Identification of surgical forceps using YOLACT++ in different lighted environments.

Abstract

Forceps tracking in laparoscopic surgery contributes to improved surgical outcomes. We identified forceps using YOLACT++ for fast and accurate segmentation. Differences in the illumination of the environment can affect the image recognition accuracy in deep learning. Therefore, we examined the speed and accuracy of YOLACT++ forceps identification in different illuminated environments. We expected that this experiment would help us understand the optimal lighted environments for YOLACT++ and to further improve the performance of the forceps identification model. The greatest accuracy was obtained under a light-shielded environment with light shining only on the suture area. Although a laparotomy with a clear view of the surgical site is easier for the physician to operate in, we concluded that the forceps identification model of YOLACT++ can be used more effectively in the laparoscopic surgical environment.Clinical Relevance- This study contributes to analyzing the cause of surgical errors in laparoscopic surgery.