and recurrent neural networks (RNNs). Such methods have achieved high accuracy, specifically distinguishing stress levels with classification rates of up to 93–94%, as demonstrated in deep learning models applied to ongoing EEG signals [58]. Although these findings were initially validated in populations such as adolescents with autism spectrum disorder, the underlying methodology is highly translatable to stroke survivors, who often experience comorbid depression and anxiety [58]. Furthermore, the integration of EEG-based BCI with other neuromodulation techniques and neuroimaging modalities – such as functional near-infrared spectroscopy (fNIRS) – enables real-time cortical monitoring and feedback, establishing a closed-loop system that supports precise, individualized neurorehabilitation strategies. This approach not only facilitates the objective assessment of emotional states but also provides continuous monitoring, which is essential for timely intervention and adjustment of therapeutic regimens in post-stroke patients [59,60]. Collectively, these technological advancements underscore the emerging role of BCI systems as valuable auxiliary tools for the objective evaluation and dynamic monitoring of emotional and psychological disturbances following stroke, paving the way for more targeted and effective rehabilitation interventions.

INTERVENTION AND TREATMENT:

BCI technologies, when combined with neurofeedback training, have shown increasing promise as innovative, non-invasive interventions for improving depressive symptoms following neurological injuries such as stroke. Mechanistically, BCI-based neurofeedback enables patients to gain voluntary control over specific brain rhythms associated with mood regulation. By providing immediate feedback, this closed-loop system allows patients to potentially correct dysfunctional neural patterns implicated in depression, fostering neuroplasticity and functional recovery. Recent studies emphasize that the integration of BCI and neurofeedback can optimize the reorganization of neural networks, fostering neuroplasticity and functional recovery in patients with central nervous system disorders, including post-stroke depression. For instance, BCI-based neurofeedback has been applied to modulate affective states, such as using real-time feedback from brain responses to auditory or visual stimuli, which can result in statistically significant reductions (P<0.05) in depression severity scores and measurable improvements in cognitive function [61]. These effects are often linked to changes in specific EEG frequency bands, such as beta and theta rhythms, which are associated with emotional and cognitive processes. Furthermore, advancements in neuroimaging and multimodal feedback, such as fNIRS combined with BCI, have enabled more precise and individualized interventions by monitoring cortical responses and establishing closed-loop strategies for neurorehabilitation [60]. Systematic reviews highlight that while neurofeedback protocols – such as Slow Cortical Potential neurofeedback – show efficacy in enhancing self-regulation, the translation of these neural gains into clinical symptom improvement remains complex and requires standardized methodologies for evaluation [62]. Clinical trials, including randomized controlled designs, have demonstrated that BCI-neurofeedback interventions can lead to significant decreases in depression scores, such as the Hamilton Rating Scale for Depression (HAM-D), compared with control conditions, supporting their therapeutic potential [61]. However, the literature also underscores the need for larger, well-controlled studies to confirm these findings, optimize intervention protocols, and clarify the mechanisms linking neural modulation to behavioral and emotional outcomes. Overall, BCI combined with neurofeedback represents a promising direction for the treatment of post-stroke depression, offering the potential for personalized, adaptive, and effective neurorehabilitation strategies.

MECHANISTIC EXPLORATION:

Recent advances in BCI technology have provided new insights into the mechanisms by which BCI interventions may modulate the prefrontal-limbic circuitry, thereby promoting neuroplasticity and facilitating emotional regulation after stroke. The prefrontal cortex, in concert with limbic structures such as the amygdala and hippocampus, forms a critical network underlying affective processing and executive control. Disruption of this circuitry following stroke often leads to neuropsychiatric sequelae, including depression and anxiety. BCI-based neurorehabilitation strategies, particularly those leveraging EEG and related neuroimaging modalities, have demonstrated the ability to provide real-time feedback and closed-loop modulation of neural activity within these networks, fostering adaptive reorganization and synaptic plasticity. For example, the integration of EEG with BCI enables the detection and classification of stress-related neural signatures, allowing for targeted interventions such as respiratory entrainment to mitigate mental stress and potentially recalibrate dysfunctional prefrontal-limbic interactions [58]. Moreover, the application of multimodal approaches – combining BCI with non-invasive neuromodulation techniques like transcranial direct current stimulation or fNIRS – has been shown to enhance cortical excitability and facilitate the reorganization of neural networks implicated in both motor and emotional recovery [14,60]. These interventions can induce neuroplastic changes not only at the site of injury but also in remote, functionally connected regions, underscoring the network-level impact of BCI-based therapies [10]. Additionally, advances in closed-loop BCI systems, which adaptively respond to ongoing neural activity, offer the potential for individualized, precise modulation of prefrontal-limbic circuits, thereby optimizing emotional regulation and functional recovery [7]. Collectively, these findings highlight the promise of BCI technologies in harnessing neuroplasticity and facilitating the restoration of the dynamic balance of prefrontal-limbic circuitry, paving the way for innovative interventions in the management of PSPD.

