Rosota, a startup developing surgical data-based artificial intelligence robotics, said it has succeeded in using AI trained on surgeons' operative experience to carry out actual surgical motions with a robot.
The company said on the 20th that it had released results from a preclinical animal study in which AI trained on laparoscopic surgery data was applied to an actual surgical robot. Under the approach, a surgeon designates the surgical step and the AI autonomously executes the movements of the robotic instruments required for that step.
The core of the technology lies in how the data was built. Rosota attached sensors it developed in-house to conventional laparoscopic instruments, recording surgical video together with instrument movement and status. That allowed the AI to learn how a surgeon moves the instruments and manipulates tissue in each scene of the video.
The researchers used 849 surgical demonstration datasets obtained from chinchilla rabbits to train the AI. After validation in simulation, actual robotic surgery was evaluated on separate animals that had not been used in training. In the execution tests, the AI performed robotic motions step by step, including approaching the target site, grasping and retracting tissue, and resecting it, as required for an appendectomy. The recovery of the test animals and any complications were also monitored after surgery. Rosota said it had become the first in the world to implement an integrated technology that applies AI trained on general laparoscopic surgery data to an actual robot and links it through to animal surgery and post-operative survival.
The study does not amount to fully autonomous surgery, in which AI independently makes judgments across the entire procedure, because the operations proceeded based on the physician's decisions. Building on the results, Rosota plans to advance development of an "AI surgical copilot" that supports medical staff in surgical decision-making and instrument handling.
"Surgical video alone makes it difficult to fully record how a surgeon actually moved the instruments and approached the tissue," Rosota CEO Seo Ye-chan said. "We are turning the experience accumulated in the operating room into data and making it learnable by AI, developing it into technology that substantively assists medical staff in their judgments and instrument handling."
