A team at Johns Hopkins University in Baltimore trained a robotic surgery system using a technique much like that of ChatGPT. The “machine” successfully performed the operation on its own on a lifelike patient. It’s the first time this has been done.
Surgical robots are now increasingly used during procedures in hospitals in the most advanced countries.
They are normally guided by humans in the operating theatre, using special, highly sophisticated “joysticks” that control the mechanical arms and probes inserted in the patient’s body. Robots enable movements that are impossible with standard laparoscopic surgery, improving outcomes for certain types of operations (especially those on the prostate, liver and rectum). What’s more, they also feature a mini camera that provides the surgeon with a 3D image of the surgical site, magnified about 10 times.
But if all this works so well, then why do researchers at Johns Hopkins University and other centres want to hand everything over to the “machine”, without human assistance?
Surgical robots could prove very useful in the event of staff shortages and emergencies, suggest the researchers. But that’s not all: when its self-learning reaches peak levels and has been certified, this technology may even be able to do some special procedures or parts of certain operations (e.g. initial preparation) more precisely and effectively than humans.
The first autonomous surgery by a surgical robot
A surgical robot has autonomously performed a lengthy phase of a gallbladder removal for the first time (later repeating the same operation eight times, with excellent results), by learning directly from videos of real-life surgeries. During the operations, which were conducted on highly realistic lifelike patients, the machine not only exhibited expertise and precision, but also responded to voice commands from surgeons in the operating theatre, just as a young surgical resident would.
Advances in robotic surgery
These results have been described in the journal Science Robotics by surgeons and engineers at Johns Hopkins University in Baltimore (United States), a leading international centre for so-called autonomous surgery. This is a major breakthrough, explained the researchers: surgical robots are now no longer simply machines that execute decisions made by human surgeons, but systems that can adapt in real time, carrying out procedures autonomously and correcting errors on their own.
With this advancement, robots “truly understand surgical procedures,” said Axel Krieger, leader of the study. “This is a critical distinction that brings us significantly closer to clinically viable autonomous surgical systems that can work in the messy, unpredictable reality of actual patient care.”
The new system, known as Surgical Robot Transformer-Hierarchy (SRT-H), is based on the same machine learning architecture that powers language models like ChatGPT. One of its strengths is its ability to interact: it can respond to voice commands such as “grab the gallbladder head” or corrections like “move the left arm a bit to the left”. The robot learns from the feedback, honing its skills through continuous interaction.
Fully autonomous surgery on a pig three years ago
This isn’t the first time that Krieger’s team has conducted trials on autonomous surgery. In 2022, the STAR – Smart Tissue Autonomous Robot – prototype performed a laparoscopic surgery on a pig. In that study, though, the conditions were strictly controlled: the tissue to be removed was specially marked and the robot followed a predetermined plan, a bit like driving along a pre-mapped route. But with SRT-H, the approach is much more ambitious: "It is like teaching a robot to navigate any road, in any condition, responding intelligently to whatever it encounters,” said Krieger.
The experiments also showed the system's ability to deal with unexpected scenarios. For example, the researchers changed the robot’s starting position and even altered the appearance of the organs using blood-like dyes: in all these circumstances, SRT-H adapted, working “unflappably” and without errors. “This work represents a major leap from prior efforts because it tackles some of the fundamental barriers to deploying autonomous surgical robots in the real world,” said the study’s lead author Ji Woong “Brian” Kim, who’s now with Stanford University.
However, there’s still a long way to go before a completely autonomous procedure is possible in the operating theatre. The researchers want to extend the training to other types of operations and expand the system's capabilities to perform a fully autonomous robotic surgery one day, from start to finish. But for now the message is clear: surgical robots are emerging from experimental laboratories to replace – at least in part – human activities. It’s only a matter of time.
