Tracking camera control in endoscopic dacryocystorhinostomy surgery

Objectives Poor camera control during endoscopic dacryocystorhinostomy (EnDCR) surgery can cause inadequate visualisation of the anatomy and suboptimal surgical outcomes. This study investigates the feasibility of using computer vision tracking in EnDCR surgery as a potential formative feedback tool for the quality of endoscope control. Design A prospective cohort analysis was undertaken comparing junior versus senior surgeons performing routine EnDCR surgery. Computer vision tracking was applied to endoscopic video footage of the surgery: Total number of movements, camera path length in pixels and surgical time were determined for each procedure. A Mann–Whitney U-test was used to test for a significant difference between juniors and seniors (P < 0.05). Setting Operating theatre. Participants Ten junior surgeons ( 100 completed procedures). Main outcome measures Total number of movements of the endoscope per procedure. Path length of the endoscope per procedure. Results Twenty videos, 10 from junior surgeons and 10 from senior surgeons were analysed. Feasibility of our tracking system was demonstrated. Mean camera path lengths were significantly different at 119 329px (juniors) versus 43 697px (seniors), P ≪ 0.05. The mean number of movements was significantly different at 9134 (juniors) versus 3690 (seniors), P ≪ 0.05. These quantifiable differences demonstrate construct validity for computer vision endoscope tracking as a measure of surgical experience. Conclusions Computer vision tracking is a potentially useful structured and objective feedback tool to assist trainees in improving endoscope control. It enables juniors to examine how their pattern of endoscope control differs from that of seniors, focusing in particular on sections where they are most divergent.