Research on Diagnosis and Localization of Moles for Robotic Mole Removal Surgery Based on YOLO

Surgical treatment is often considered the first choice for mole removal. Treatments such as laser mole removal are widely performed, but intraoperative navigation remains a significant challenge, relying heavily on the experience of surgeons. Additionally, when multiple moles are required to be removed, a more repetitive workload can occur in the surgical procedures. To address these issues, this study aims to achieve precise intraoperative navigation and to achieve modes of automatic diagnosis and localization for robotic mole removal surgery. By training the YOLO model on an open-source categorical mole dataset, this study achieves the automatic diagnosis and localization of the moles during surgeries. In the evaluation of the study, a precision of 0.839 and an average precision mAP of 0.862 which are at a high level are observed. This study verifies that the YOLO can play a critical role in enabling the automatic diagnosis and localization of robotic mole removal surgery.