Machine learning distinguishes between skilled and less-skilled psychological performance in virtual neurosurgical performance

Background: Virtual reality surgical simulators have facilitated surgical education by providing safe training environments. Electroencephalography (EEG) has been used to assess neural activity during surgical performance. Machine learning (ML) has been applied to analyze EEG data split into frequency bands. We hypothesized that ML could distinguish between skilled and less-skilled surgeons based on EEG recordings. Methods: Twenty-one participants performed 3 simulated brain tumour resection procedures on the NeuroVR platform (CAE Healthcare, Montreal, Que.) while EEG data were recorded. Participants were divided into less-skilled (medical students and postgraduate year [PGY] 1-3) and skilled (PGY4+) groups. Thirteen EEG frequency band metrics (e.g., alpha, beta, theta:beta ratio) were generated and statistically compared across expertise groups. ML models were trained using these metrics to differentiate between skilled and less-skilled groups. Metrics were arranged in order of importance using model interpretability methods. Results: An artificial neural network achieved a testing accuracy of 100% (area under the receiver operating characteristic = 1.0). Model interpretation identified low alpha (8-10 Hz), which is associated with neural efficiency, as the most important metric for classifying expertise, with skilled surgeons exhibiting higher low alpha than those who were less skilled (p = 0.044). Furthermore, skilled surgeons displayed significantly lower theta:beta ratios (p = 0.048) and significantly higher beta (13-30 Hz, p = 0.049), beta1 (15-18 Hz, p = 0.014), and beta2 (19-22 Hz, p = 0.015), which are associated with technical expertise, attention, memory formation, and anxiety, respectively. Conclusion: This novel methodology elucidates the psychological profile of expert surgeons and the relative contribution of each EEG band to expertise, allowing the development of a neurofeedback protocol for surgical training.