Proposal

The goal of graduate surgical education is to prepare residents to become competent, independent surgeons who can deliver safe and effective patient care. But how do we know when residents are competent? Traditional assessments of surgical skill rely on subjective evaluations by attendings which are prone to implicit bias—unconscious attitudes that contradict stated beliefs and influence actions. The most studied implicit bias in surgical education is gender bias. Qualitative evaluations of surgical residents’ performance may be implicitly gender biased, and even the perception of gender bias and stereotypes may negatively influence female surgical trainees’ skill performance and career engagement. Gender disparities may also be present in faculty trust and surgical autonomy. 

Our central goal is to develop an AI-aided video-based assessment (VBA) tool that automates accurate and objective assessment of surgical skill in recorded mastoidectomy videos compared with human expert subjective evaluations (Figure 1). We will leverage our AI models for surgical instrument tracking to achieve this goal. We will pursue objectives of (1) developing an AI-aided VBA tool for objectively assessing mastoidectomy surgical skills, and (2) prospectively validating the accuracy of this tool using ground truth labels from averaged blinded attending ratings. Our long-term goal is to develop objective measures of surgical competency.