Surgical Instrument Detection and Tracking in Orthopaedic Surgery Using Computer Vision.

Abstract

The purpose of this study is to establish a methodology to digitalise instrument and object tracking analysis in orthopaedic surgery and generate insights into surgical procedures using spatiotemporal data, without introducing costly equipment into the operating theatre. This approach aims to produce descriptive statistics that demonstrate potential improvements in surgical safety and efficiency, presenting opportunities for further research and application in orthopaedic surgery.MethodsThe proposed method utilises a computer vision-based approach by applying instrument tracking and object detection software on a video of a surgical procedure. The model was trained on 23,398 images representing 14 surgical instruments. The model achieved a mean Average Precision of 95.9% on the validation and 96.2% on the test set. A feasibility test was conducted using a surgical video (275 frames, 11 seconds) to detect and track surgical tools. Movement metrics such as distance and spatial distribution were computed to quantify instrument motion patterns.ResultsAll instruments were successfully detected and tracked. Retractors showed lowest (mean total movement <50 px), while Metzenbaum scissors demonstrated the highest activity (>150 px). Violin plots revealed variation in vertical movement among instruments. The system generated detailed spatiotemporal data suitable for surgical workflow analysis and training support.ConclusionThis method produces statistical outputs that may enhance surgical safety by improving procedural organisation and resource allocation. The proposed computer vision framework represents a cost-effective advancement in surgical digitisation, enabling accurate instrument tracking using standard video data and supporting future optimisation of orthopaedic surgical practice with significant implications for clinical research.