Multimodal Approach to Assess a Virtual Reality-based Surgical Training Platform.

Area of Science:

• Medical Education
• Human-Computer Interaction
• Physiological Monitoring

Background:

• Virtual reality (VR) offers significant advantages for learning.
• Integrating physiological sensors with VR enhances skill assessment capabilities.

Purpose of the Study:

• To investigate physiological (ECG) and behavioral differences in trainees during VR surgical training.
• To correlate performance metrics with workload and physiological responses.
• To classify participants as gamers or non-gamers based on VR interaction data.

Main Methods:

• Trainees underwent virtual reality-based surgical training.
• Physiological data (ECG) and performance scores (VR-Score) were collected.
• NASA-TLX was used for workload assessment.
• Support Vector Machine (SVM) and Logistic Regression were employed for gamer classification.

Main Results:

• A significant negative correlation was found between VR-Score and NASA-TLX workload (R²=0.15, P<0.03).
• Time-domain ECG metrics, RMSSD (R²=0.16, P<0.05) and pNN50 (R²=0.15, P<0.05), positively correlated with higher VR-Scores.
• Both SVM and Logistic Regression achieved 83% accuracy in classifying gamers and non-gamers, with 88% precision and 83% recall/F1-score.

Conclusions:

• VR surgical training performance is linked to reduced perceived workload and specific physiological responses.
• Objective classification of user types (gamers vs. non-gamers) is feasible using VR interaction data.
• Characterizing physiological and behavioral profiles in VR is crucial for developing advanced training and assessment tools.