Samsung Medical Center Develops AI Technology to Guide Excision Paths in Breast Cancer Surgery

Real-Time Analysis of Surgical Video Guides the "Safe Excision Margin"

The research team at Samsung Medical Center has developed a technology that allows artificial intelligence (AI) to provide real-time guidance for safe excision paths during robotic breast cancer surgery.

Professor Jinsoo Yoo, Department of Transplant Surgery at Samsung Medical Center (left in the photo), is consulting with medical staff on the surgical process for liver transplant patients using the AI-based 3D surgical planning platform Liveraiz. Samsung Medical Center

On March 5, Samsung Medical Center announced that a research team led by Professors Yoo Jaemin and Park Woongki from the Department of Breast Surgery, along with Professors Yoo Jinsoo and Oh Namki from the Department of Transplantation Surgery, has developed an AI navigation system that analyzes surgical videos during robot-assisted nipple-sparing mastectomy to guide surgeons toward a safe excision margin. The team also confirmed the system's clinical applicability through external multicenter validation. The results were recently published in the European Journal of Surgical Oncology, the academic journal of the European Society of Surgical Oncology.

Robot-assisted nipple-sparing mastectomy is a procedure where a small incision is made near the armpit, and a robotic arm is used to remove only the breast tissue while preserving the nipple and skin. This method leaves virtually no scar and offers high cosmetic satisfaction, but it has the limitation of making it difficult to distinguish tissue boundaries due to the lack of tactile feedback.

In particular, it is challenging to accurately differentiate the boundary between the subcutaneous fat layer and the glandular tissue. If the excision is too shallow, some breast tissue may be left behind; if it is too deep, there is a risk of damaging the skin's blood supply.

The research team developed a system in which AI analyzes intraoperative video to display the boundary between the fat layer and the glandular tissue—that is, the safe excision margin—on the monitor. Similar to a car navigation system, this provides the surgeon with a visual guide to the excision path.

For AI training, 1,996 frames extracted from surgical videos of 29 robot-assisted mastectomies performed at Samsung Medical Center were used. Breast surgery specialists marked the safe excision margin in each frame, and these annotations were used to train the AI model.

In internal validation, the model achieved a Dice Similarity Coefficient (DSC) of 74.0%. In external validation using eight surgical videos from Samsung Changwon Hospital, the model showed similar performance at 70.8%, demonstrating its potential to operate reliably in surgical environments at other institutions.

The research team had previously developed an AI navigation system for laparoscopic living donor liver transplantation and published their findings in the international journal Scientific Reports. In that study, they analyzed videos from 48 surgeries performed at three hospitals and introduced a technology that displayed the structure of blood vessels around the liver and the safe dissection plane in real time.

Professor Yoo Jaemin stated, "This study developed an AI navigation system for robotic mastectomy and carried out multicenter external validation. It has the potential to enhance precision and safety by guiding the safe excision margin during surgery."

This content was produced with the assistance of AI translation services.