CLOSED-LOOP BRAIN–COMPUTER INTERFACES WITH AI-DRIVEN NEUROFEEDBACK FOR MOTOR AND COGNITIVE REHABILITATION: COMPREHENSIVE EXTENDED REVIEW AND MINDAFFECT DATASET ANALYSIS
International Journal of Grid Computing & Applications · 28.03.2026 · 17 (1) · 14-31
DOI: 10.5121/ijgca.2026.17102 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Grid Computing & Applications, online erschienen am 28.03.2026. Verfasst von Sungho Kim und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Closed-loop brain–computer interfaces (BCIs) represent a revolutionary advancement in the fields of motor and cognitive rehabilitation, offering transformative opportunities to enhance recovery outcomes for patients with neurological impairments. Unlike traditional open-loop systems, closed-loop BCIs are capable of continuously monitoring neural activity in real time, allowing for the delivery of adaptive neurofeedback that is dynamically tailored to the user’s current brain state. By leveraging this real-time monitoring and feedback capability, these systems can effectively harness neuroplasticity, reinforcing beneficial neural patterns and facilitating the restoration or enhancement of motor, cognitive, and emotional functions. The integration of artificial intelligence (AI) algorithms, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory networks (LSTMs), and reinforcement learning approaches, enables highly accurate decoding of neural signals and adaptive adjustment of therapeutic interventions based on individual performance. This combination of AI and closed-loop neurofeedback provides a powerful framework for personalized rehabilitation that can adapt to the unique needs, progress, and neural characteristics of each patient. This extended review synthesizes the state-of-the-art in closed-loop BCIs, providing a comprehensive overview of the major components and methodologies that underpin these systems. We begin by examining neural signal modalities, such as electroencephalography (EEG), electrocorticography (ECoG), magnetoencephalography (MEG), and functional near-infrared spectroscopy (fNIRS), highlighting the advantages and limitations of each modality in terms of spatial and temporal resolution, invasiveness, portability, and cost. In addition, we discuss the preprocessing techniques that are essential for highquality neural decoding, including artifact removal, bandpass filtering, independent component analysis (ICA), epoch segmentation, and feature extraction using time-frequency analysis and spatial covariance metrics. The review also explores AI-based decoding algorithms, emphasizing how advanced machine learning and deep learning models can identify and classify complex patterns in neural data, supporting real-time interpretation of motor intentions, cognitive states, or emotional responses. A key aspect of closed-loop BCIs is the design and implementation of feedback mechanisms, which can range from robotic exoskeleton actuation and functional electrical stimulation (FES) to immersive virtual reality (VR) environments. These feedback modalities are critical for engaging the user, reinforcing correct neural activity, and optimizing rehabilitation outcomes. We also highlight personalized rehabilitation strategies, demonstrating how adaptive systems can modulate task difficulty, adjust feedback intensity, and deliver individualized interventions based on longitudinal monitoring of performance and neuroplasticity. The review includes an analysis of the MindAffect BCI dataset, providing a concrete example of practical implementation and evaluating classification performance in a real-world scenario. This dataset serves as a benchmark for assessing decoding accuracy, system responsiveness, and adaptive feedback effectiveness, offering insights into the feasibility of deploying closed-loop BCIs in clinical and home-based settings. Furthermore, we present comparative discussions on open-loop versus closed-loop systems, underscoring how closed-loop designs exploit feedback and learning mechanisms to improve skill acquisition, accelerate recovery, and enhance long-term retention. The potential of multimodal integration is explored, showing how combining EEG with additional physiological or kinematic data can enhance decoding accuracy, reduce susceptibility to noise, and enable richer and more nuanced feedback. The review also addresses emerging trends in home-based rehabilitation, highlighting the importance of remote monitoring, telerehabilitation platforms, and user-friendly interfaces that support high-frequency training in the patient’s natural environment. Finally, we discuss clinical protocols and ethical considerations, including standardization of intervention procedures, longterm validation, patient adherence, brain data privacy, algorithmic transparency, and mitigation of unintended cognitive or emotional effects. This work aims to provide a comprehensive roadmap for future research and clinical deployment, guiding efforts toward scalable, patient-centered, and ethically responsible closed-loop BCI solutions. By synthesizing current knowledge, identifying technical and clinical challenges, and outlining best practices for AI-driven adaptive neurofeedback, this review contributes to the development of robust, effective, and accessible neurorehabilitation systems that can improve patient outcomes and transform the landscape of neurological therapy. The insights provided herein are intended to inform researchers, clinicians, and developers in the design, optimization, and ethical deployment of next-generation BCIs capable of enhancing motor, cognitive, and emotional recovery in diverse patient populations.
Zitieren
APA
Kim, S., & Kim, S. (2026). CLOSED-LOOP BRAIN–COMPUTER INTERFACES WITH AI-DRIVEN NEUROFEEDBACK FOR MOTOR AND COGNITIVE REHABILITATION: COMPREHENSIVE EXTENDED REVIEW AND MINDAFFECT DATASET ANALYSIS. International Journal of Grid Computing & Applications, 17(1), 14-31. https://doi.org/10.5121/ijgca.2026.17102
Vancouver
Kim S, Kim S. CLOSED-LOOP BRAIN–COMPUTER INTERFACES WITH AI-DRIVEN NEUROFEEDBACK FOR MOTOR AND COGNITIVE REHABILITATION: COMPREHENSIVE EXTENDED REVIEW AND MINDAFFECT DATASET ANALYSIS. International Journal of Grid Computing & Applications. 2026;17(1):14-31. doi:10.5121/ijgca.2026.17102
BibTeX
@article{Kim2026,
author = {Kim, Sungho and Kim, Sarah},
title = {CLOSED-LOOP BRAIN–COMPUTER INTERFACES WITH AI-DRIVEN NEUROFEEDBACK FOR MOTOR AND COGNITIVE REHABILITATION: COMPREHENSIVE EXTENDED REVIEW AND MINDAFFECT DATASET ANALYSIS},
journal = {International Journal of Grid Computing & Applications},
year = {2026},
volume = {17},
number = {1},
pages = {14--31},
doi = {10.5121/ijgca.2026.17102}
}
EUVIMED