An adaptive hand exoskeleton rehabilitation training system integrating virtual reality and an AI-based assessment...

Post-stroke hand motor impairment is a major cause of long-term functional disability and reduced quality of life, with approximately 70% of stroke survivors experiencing persistent limitations in fine motor control. Conventional rehabilitation is constrained by low adherence, subjective assessment, and insufficient individualization, which limits exploitation of the neuroplasticity window for motor relearning. To address these challenges, we propose a bio-AI-VR integrated hand rehabilitation system that fuses biosignal sensing (bio), AI-based analysis, and virtual reality (VR) interaction to realize an efficient, adaptive, and quantifiable closed-loop training process. The integration rationale is grounded in three theoretical pillars: (i) multimodal data fusion theory-combining heterogeneous biosignal and behavioral data through AI to overcome single-modality limitations; (ii) closed-loop adaptive control theory-dynamically balancing challenge and capability via real-time feedback; (iii) neuroplasticity multisensory enhancement theory-coordinating visual, proprioceptive, and motor pathways to strengthen cortical reorganization. This work addresses three testable hypotheses: (RQ1) Can multimodal biosignal fusion achieve real-time assessment with $R^2\ge 0.65$ and latency $<50$ ms? (RQ2) Does bio-AI-VR integration yield FMA-UE improvement $\ge 6$ points (minimal clinically important difference) with effect size $d\ge 0.8$ ? (RQ3) Are all three components (bio, AI, VR) necessary, with ablation causing $\ge 15\%$ performance degradation?

A lightweight hand exoskeleton ( $<400$ g, 3 DoF/finger) integrates a 6-axis IMU (100 Hz) and 16-channel sEMG (1 kHz) to synchronously acquire kinematics and muscle activation. Extended Kalman filtering fuses sensor streams before AI processing. Features include range of motion (ROM), smoothness metrics (SPARC, LDLJ), sEMG root-mean-square (RMS), median frequency (MDF), and co-contraction index (CCI). A hybrid model combining random forests (200 trees, depth 8) and support vector regression (RBF kernel, $\gamma =0.01$ , $C=10$ ) outputs a real-time composite score $S_t\in [0,1]$ via multi-task learning with GroupKFold cross-validation, mapped to clinical scales through Sigmoid normalization. FMA-UE proxy labels for window-level training were constructed via linear interpolation (80%), biomechanical anchoring (15%), and expert annotation (5%, inter-rater $\kappa =0.78$ ). A cloud AI engine communicates bidirectionally with Unity-based VR over MQTT to close the perception-assessment-assistance loop. The assistance-as-needed (AAN) algorithm adjusts exoskeleton torque ( $u_t$ ) and VR difficulty ( $d_t$ ) using $S_t$ as control input with hysteresis, dead zone, and rate limiting to ensure smooth adaptation. Twenty-four stroke survivors (3-12 months post-stroke, FMA-UE 15-50) underwent 4-week training (5 sessions/week, 20 min/session). Outcomes included FMA-UE (primary), ARAT, grip strength, normalized ROM, task success rate, and System Usability Scale (SUS). Statistical analysis employed paired $t$ -tests with Hedges' correction for effect sizes, Bonferroni adjustment for multiple comparisons, and leave-one-subject-out cross-validation (LOSOCV) to assess model generalization.

All 24 participants completed the study with one missed session (479 of 480 scheduled sessions, 8,946 annotated segments); end-to-end latency median 38 ms (IQR 33-42 ms), decomposed as: sampling $8\pm 2$ ms, preprocessing $4\pm 1$ ms, network $12\pm 3$ ms, AI inference $5\pm 1$ ms, control $2\pm 0.5$ ms, command return $7\pm 2$ ms. Offline model performance: $R^2=0.72$ (GroupKFold), MAE $=$ 3.2 points, Spearman $\rho =0.68$ with FMA-UE ( $p<0.001$ ); LOSOCV: $R^2=0.68\pm 0.09$ ; test-retest ICC(2,1) $=$ 0.84 [0.76, 0.91]. AAN algorithm reduced assist torque $62\%\to 45\%$ ( $-27.4\%$ ), increased VR difficulty $0.42\to 0.69$ ( $+64.3\%$ ), improved task success $61.3\%\to 82.1\%$ ( $+20.8$ pp). Clinical outcomes (paired $t$ -test): FMA-UE $+9.1$ [6.7, 11.5], $d=0.98$ ; ARAT $+7.6$ [5.2, 10.0], $d=0.93$ ; grip $+4.1$ kg [2.5, 5.7], $d=0.72$ ; ${\mathrm{ROM}}_n$ $+0.14$ ; SPARC $-0.16$ . Subgroup analysis: moderate-to-severe ( $n=13$ ) showed greater FMA-UE gain ( $+10.7$ vs $+7.2$ in mild-to-moderate, $p<0.05$ ). Ablation experiments confirmed synergistic necessity: Bio only ( $R^2=0.45$ , FMA-UE $+4.3$ ), VR only ( $R^2=0.38$ , $+3.9$ ), Bio-AI ( $R^2=0.70$ , $+7.2$ , compliance 68%), complete system ( $R^2=0.72$ , $+9.1$ , compliance 88%). SUS $84\pm 6$ ; no serious adverse events.

Results validate all three hypotheses: (i) multimodal fusion exceeded technical targets ( $R^2=0.72>0.65$ , latency 38 ms $<50$ ms); (ii) clinical efficacy surpassed MCID with large effect sizes (FMA-UE $+9.1>6$ points, $d=0.98>0.8$ ), exceeding published spontaneous recovery rates (2-4 points); (iii) ablation experiments demonstrated $\ge 15\%$ degradation when removing any component, confirming non-additive synergistic effects of bio-AI-VR integration. Compared to recent brain-computer interface systems using EEG-based motor imagery, this approach achieves paradigm shift toward execution-based rehabilitation with direct motor intent capture and real-time physical feedback. The AAN control law elevates from low-level motion control to high-level rehabilitation strategy, spanning multiple temporal scales (window to course-level) and dual channels (physical assistance $+$ cognitive challenge). Limitations include single-arm design limiting causal inference, small sample size ( $n=24$ ), short intervention period (4 weeks), FMA-UE proxy construction via linear interpolation, and controlled clinical setting vs. real-world deployment. Future work requires larger RCTs with active control arms, extended follow-up (3-6 months), dense longitudinal assessments, exploration of deep learning architectures for temporal modeling, and validation in home-based telerehabilitation settings. The bio-AI-VR system demonstrates feasibility of data-driven, multimodal closed-loop rehabilitation, offering a wearable, low-latency, and personalized solution for post-stroke hand recovery that bridges the gap between laboratory innovation and clinical translation.