Laiwang Zheng, Xu Niu, Hongfeng Zhang, Junpei Liu, Junjie Li

Tianjin College, University of Science and Technology Beijing

Abstract：

This study proposes an intelligent detection method for laparoscopic surgical instruments based on the fusion of the YOLOv8 object detection algorithm and the Deep SORT multi-object tracking algorithm. Using the m2cai16-tool dataset, the model was trained following data preprocessing. Experimental results show that the detection accuracy of the model reached 96.2%, while Deep SORT achieved a tracking accuracy of 80% and a successful tracking rate of 79%. In addition, the model’s counting accuracy reached 90%, and the detection accuracy of retained-instrument risk was 85%. This research demonstrates broad prospects for clinical application.

Key Words：

 object detection; multi-object tracking; data preprocessing; clinical practice