Experimental surgery performed by AI-driven surgical robot

Intuitive Surgical, an American biotechnology company, introduced Davinci surgical robots in the late 1990s, and they became revolutionary teleoperation equipment. Expert surgeons could operate on patients remotely, handling robotic arms and their surgical tools based on a video flow of Davinci’s integrated cameras and endoscopes.

Now, researchers from the University of John Hopkins have put a Chatgpt type AI in charge of a Davinci robot and have taught him to carry out a gallbladder elimination surgery.

Kuka surgeries

The idea of putting a computer behind the wheel of a surgical robot is not entirely new, but these were mainly based on the use of preprogrammed actions. “The program told the robot exactly how to move and what to do. To improve this, a team led by Axel Krieger, assistant professor of mechanical engineering at John Hopkins University, Build Star: The Smart Tissue Autonomous Robot. In 2022, he succeeded in surgery on a living pig.

But even star could not do it without specially marked fabrics and a predetermined plan. The main star difference was that his AI could make adjustments to this plan according to the flow of cameras.

The new robot can do much more. “Our current work is much more flexible,” says Kim. “It’s an AI that learns from demonstrations.” The new system is called SRT-H (surgical robot transformer) and was developed by Kim and his colleagues, added Krieger.

The first change they brought was to the equipment. Instead of using a personalized robot as a star, the new work was based on the Davinci robot, which has become a de facto standard in teleoperation surgeries, with more than 10,000 units already deployed in hospitals in the world. The second change was the software that stimulated the system. It was based on two models of transformers, the same architecture that feeds Chatgpt. One was a high -level policy module, which was responsible for planning tasks and the procedure guarantee. The low level module was responsible for the execution of the tasks issued by the high -level module, reflecting its instructions into specific trajectories for robotic arms.