Towards determining clinical factors influencing critical structure identification using Artificial Intelligence

Studys into factors influencing critical view of safety achievement depends on large volumes of video data and granular anatomical annotations, which are limited by the burden of inefficient manual work. Artificial intelligence (AI) has the potential to radically scale the size of clinical studies by automating operative video analysis.

481 videos of laparoscopic cholecystectomy were recorded at Christchurch Hospital, New Zealand over three years. AI algorithms analysed the videos, marking time points where the cystic duct and cystic artery were visible and operative phases. Metrics were stratified by surgeon experience (trainee or consultant) and case complexity (North Shore Grading scale). Nine timing metrics were derived based on the outputs of the AI algorithms and compared against the clinical variables.

Operative time increased with increasing operative difficulty. Significantly consultants demonstrated a higher proportional duration of anatomy visualisation than trainees in complex patients The cystic duct was commonly identified prior to the cystic artery independent of complexity grade.

Surgical video review offers the potential of significant new insights with substantive benefits to patients but is often limited by the costly effort of manual analysis. This paper correlates AI-derived analytics with clinical factors demonstrating real-world utility of AI video analysis.