450 Estimating Cognitive Load Via Pupil Diameter: A Novel Metric for Assessing Neurosurgical Residency Applicants

INTRODUCTION: Neurosurgical residency applicants are traditionally evaluated on metrics that do not reflect technical ability. While dexterity tests are sometimes used, they do not reveal the internal processing behind a test outcome. Cognitive load (CL) can provide insight into the bandwidth required to perform tasks and whether residual capacity exists for increasing task burden. The baseline change in pupil diameter (BCPD) can be used to calculate CL, providing insights into mental effort and decision-making processes. METHODS: We compared CL between six preclinical and six clinical medical students and four residents performing surgical knot-tying and finger dexterity tests while wearing point-of-view and eye-tracking cameras. We defined BCPD by computing a reference average pupil diameter established during the first task. We calculated task CL by measuring the average difference between baseline and current pupil diameter. RESULTS: Preclinical students had higher average BCPD values (−0.07) than clinical students (−0.56). Resident BCPD averaged −0.16. Increasingly negative values suggest lower CLs. One resident’s BCPD was 1.28 (test score >93rd percentile); when removed, the average resident BCPD was -1.9. Low CL was weakly correlated (0.029) with increased task completion time. Tool motion was not significantly different between groups. Residents had more stable gaze entropy (0.33) than preclinical (0.46) and clinical students (0.58). CONCLUSIONS: With increasing clinical experience, CL during simulated dexterity tasks declined significantly but did not reliably predict task completion time. This suggests that working memory may be occupied despite rapid completion time, making it difficult to attend to unexpected events. Incorporating measures of CL alongside other machine learning-enabled metrics may provide a more objective means of evaluating residency applicants while identifying opportunities for personalized education. The next steps include ongoing data acquisition from more students, residents, and attendings.