Application of Brain-Computer Interface in Post-Stroke Cognitive Impairment and Anxiety Disorders

BCI-ASSISTED REHABILITATION FOR COGNITIVE IMPAIRMENT:

Recent advances in BCI technology have significantly expanded its application from motor rehabilitation to the domain of post-stroke cognitive impairment, particularly in the areas of cognitive training, working memory, and attention. BCI can translate brain activity into computer-recognized instructions, enabling tailored cognitive training paradigms that bypass motor or verbal limitations, which is especially beneficial for stroke survivors with severe disabilities [41]. Studies have demonstrated that BCI-based neurofeedback and cognitive training can enhance attention and working memory in stroke patients. For instance, neurofeedback training using EEG-based BCI games has shown measurable improvements in attention and working memory scores among both stroke patients and patients with mild cognitive impairment, as evidenced by increased EEG-based attention scores and improved task performance over repeated sessions [63]. A randomized controlled trial specifically assessing P300-based BCI cognitive training found significant improvements in attention, naming, and abstraction domains on the MoCA scale. Notably, the Attention domain scores in the BCI group increased from 2.3±1.24 to 5.2±1.16 (P<0.05) post-intervention, significantly outperforming the traditional rehabilitation group [64]. These behavioral gains are supported by neuroimaging and electrophysiological evidence. Functional MRI and EEG studies reveal that BCI-based attention training can rapidly modulate functional brain connectivity in stroke patients, strengthening neural circuits associated with attention and executive control [65]. Moreover, EEG and fNIRS-based multimodal BCI have been used to identify neural markers of cognitive improvement, such as increased P300 amplitude and changes in late positive potential, which correlate with enhanced cognitive processing in both healthy and cognitively impaired populations [66]. Integration of BCI with virtual reality and multimodal feedback further accelerates sensory-motor and cognitive pathway facilitation, providing immersive and engaging contexts for cognitive training [67]. Collectively, these findings indicate that BCI-assisted cognitive rehabilitation is a promising avenue for improving attention, working memory, and overall cognitive function in post-stroke patients, supported by preliminary behavioral and neuroimaging evidence. However, future research should focus on optimizing training protocols, integrating AI for individualized adaptation, and conducting larger-scale clinical trials to validate long-term efficacy [68,69].

BCI INTERVENTION FOR ANXIETY DISORDERS:

The integration of BCI technology with relaxation training and meditation has shown promising effects in alleviating anxiety symptoms, particularly through non-pharmacological, neurophysiological interventions. EEG-based BCI facilitates real-time feedback during relaxation or mindfulness practices, allowing individuals to learn to regulate arousal-related brain activity. This process targets the physiological over-arousal characteristic of anxiety disorders, often leading to measurable symptom improvements [70]. Randomized controlled trials further support the efficacy of BCI-based interventions: in one RCT, participants with anxiety and depressive spectrum disorders who underwent BCI-guided audio-visual entrainment experienced significant reductions in anxiety and depression scores, as measured by standardized rating scales, compared with active control groups. These improvements were associated with specific changes in EEG frequency bands, suggesting a neurophysiological basis for the observed therapeutic effects [61]. Additionally, closed neurofeedback systems using EEG have been shown to be able to efficiently classify and detect anxiety states, achieving an average classification accuracy of 87.18% in patients with anxiety disorders, thus providing objective assessment tools and potentially enhancing the personalization of interventions [71]. Systematic reviews emphasize that combining BCI with virtual reality or other immersive environments can further enhance the efficacy of relaxation and meditation-based protocols, leveraging neuroplasticity for cognitive and emotional rehabilitation, including the treatment of phobias and anxiety disorders [72]. However, methodological limitations, such as sample size and heterogeneity in study design, highlight the need for further large-scale, high-quality RCTs to validate these findings and optimize BCI protocols for clinical use in anxiety disorders [70].

