Explaining Motion Control Precision via sEMG for Physiological Feedback in Robotic Surgery Training

Abstract

Automated performance metrics (APMs) are widely used for objective assessment in robotic surgical training. However, while they provide numerical scores, they often lack interpretable insights into the underlying motor behavior needed to deliver actionable guidance for skill improvement. To bridge this gap, this study investigates the integration of surface electromyography (sEMG) to explain motion control precision and derive interpretable physiological feedback. We collected sEMG signals from thirteen novice subjects performing canonical motion primitives in a VR-based surgical robot simulation. Twenty sEMG features from time, frequency, and wavelet domains were extracted and analyzed via linear regression to link neuromuscular activation with motion control precision. The results identified Wavelet Peak Count as the most discriminative predictor, achieving an $R^{2}$ score of 0.92 in Leave-One-Subject-Out cross-validation. Crucially, muscle contribution analysis revealed that lower precision was associated with significantly higher sEMG signal variability and excessive activation, particularly in the Deltoid complex during vertical motions $(\mathbf{p}<\mathbf{0. 0 0 8 3})$. These findings confirm that specific sEMG features can effectively characterize the neuromuscular control patterns underlying surgical maneuvers. By translating these features, the proposed approach identifies muscle-specific root causes of error (such as excessive tension or proximal instability) thereby facilitating the generation of targeted, semantically meaningful physiological feedback to support surgical technical skill acquisition.