Surface EMG-Based Assessment of Stroke Survivors: From Impairment Status to Fugl-Meyer Upper Extremity Scores

Background Stroke is a leading cause of long-term disability worldwide, and with rising incidence and a global shortage of rehabilitation professionals, there is a growing need for scalable methods to assess motor impairment. Surface electromyography (sEMG) has emerged as a promising modality for capturing motor function. Despite its potential, sEMG remains underused in clinical practice, and its ability to complement standardized assessments is not yet well established. Methods Bilateral sEMG was recorded from 23 subacute stroke participants as they performed four standardized wrist and hand tasks. Task-level features were extracted from raw sEMG recordings for impairment status prediction and impairment level estimation. All models were evaluated using leave-one-subject-out cross-validation (LOSO-CV). Benchmark regression models used therapist-rated subscores as inputs and paired t-tests compared their absolute errors against those of the sEMG-based models. Model interpretability was examined using SHAP values to identify sEMG features contributing most strongly to predicted impairment levels. Results Among individual FMA wrist and hand tasks, wrist extension yielded the highest performance in distinguishing impairment status (accuracy 0.87 ± 0.16; AUC-ROC 0.94), while combining all four tasks further improved accuracy to 0.90 ± 0.14 and AUC-ROC to 0.96. For impairment level estimation, the sEMG-based predictions reached a RMSE of 3.12 for FMA wrist and hand subscore (FMA-WH) and 6.68 for FMA-UE full score. SHAP analysis with the sEMG-based model revealed that higher extensor activation corresponds to higher predicted scores. Conclusions Consumer-grade sEMG signal collected during FMA-UE hand and wrist tasks enabled estimation of partial and full FMA-UE scores with prediction errors below the minimally clinically important difference (MCID), indicating the potential for low-cost assessment in non-expert settings. However, validation in larger and longitudinal cohorts is needed to establish robustness over time, and future work should assess whether this approach can be reliably deployed without therapist assistance.