Brain-computer interfaces: the innovative key to unlocking neurological conditions

Neurological disorders contribute substantially to global mortality and disability, with stroke and other nervous system diseases representing leading causes of death and disability. As populations age, the incidence and burden of neurological conditions are expected to grow, leading to significant social, economic, and quality-of-life impacts for patients and caregivers. Traditional therapies often fall short in addressing complex neurological dysfunctions, underscoring the need for innovative approaches. Brain-computer interface (BCI) technology enables direct, real-time decoding of brain signals and translation into device commands, offering diagnostic insights, therapeutic restoration of function, and rehabilitation via targeted activation of neural circuits. The integration of AI and machine learning promises more adaptive, personalized BCIs. This review synthesizes the evolution, classification, working principles, and paradigms of BCI technology, and critically examines recent clinical applications across motor disturbances, disorders of consciousness, cognitive/mental disorders, and sensory disorders, highlighting prospects, ethical considerations, and the pivotal role of neurosurgery in clinical translation.

The review traces BCI development through three phases: Academic Exploration (1924–1970), Scientific Validation (1970–2000), and Experimental Application (2001–present). BCIs are classified by invasiveness (non-invasive: EEG, fMRI, fNIRS, MEG; semi-invasive: ECoG; invasive: SEEG) and by directionality (unidirectional vs bidirectional). It details the BCI pipeline—signal acquisition, preprocessing, feature extraction, classification (SVM, LDA, ANN, deep learning), device control, and feedback (VR, robotics, haptics). Paradigms covered include motor imagery (MI-BCI), SSVEP, auditory evoked potentials (AEP), and P300. Neuroscience fundamentals are summarized (CNS/PNS structure, neuronal signaling, synaptic transmission). The review compiles recent clinical and translational studies across neurological indications: PD and DBS advancements (including closed-loop concepts), stroke rehabilitation using MI-BCI and hybrid systems, SCI motor recovery via multimodal BCIs and implanted arrays, communication solutions for locked-in syndrome (eye-tracking, HMM/DNN, FES integration), epilepsy seizure prediction/localization and responsive neurostimulation, DoC assessment and communication (EEG/fMRI, ERPs, P300 spellers), cognitive rehabilitation in AD (EEG biomarkers, neurofeedback, conditioning-based BCIs), depression (DBS targets and EEG-based BCIs), and sensory disorders (auditory and visual BCIs, speech synthesis from ECoG).

This article is a comprehensive descriptive/narrative review synthesizing advances in BCI technology and clinical applications across neurological conditions. The authors summarize technological classifications, operating principles, and paradigms, and critically appraise recent clinical studies, randomized controlled trials, and translational efforts. Evidence sources include peer-reviewed studies on signal acquisition modalities (EEG, fMRI, fNIRS, MEG, ECoG, SEEG), decoding algorithms (SVM, LDA, ANN, deep learning), feedback/control systems (VR, robotics, haptics), and clinical outcomes reported in supplemental tables for movement disorders (Table S1), cognitive/mental disorders (Table S2), and sensory disorders (Table S3). No primary data collection or meta-analytic statistical synthesis is reported; instead, the review integrates findings to identify trends, challenges, and future directions.

