A Novel Mental Workload-driven Adaptive Training Framework in Robotic Surgical Skill Acquisition

This study proposes and evaluates a mental workload (MWL)–driven adaptive training paradigm for robotic-assisted surgery (RAS). RAS enhances dexterity and sensorimotor control compared with conventional laparoscopy, but limited access to structured training may contribute to errors and safety risks. We developed a multimodal MWL model using brain activity and eye-movement features and integrated it into an adaptive training framework that adjusts task difficulty in real time. In a between-subject experiment, participants received either MWL-driven adaptive training (MATF, n = 10) or self-directed training (n = 10). Mixed-effects models showed that the MWL model classified workload with 83.8% accuracy and that MATF reduced task completion time and errors relative to self-directed training. Gaze behavior and neurophysiological measures were consistent with lower perceived MWL under MATF. These results support workload-aware, closed-loop adaptation as a scalable approach for personalized RAS skill acquisition.