LSTM Cognitive Load Assessment for VR Surgical Training Based on E4 Wristband Signals

The integration of Virtual Reality (VR) into surgical education has revolutionized thepedagogical landscape, offering a risk-free environment for residents to acquire complexpsychomotor skills. However, the efficacy of VR training is heavily contingent upon themanagement of cognitive load. Excessive cognitive load can lead to skills decay and poorretention, while insufficient load may result in disengagement. Traditional methods ofassessing cognitive load, such as self-report questionnaires, disrupt the training workflowand lack temporal resolution. This paper proposes a novel framework for continuous,objective cognitive load assessment using Long Short-Term Memory (LSTM) networksapplied to physiological signals captured by the Empatica E4 wristband. We collectedElectrodermal Activity (EDA), Blood Volume Pulse (BVP), and Skin Temperature (ST) datafrom forty medical students performing simulated laparoscopic procedures. By leveragingthe temporal memory capabilities of LSTM architectures, we modeled the sequentialdependencies in physiological responses to varying levels of surgical difficulty. Theproposed model demonstrates superior classification accuracy compared to traditionalmachine learning baselines, establishing a robust foundation for adaptive VR surgicaltraining systems. This research bridges the gap between affective computing and medical education, providing a pathway toward real-time cognitive state monitoring.