MULTIMODAL BCI AND COMPREHENSIVE INTERVENTION FOR MENTAL DISORDERS:

The integration of multimodal BCI interventions, particularly those that combine EEG and fNIRS, represents a promising frontier in the comprehensive intervention of post-stroke mental disorders. EEG-based BCI has already demonstrated significant potential in real-time assessment and mitigation of mental stress, as evidenced by deep learning approaches capable of distinguishing stress states with high accuracy in diverse populations, including adolescents with neurodevelopmental disorders [58]. However, EEG alone is sometimes limited by its spatial resolution and susceptibility to artifacts. The addition of fNIRS, a non-invasive optical imaging technology that measures cerebral hemodynamics, addresses these limitations by providing complementary information on cortical activation and allowing for robust monitoring even in motion-rich environments [60]. This synergy enables more precise, individualized neurofeedback and closed-loop interventions, as fNIRS can offer real-time feedback on cortical responses to neurostimulation or behavioral interventions, thereby optimizing the modulation of neural circuits involved in mood and cognition. Such multimodal approaches are especially relevant for post-stroke patients, who often experience a complex interplay of motor and psychiatric symptoms that require nuanced, adaptive therapeutic strategies. By leveraging the strengths of both EEG and fNIRS, multimodal BCI can facilitate the development of tailored interventions that not only enhance neuroplasticity but also provide ongoing evaluation and adjustment of therapeutic protocols. This integrated methodology supports the overarching goal of individualized, precise neurorehabilitation and holds promise for improving functional recovery and quality of life in patients with central nervous system diseases, including those suffering from post-stroke psychiatric sequelae [60]. Future research should continue to refine these technologies, establish standardized protocols, and explore their efficacy in large-scale clinical trials to fully realize their potential in comprehensive mental health interventions. To provide a structured overview of the clinical evidence discussed in this section, Table 2 summarizes representative studies, detailing their target disorders, technical features, and key outcomes.

Challenges and Future Directions in the Application of Brain-Computer Interface Technology

TECHNICAL CHALLENGES:

The application of BCI technology in the rehabilitation of PSPD has several significant technical challenges, including signal noise, individual variability, and long-term user compliance. First, the issue of signal noise remains a core obstacle in BCI systems, especially when using EEG as the primary signal acquisition modality. Although EEG offers excellent temporal resolution, its spatial resolution is relatively poor, and the signals are highly susceptible to various artifacts such as muscle movements, environmental electromagnetic interference, and electrode displacement. Recent advancements, such as the adoption of high-density EEG systems and innovative signal analysis techniques, have improved the situation, but noise remains a limiting factor for real-time and accurate BCI applications in clinical settings [59]. Second, individual differences in neurophysiological responses pose another major challenge. Variability in brain structure, neuroplasticity, and the extent of stroke-induced damage can lead to significant heterogeneity in BCI performance and rehabilitation outcomes. This necessitates the development of highly individualized BCI protocols and adaptive algorithms to ensure efficacy across diverse patient populations [10]. Moreover, machine learning-based classifiers, though promising for tasks such as mental state detection, often require subject-dependent models, further complicating the generalizability and scalability of BCI solutions [58]. Third, ensuring long-term compliance and usability is critical yet problematic. Many BCI systems require frequent calibration, are cumbersome to set up, or demand sustained attention and engagement from users, which can be particularly challenging for stroke survivors with cognitive or psychiatric comorbidities. The lack of standardized protocols and the need for continuous technical support also hinder widespread adoption and consistent long-term use in real-world rehabilitation environments [7]. Addressing these technical challenges through interdisciplinary collaboration, advances in multimodal signal integration, and the development of more user-friendly, robust systems is essential for the successful clinical translation and sustained application of BCI technology in post-stroke psychiatric rehabilitation.

BARRIERS TO CLINICAL TRANSLATION:

Despite the promising advancements in BCI technologies for neurorehabilitation and the treatment of post-stroke neuropsychiatric disorders, significant barriers remain in the clinical translation of these innovations. One major limitation is the lack of large-sample, multicenter clinical validation studies. Most current research is restricted to small cohorts or single-center trials, which hinders the generalizability and reproducibility of results. This limitation is highlighted in recent studies, where the need for larger-scale investigations is emphasized to establish robust evidence for the efficacy and safety of BCI-based interventions, especially in chronic post-stroke rehabilitation [14]. Furthermore, while technological advances such as high-density EEG, innovative analysis techniques, and integration with other modalities have expanded the clinical potential of BCI, the absence of standardized protocols and long-term outcome data presents additional challenges for widespread clinical adoption [7,59]. Ethical and safety concerns also pose significant obstacles. The direct interaction between neural systems and external devices raises questions regarding data privacy, informed consent, and the potential for unintended neurophysiological effects. The need for interdisciplinary collaboration and the development of evidence-based guidelines is repeatedly underscored in the literature, as these are crucial for addressing ethical dilemmas and ensuring patient safety [7]. Additionally, the lack of clear regulatory frameworks and consensus on best practices further complicates the path to clinical translation. In conclusion, overcoming these translational barriers will require concerted efforts to conduct large, multicenter clinical trials, establish standardized protocols, and address ethical and safety issues through comprehensive guidelines and interdisciplinary collaboration. Only through such measures can the full therapeutic potential of BCI in post-stroke neuropsychiatric disorders be realized in routine clinical practice [7,10,59].

