A research team from Johns Hopkins University (Johns Hopkins University) has successfully developed an advanced surgical robot that was able to carry out a critical stage of cholecystectomy almost autonomously.
This robot, named SRT-H (Surgical Robot Transformer-Hierarchy), relies on machine learning techniques similar to ChatGPT techniques, which enabled it to provide performance close to the level of expert surgeons even in difficult and changing conditions.
The researchers see this achievement as an important step towards the development of autonomous and reliable surgical systems, capable of assisting or performing operations completely autonomously, which could revolutionize healthcare. The results of the study of the robot test were published on July 9, 2025, in the journal Science Robotics.
From programmed robots to intelligent surgical systems
“This development moves us from the stage of robots that carry out previously defined surgical tasks to robots capable of understanding and comprehending surgical procedures in depth, a fundamental advance that brings us closer to clinically applicable autonomous systems in realistic and unpredictable medical environments,” says medical robotics expert Axel Krieger, supervisor of the research.
The same research team led by Axel Krieger had developed in 2022 a robot called STAR (Smart Tissue Autonomous Robot) carried out the first autonomous surgery on a living animal, but it was in a strictly controlled environment and relied on distinctive marks on tissues and a previously prepared surgical plan so that the robot could accomplish the required tasks accurately.
Axel Krieger likens the difference between the two generations of robots by saying: “If STAR training is like teaching a robot to drive a car on a carefully drawn road, SRT-H training is like teaching him to drive on any road and under any conditions, with the ability to intelligently respond to emergency situations.
How to teach SRT-H robot surgery
The researchers trained the SRT-H robot to watch videos of Johns Hopkins surgeons performing cholecystectomies on pig carcasses, in addition to supporting its training with text comments explaining the details of the surgical steps.
Instant adaptation to anatomical differences between patients.
Make immediate decisions and self-correct his mistakes when an emergency occurs.
Respond to voice commands instantaneously, such as: hold the head of the gallbladder “or” move the left arm slightly towards the left.
Co-researcher Ji Woong (Brian) Kim, the lead author of the study, says: This work represents a big leap compared to previous attempts, because it overcame fundamental obstacles that were preventing the adoption of autonomous surgical robots in practice. We have proven that AI models can be reliable enough to perform surgery autonomously, he said.
From simple tasks to a complex process
Krieger's team had previously trained robots for simple tasks, such as handling needles, lifting tissues, and sewing, which are short tasks that only take a few seconds to complete. But the operation of cholecystectomy is more complicated, as it involves 17 consecutive steps, including grasping fine arteries, installing clamps, and cutting tissue using surgical scissors.
Researchers have tested the robot's capabilities in unexpected circumstances, such as changing its position before the operation begins, or adding blood-like dyes to change the appearance of tissues. However, he managed to carry out the operation with high accuracy and without errors.
As residents learn surgical procedures gradually and at different speeds, this robot shows great promise in developing systems capable of performing operations autonomously in the future, says surgeon Jeff Jopling, co-author of the study.
The team plans to expand the use of the new system to other types of complex surgeries, while developing its capabilities so that in the future it can perform complete surgeries autonomously in real clinical environments.