First surgery performed by a robot without human assistance

For the first time, a robot has successfully performed surgery on a simulated patient without any human assistance. Trained with videos of surgical procedures, the SRT-H (Surgical Robot Transformer-Hierarchy) robot was able to remove a gallbladder by adapting in real time to the anatomical characteristics of the simulated patient, making on-the-fly decisions and self-correcting when things didn't go as planned. The result was published in Science Robotics by a research team led by Axel Krieger of Johns Hopkins University (USA). The same team had previously developed another surgical robot called STAR (Smart Tissue Autonomous Robot), which in 2022 performed the first autonomous robotic surgery on a living animal. However, that procedure was laparoscopic, and the robot followed a rigid, predetermined surgical plan, moving through tissues that had been specially marked to facilitate its work.

A pre-established path

"It was like teaching a robot to drive along a carefully mapped path," Krieger explains. With the new system, however, "it's like teaching a robot to navigate any road, in any conditions, responding intelligently to whatever it encounters. This advancement takes us from robots capable of performing specific surgical tasks to robots that truly understand surgical procedures. This is a fundamental distinction that brings us significantly closer to clinically viable autonomous surgical systems, capable of operating in the chaotic and unpredictable reality of actual patient care."

An interactive robot

Built with the same machine learning architecture as ChatGpt, Srt-h is also interactive: during the procedure, it responds to voice commands (such as "grab the head of the gallbladder") and follows prompts (such as "move your arm slightly to the left"), just like a novice surgeon working with a mentor. The gallbladder removal procedure performed in the study involves a sequence of 17 tasks lasting several minutes. The robot had to identify specific ducts and arteries, grasp them precisely, strategically place clips, and sever the parts with scissors. Srt-h learned to perform each action by watching videos of Johns Hopkins surgeons operating on pig cadavers. The visual training was further reinforced with captions describing the tasks. After watching the videos, the robot performed the surgery with 100% accuracy: during the various tasks, it acted imperturbably and with the expertise of an experienced human surgeon, even during unexpected scenarios typical of real medical emergencies.

Although the robot took longer than a human surgeon to perform the procedure, the results were comparable to those of an experienced surgeon. "Our work," observes the study's first author, Ji Woong "Brian" Kim , "demonstrates that AI models can be made reliable enough for autonomous surgery, a goal that once seemed distant but is now clearly achievable."