Robust Surgical Tool Detection in Laparoscopic Surgery using YOLOv8 Model

Abstract

Surgica1 tool detection involves identifying the position and type of instruments in an image. This is one of the significant issues in automatic video analysis that can aid in evaluating the surgical skills of doctors or automating the process of controlling the viewing angle of the endoscopic camera. This paper presents a robust method for detecting surgical tools using the YOLOv8 model. We trained four different versions of YOLOv8, evaluated their effectiveness, and compared them with previous models. The experimental results indicate that the YOLOv8 models have an average mAP50 greater than 95.6% across all classes, and are significantly better than some previous research findings.