FUTURE DEVELOPMENT TRENDS:

The future of BCI systems in the management of PSPD is characterized by a marked shift towards intelligent and individualized solutions, driven by the integration of AI and big data analytics. The application of advanced machine learning algorithms, such as deep learning models, has already demonstrated impressive accuracy in real-time mental state classification based on EEG signals, enabling closed-loop interventions tailored to individual neurophysiological responses [58]. This trend is expected to continue, with AI-powered BCI offering highly adaptive and personalized therapeutic strategies that can dynamically adjust to the evolving needs of post-stroke patients. In addition, the increasing availability of high-density EEG and multimodal neuroimaging, combined with sophisticated data analysis techniques, is poised to enhance the spatial and temporal resolution of brain monitoring, further supporting precision medicine approaches in neurorehabilitation [59].

Interdisciplinary collaboration is another key driver propelling the clinical application of BCI in PSPD. The convergence of neuroscience, engineering, computer science, clinical medicine, and rehabilitation science is fostering the development of more robust, user-friendly, and effective BCI systems [7]. Such collaboration not only accelerates technological innovation but also facilitates the establishment of standardized protocols and evidence-based guidelines, which are essential for the widespread adoption of BCI in clinical practice [10]. Furthermore, the integration of non-invasive neuromodulation technologies and real-time neurofeedback mechanisms – guided by advanced neuroimaging modalities like fNIRS – opens new avenues for individualized, closed-loop neuromodulation strategies [60]. These developments collectively underscore a future in which BCI systems are not only more intelligent and personalized but also potentially integrated into multidisciplinary care pathways, ultimately improving functional and psychiatric outcomes for stroke survivors.

Crucially, the clinical implementation of these technologies must align with the framework of the ‘Individual Rehabilitation Project’ as advocated by the European Union of Medical Specialists section on Physical and Rehabilitation Medicine (UEMS-PRM). This approach ensures that BCI interventions are not isolated technical exercises but are embedded within a holistic, patient-centered rehabilitation plan that prioritizes the specific functional goals and quality of life of each stroke survivor [73].

Conclusions

This narrative review summarizes the evolving role of BCI in PSPD, proposing an integrated framework of “Closed-Loop Precision Neurorehabilitation”. Unlike traditional interventions that operate in isolation, BCI systems bridge the gap between neurophysiological dysfunctions and behavioral sequelae by establishing a direct, adaptive feedback loop. This framework posits that BCI does not merely alleviate symptoms but actively drives the restoration of the prefrontal-limbic circuitry through Hebbian plasticity, offering a neurobiologically grounded alternative to the “trial-and-error” nature of pharmacotherapy.

From a translational perspective, current evidence suggests that non-invasive EEG-based BCI modalities are currently the closest to routine clinical implementation due to their safety profile and portability. Clinically, these systems appear most beneficial for stroke survivors with residual cognitive capacity but severe motor impairments, or those exhibiting pharmaco-resistance to traditional antidepressants. For these patient profiles, BCI should be considered a promising adjunctive therapy to standard rehabilitation.

Despite these advancements, a critical evaluation reveals significant gaps hindering widespread adoption. While immediate neural modulation effects are well-documented, the long-term durability of these neuroplastic changes remains unproven. Most current findings rely on small-scale pilot studies with heterogeneous protocols, limiting their generalizability. Furthermore, existing research heavily prioritizes electrophysiological metrics and clinical scales over functional outcome measures. There is a paucity of data concerning whether BCI-driven improvements translate into tangible benefits in Activities of Daily Living or social participation.

Consequently, future research must pivot from proof-of-concept studies to large-scale, multicenter randomized controlled trials. These trials should prioritize functional endpoints alongside neural markers to validate real-world impact. Ultimately, the successful translation of BCI technology depends on moving beyond the laboratory to create robust, user-friendly systems embedded within Individual Rehabilitation Projects, ensuring that technological innovation directly translates into improved quality of life for stroke survivors.