AI assessment of surgical technical skill adaptation across depth levels in simulated tumor resection: a case series study.

OBJECTIVE: Surgical procedures involving varying tissue depths present challenges to surgeons regarding accessibility and precision, restricting instrument movement and increasing the risk of tissue injury. Understanding how experts navigate varying depths is essential, yet research on this issue is limited. Artificial intelligence (AI)-powered systems enable real-time analysis of 3D psychomotor performance during virtual reality simulation tasks. In this study, the authors evaluated performance in a complex brain tumor resection simulation, testing two hypotheses: 1) neurosurgeons' performance scores would remain at an expert level across varying depths, and 2) trainees' scores would decline as they navigated into deeper and more challenging areas. METHODS: Participants included neurosurgeons (n = 14), senior trainees (n = 14), junior trainees (n = 10), and medical students (n = 12). Five left-handed participants were excluded to avoid confounding due to hand dominance, resulting in a final analyzed sample of 45 participants. The Intelligent Continuous Expertise Monitoring System, an AI-powered real-time performance assessment system, assessed surgical performance and measured metrics such as instrument tip separation distance, bleeding risk, healthy tissue injury risk, aspirator force applied, bipolar cautery force applied, and an overall composite score. An average score for each metric at each depth interval (0-15 mm) was calculated across expertise levels for statistical comparison in a retrospective single-center analysis. RESULTS: Neurosurgeons maintained their performance score across varying depths, demonstrating their expertise. Senior trainees had lower scores with increased depth. Surprisingly, increased depth resulted in higher composite scores among medical students and junior trainees, as they had to adapt better instrument techniques in deeper surgical sites. However, their scores remained in the novice spectrum. There was an increasing trend in bleeding risk with greater depth regardless of the expertise level, indicating the more challenging nature of deeper sites. CONCLUSIONS: The unique responses observed at varying depths at each expertise level indicate the necessity for adaptive training modules that accommodate trainee skill set levels and individual learning curves, ensuring development of the competencies required for mastering challenging tasks.