Abstract
INTRODUCTION: Coronary artery bypass grafting (CABG) is a technically demanding procedure where surgical skill directly influences outcomes. Traditional evaluation relies on expert subjective judgement, which is resource-intensive and lacks scalability. The emergence of computer vision and deep learning offers potential for objective, automated skill assessment. Prior research has explored phase recognition and gesture classification in surgery; however, few studies have applied AI-driven evaluation in high-stakes cardiac procedures. Therefore, the objective of this study is to develop and validate an artificial intelligence (AI)-based framework for the automated assessment of surgical technical skills in CABG using real-world surgical videos, benchmarked against expert ratings. METHODS AND ANALYSIS: This study is a prospective, single-centre observational study conducted in a high-volume surgical hospital. Eligible participants are adult patients undergoing elective CABG with complete intraoperative video data. Videos are analysed using a hybrid AI pipeline to generate scores based on visual impression and tool trajectory accuracy. The primary outcome is the feasibility of AI annotation, that is, the intraclass correlation coefficient value of AI predicted score and human rating data. Secondary outcomes include the consistency between AI and expert skill assessments, analysis of surgical instrument trajectories and the correlation of AI-derived skill scores with intraoperative graft flow and resistance. Exploratory outcomes aim to correlate surgical skill with graft patency at 1 year and major adverse cardiovascular events within 6 months and 12 months postoperatively. ETHICS AND DISSEMINATION: The Ethics Committee in Fuwai hospital approved this study (2024-2563). The results of the study will be submitted for publication in a peer-reviewed journal. TRIAL REGISTRATION NUMBER: NCT06739005.