A Korean research team has developed an AI-powered tool that helps surgeons navigate robotic breast cancer procedures with expert-level precision.
The study was led by Professor Lee Jee-a of surgery at Uijeongbu Eulji Medical Center, under the supervision of Professor Park Hyung-seok of Yonsei University College of Medicine. Developed in collaboration with Yonsei, the model identifies skin flap dissection planes in real time during robot-assisted mastectomies.
The system is the first reported use of deep learning to guide intraoperative decisions in robotic breast surgery, according to findings published in [**Breast Cancer Research**](https://breast-cancer-research.biomedcentral.com/articles/10.1186/s13058-025-01981-3).
From left: Park Hyung-seok of the Department of Surgery, Yonsei University College of Medicine; Kim Nam-kug of the Department of Convergence Medicine, University of Ulsan College of Medicine, Asan Medical Center; Lee Jee-a of the Department of Surgery, Uijeongbu Eulji Medical Center; and Ham Sung-won of the Healthcare Readiness Institute for Unified Korea, Korea University Ansan Hospital, developed a surgical assistance system applying AI deep learning technology to robotic surgery. (Courtesy of Severance Hospital)
"Because of the lack of haptic function of the robotic surgical system, a surgical guide for the dissection of the skin flap could improve the postoperative complications and local recurrence rates of breast cancer," researchers wrote in the study.
Robot-assisted mastectomy, performed with robotic arms as small as 8 millimeters, allows for more precise incisions and better cosmetic outcomes compared to open surgery. But the learning curve is steep, and younger surgeons often struggle to locate safe dissection boundaries with limited tactile feedback and field of view, according to a Thursday release from Severance Hospital.
To build the AI model, Lee and her team extracted 8,834 video frames from 10 robotic mastectomies performed between 2016 and 2020 using Intuitive Surgical’s da Vinci Xi system. The deep learning system was trained to detect optimal dissection planes based on manually annotated data from expert surgeons.
In tests, the AI matched surgeon-drawn dissection lines with a Dice coefficient of up to 0.828. The median Hausdorff distance—a metric of spatial accuracy—was under 10 millimeters. No intraoperative complications or cancer recurrences were reported in the studied cases.
The researchers said they plan to expand the model with additional video data to improve its generalizability.
“Because breast surgery often impacts a woman’s self-esteem, demand for minimally invasive robotic procedures will continue to grow,” Park said in Thursday's release. “We developed this technology to help young surgeons train more effectively.”