Mount Sinai researchers have developed an artificial intelligence-driven model to train surgical residents in performing complex procedures without the need for a human instructor. In a recent study, all 17 trainees using this system successfully completed a simulated partial nephrectomy, a surgery to remove part of a kidney affected by cancer.
The findings, published in the Journal of Medical Extended Reality, suggest that AI combined with extended-reality technology could help standardize and improve the training process for surgeons. The model uses deep learning algorithms linked to an extended-reality headset, allowing trainees to receive real-time instructions and feedback while practicing surgical steps on synthetic models.
“For the first time, we created an AI model linked to an extended-reality headset to prove that a critical step in a kidney cancer procedure could be done with 99.9 percent accuracy,” said Nelson Stone, MD, Clinical Professor of Urology, Radiation Oncology, and Oncological Sciences at the Icahn School of Medicine at Mount Sinai and corresponding author of the study. “We believe our study offers early proof that AI programs that substitute for proctors, who teach resident physicians, can reduce training costs and ultimately improve the quality, efficiency, and standardization of that instruction.”
Traditionally, surgical residents are taught by instructors present in operating rooms. This approach can lead to inconsistent skill levels among trainees due to varying teaching methods and limited availability of supervisors. The Mount Sinai team’s new system replaces live instruction with an AI program called ESIST (educational system for instructionless surgical training). The system streams visual instructions through headsets while monitoring each trainee’s actions via cameras. Feedback is given instantly if errors are detected during practice on specially designed 3D-printed kidney models.
“Above all, our study proved that a complex procedure like a partial nephrectomy could be effectively taught to surgical trainees using a simulated model, without the presence of an instructor,” Dr. Stone noted. “This finding addresses an urgent need resulting from the shortage of trainers and supervisors to educate physicians on new medical devices and techniques, and from the severe time constraints on attending physicians to train residents pursing surgical careers.”
Dr. Stone also emphasized another benefit: reducing risks associated with inexperienced surgeons performing new procedures directly on patients by allowing more practice outside clinical settings. “From the patient’s point of view, we hope this study will provide reassurance that the technology can be leveraged to greatly improve surgical proficiency while reducing surgical errors,” he said.
The research team plans further development by expanding their AI-based training platform to cover complete surgeries rather than individual steps as tested in this initial trial. According to survey results after training sessions concluded, all participants found significant educational value in this approach.
“Our investigation suggests that AI systems could indeed play an important complementary role in shaping the future of surgical education in this country,” asserts Dr. Stone. “The public should be reassured that the pathway to autonomous learning we investigated in this small study could eventually lead to significant cost savings and improved patient outcomes and importantly—to the cultivation of a highly skilled new generation of surgeons.”
Authors listed on the publication include Jonathan J. Stone, Nelson N. Stone, Steven H. Griffith, Kyle Zeller (the only author not holding equity in Viomerse), and Michael P. Wilson.
Funding for this research came from grants provided by both federal agencies: the National Institute of Biomedical Imaging and Bioengineering (grant 1R41EB026358-01A1) and the National Science Foundation (grant 1913911).
