Presentation
Modeling Surgeons' Mental Workload in Robot-Assisted Surgery: A Bidirectional Overload-Recovery Model Using Neurophysiological Measures
SessionHC6: Clinician Workloads
DescriptionRobotic-assisted surgery (RAS) improves surgical dexterity and visualization but can impose high mental workload (MWL) due to complex interfaces, demanding task coordination, and reduced team interaction. This study presents a real-time MWL assessment framework that models MWL dynamics as bidirectional transitions between overload and recovery. Physiological data were collected from two phases: cognitively demanding gaming tasks used as the source domain, and three RAS training tasks, including suturing, anastomosis, and chicken wing dissection. EEG signals and eye-tracking measures were segmented into 10-second sliding windows with 5-second steps. Instead of treating MWL as a static classification problem, the proposed Dual-Cox framework uses two time-varying Cox proportional hazards models to separately estimate the instantaneous risk of entering a high-MWL state and returning to a lower-MWL state. This structure captures the asymmetric nature of overload accumulation and recovery. Compared with logistic regression, SVM, and random forest classifiers, the Dual-Cox model achieved the highest accuracy across all three RAS tasks and improved precision in identifying MWL states. These findings suggest that modeling MWL as concurrent overload-recovery transitions can provide more robust real-time monitoring for surgical training and adaptive assistance.
Event Type
Lecture
TimeThursday, October 22nd3:40pm - 4pm PDT
Location