Design of an adaptive lower limb rehabilitation system with virtual reality and multimodal biosignals

Objectives To develop and evaluate a virtual-reality-enhanced lower-limb rehabilitation framework that combines EEG-based motor-intention detection, robotic gait assistance, trunk-motion assessment, and lumbar surface electromyography to support multimodal assessment, interactive training, and therapist-oriented performance monitoring. Methods Nine healthy participants completed VR-enhanced EEG motor-imagery training, open-loop calibration, closed-loop exoskeleton-triggering, trunk-attitude assessment, and lumbar-sEMG evaluation in separate experimental sessions. A four-degree-of-freedom hip–knee exoskeleton was integrated with participant-specific CDF threshold calibration, treadmill synchronization, VR rehabilitation tasks, attitude sensing, and RMS/MPF analysis. Participant identity was consistent across modalities, although the signals were not acquired simultaneously. Results The seated VR motor-imagery condition produced the highest open-loop command accuracy of 75.35 ± 4.21 %. Mean closed-loop gait-triggering accuracy was 76.7 ± 10.0 %, ranging from 60 % to 90 %. Mean trunk-movement magnitudes were 30.8°, 30.7°, 91.3°, and 31.0° for left lateral flexion, right lateral flexion, forward flexion, and backward extension, respectively. System prompting increased mean RMS from 0.094759 to 0.103330 mV and reduced mean MPF from 111.2023 to 103.0729 Hz. Conclusions The findings support the feasibility of integrating intention-driven exoskeleton control, VR interaction, postural assessment, and muscular-response monitoring within a unified rehabilitation framework for adaptive, multimodal rehabilitation.