Back to 2026 Abstracts
Quantifying Longitudinal Surgical Skill Progression in Residents Using AI-based Video Analysis
Sofia Volpi1, Guillaume Kugener1, Zeynep Bilgili1, Mohammed Azad Khan Abdus Salam1, Gabriel Brat1,2 1. Department of Surgery, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA 2. Department of Biomedical Informatics, Harvard Medical School, Boston, MA
Background: Open surgical skills evolve continuously throughout residency, yet objective measurement of longitudinal improvement remains limited. We demonstrate an Artificial Intelligence (AI)-based approach to detect technical progression in surgical interns. Methods: From 2023 to 2025, surgical interns at a single academic center were recorded performing suturing tasks during their first 9 weeks of residency; with experts recorded as reference. Using an AI hand-tracking model, we extracted kinematic metrics (velocity, distribution, and jitter) and trained a logistic regression model to estimate probability of expert. Performance was evaluated using 5-fold cross-validation. Longitudinal improvement was assessed using mixed-effects models examining changes in expert probability by (1) weeks into residency and (2) recording session (practice exposure). Results: A total of 554 videos from 40 interns were recorded, with 83% captured at least twice; 8 surgeons and 5 senior residents served as reference. The logistic model significantly discriminated experts from week-1 interns (Area-Under-Curve 0.74, 95% CI 0.63-0.85; p = 0.005). In mixed-effects models, expert probability increased by 4.6% per recording session (β = 0.046, p < 0.001) and 2.5% per week into residency (β = 0.025, p < 0.001). However, when both predictors were standardized and included simultaneously, only recording session remained associated with improvement (β = 0.064 per 1 SD increase, p = 0.014), whereas calendar week was not (β = 0.005, p = 0.858) Conclusion: AI-derived motion metrics show trainees progressively resemble experts, with early improvement more closely associated with practice exposure than time in residency. This framework enables objective tracking of surgical skill progression.
Back to 2026 Abstracts
