1000 Section - Surface EMG-Based Evaluation of Muscle Activation Patterns in Stroke Patients Under Four Robot-Assisted Training Modes

 To quantify the effects of four RAT training modes on upper-limb muscle activation using three derived sEMG metrics: ΔGap% (change in paretic–nonparetic iEMG activation gap), role conformity (prime mover / synergist / antagonist), and side-specific activation expectations.

Design:

Thirty-eight post-stroke patients were randomly assigned to four RAT modes and completed 20 training sessions. sEMG signals were recorded from 16 upper-limb muscles during shoulder and elbow tasks. ΔGap% and role conformity were computed based on expected activation patterns: prime movers increasing during active tasks, antagonists and non-paretic muscles decreasing, and all muscles decreasing during passive tasks. Group differences were analyzed using the Kruskal–Wallis test.

Results:

ΔGap% varied widely among participants. Group means showed no statistically significant differences (p ≈ 0.62). Overall role conformity also showed no significant group differences (prime movers p≈0.44, synergists p≈0.85, antagonists p≈0.70). For shoulder tasks, all role categories remained nonsignificant (p >0.33). For elbow tasks, prime-mover conformity approached significance (p≈0.054). During passive tasks, most muscles demonstrated expected decreases, with no significant group differences (p≈0.95).

Conclusion:

Across 38 patients, the four RAT modes did not produce statistically significant differences in ΔGap% or role conformity. However, physiological trends were observed: passive training most effectively reduced abnormal activation; active training showed lower prime-mover conformity; and mirror training showed relatively improved activation alignment. sEMG-derived activation metrics may support individualized RAT mode selection for optimizing recovery.