• PD: DBS significantly improves tremor, rigidity, bradykinesia, and balance; in advanced PD, DBS vs best medical therapy increased 'on' time and improved quality of life; in early PD, DBS plus medical therapy improved PDQ-39 by +7.8 points vs −0.2 in controls, reduced levodopa equivalent dose by 39% vs a 21% increase in controls, and reduced motor complications. STN-DBS at 80 Hz improved upper limb assembly performance; higher frequencies reduced rigidity/tremor without affecting bradykinesia. Closed-loop DBS using subthalamic beta activity and dopamine signals shows promise; parameter selection and motor state critically influence adaptive DBS efficacy. • Stroke: In an RCT (n=28 subacute patients), BCI-enhanced motor imagery training outperformed conventional MI, with EEG showing greater alpha/beta desynchronization ipsilaterally. Hybrid EEG/EOG BCIs enabled integrated wheelchair and robotic arm control; in tests with 22 subjects, 5 completed an autonomous beverage task, demonstrating high accuracy and practical utility. BCI neurofeedback targeting theta/alpha bands improved memory encoding and cognitive functions. • SCI: Multimodal BCIs enhanced lower limb movement patterns and alleviated pain; BCI-assisted MI improved upper limb function, particularly early in rehabilitation. Implanted microelectrode arrays enabled precise control of bionic limbs, improving accuracy and flexibility; invasiveness remains a barrier to widespread use. • Locked-in syndrome: Eye-tracking integrated with HMM/DNN improved character recognition and input speed; combining BCI with FES may enhance motor and communication functions. fMRI-guided targeting improved placement for implantable BCIs; standardized methodologies for daily-life implementation are emerging. • Epilepsy: Machine learning on EEG/ECoG supports seizure prediction and localization; bidirectional interfaces with responsive neurostimulation improve seizure control via long-term network modulation. A wireless ECoG prosthesis (ECOGIW-16E) enabled long-term recording and stimulation (up to 6 months) in primates. Memristor-based analog neural signal analysis achieved 93.46% accuracy with ~400× power efficiency over leading CMOS systems. • DoC: EEG/fMRI BCIs detect covert consciousness and differentiate VS from MCS; misdiagnosis rates up to 43% with behavior-only methods can be mitigated. P300 spellers enabled communication in MCS. Neurofeedback improved attention and memory, supporting cognitive recovery. • AD: EEG biomarkers (elevated theta, altered alpha) aid early detection and prognostication; neurofeedback can stabilize or enhance cognition (attention, memory). Innovative conditioning-based BCI models support basic communication and rehabilitation. Emerging concepts include minimally invasive photon-based monitoring. • Depression: DBS trials in subcallosal cingulate showed safety and feasibility but no superiority over sham at 6 months; NAc-DBS reduced depressive-like behaviors in mice via GABA-mediated mechanisms; HB-DBS in 7 patients reduced depressive/anxiety symptoms by ~49% at 1 month, sustained up to 12 months, with oscillatory HB activity correlating with symptom severity. EEG-based BCIs using ResNets improved depression classification; beta-band activity was most discriminative; BCI-enabled psychoneurotherapy reduced high-beta (18–30 Hz) in prefrontal cortex alongside symptom improvements. • Sensory disorders: Auditory BCIs using ERP-based stream segregation achieved up to 95% detection accuracy; auditory P300 BCIs enable communication in ALS; hybrid ASSR+P300 paradigms improved accuracy and stability. Visual BCIs using high-frequency SSVEP with computer vision controlled robotic arms for object manipulation. Speech BCIs: Long-term ECoG-based systems decoded and synthesized intelligible speech in ALS; stable decoding without recalibration for 3 months was demonstrated; optimal decoding observed in the ventral sensorimotor cortex with unilateral surface electrodes. • Anesthesia monitoring: Median nerve stimulation (MNS)-based BCI improved MI vs rest classification accuracy for intraoperative awareness detection, leveraging ERD/ERS modulation. Under propofol sedation, offline classifiers detected movement attempts with accuracies up to 85% and 83%; awake-state trained classifiers generalized above chance to sedated states.

The review underscores BCIs as transformative tools to diagnose, treat, and rehabilitate across diverse neurological conditions by directly interfacing with neural activity. Findings show that BCIs enhance motor recovery (stroke, SCI), enable communication in severe paralysis and DoC, support seizure prediction and closed-loop control in epilepsy, and provide novel assessments and therapies in cognitive and psychiatric disorders (AD, depression). Integration with AI/deep learning improves signal decoding and personalization, while bidirectional interfaces add sensory feedback for more natural control. Convergence with neurosurgical techniques (e.g., DBS, ECoG implants) accelerates translation, though significant challenges remain in long-term biocompatibility, stability, standardized protocols, and validation of clinical efficacy. Addressing ethical, privacy, and safety considerations is vital as BCIs evolve from experimental systems to clinical tools.

BCI technology holds substantial promise to revolutionize diagnosis, therapy, and rehabilitation in neurology. Realizing this potential requires advances in biocompatible, long-term stable implantable electrodes; miniaturized, portable systems; high-accuracy, low-latency decoding; intuitive user interfaces; and rigorous clinical validation. Priority areas include bidirectional and high-performance BCIs, closed-loop stimulation systems, and robust ethical and regulatory frameworks. International collaboration and standardization will facilitate data sharing, interoperability, and faster clinical translation. With careful, responsible development and strong interdisciplinary collaboration led in part by neurosurgery, BCIs can significantly improve quality of life for patients with neurological disorders and drive the broader advancement of neurotechnology.

As a narrative review, the article does not conduct a systematic meta-analysis or provide pooled quantitative estimates. Many applications remain experimental, with gaps in long-term clinical efficacy, generalizability, and standardized protocols. Technical limitations include signal acquisition noise (non-invasive methods), invasiveness and biocompatibility concerns (implants), decoding accuracy, and device usability. Translational barriers persist between laboratory prototypes and routine clinical practice. Ethical, privacy, and security issues around neural data require robust governance.