Almost 40 years ago, the Agency for Advanced Defense Research Projects (DARPA) and NASA began promoting projects that allow remote surgical operations to be performed on the battlefield and in the space. Among these initial efforts, robotic surgical systems like Da Vinci, which act as an extension of the surgeon, have emerged, allowing minimal invasive procedures to be performed using remote control and 3D vision. But this involves humans using sophisticated tools. Today, the integration of artificial intelligence and machine learning generated in the control of systems like Da Vinci may bring autonomous surgical robots closer to reality.
This Wednesday, the Journal Science robot We have published the results of a study conducted by Johns Hopkins and Stanford University researchers. They present a system that, like health residents, can learn from human videos that operate and receive in natural language and autonomously perform several steps of a surgical procedure.
Just like human learning, a team of scientists were gradually teaching the robots the steps needed to complete the surgery. Last year, the Johns Hopkins team led by Axel Krieger performed three basic surgical tasks: training robots, handling needles, lifting tissue, and sutures. This training was done through imitation and machine learning systems similar to those used in ChatGpt, but instead of words and text, it uses a robotic language that converts the angle of machine movement into mathematical data.
In the new experiment, two experienced human surgeons performed a demonstration of gallbladder removal surgery on pig tissue outside the animal. They used 34 gallbladders to collect 17 hours of data and 16,000 trajectories. The robot was then able to perform some of the 17 tasks needed to remove organs with 100% accuracy, using eight gallbladders that had never been seen before, without human intervention, identifying specific ducts and arteries, holding them accurately, strategically positioning clips, and cutting them with masses. During the experiment, the model was able to correct its own mistakes and adapt to unexpected situations.
In 2022, this same team performed their first autonomous robotic surgery on living animals: laparoscopy of pigs. However, the robot needed tissue to have special markers, was in a controlled environment and followed a pre-established surgical plan. In a statement from his facility, Krieger said it was like teaching a robot to drive a carefully mapped route. The new experiment that has just been presented is that for robots, relying solely on a general understanding of how cars drive, like driving on roads that they have never seen before.
José Granell, director of the ENT and Neck and Neck Surgery at HLA Moncloa University Hospital and professor at the European University of Madrid, believes that the work of the Johns Hopkins team is “starting to approach something that resembles a real-life surgery.”
“The problem with robotic soft tissue surgery is that biology has a lot of inherent variation. Even if you know the techniques, there are many possible scenarios in the real world,” explains Granel. “It's easy to ask a robot to carve the bones, but in soft tissues, moving makes everything difficult. When you push, you can't predict how much it will move or whether the artery will tear if it pulls violently.
For Krieger, this advancement takes us “from robots that can perform certain surgical tasks to robots that really understand surgical procedures.” The team leader behind this innovation made possible by generative AI argues:
Francisco Kraska, professor of human anatomy and embryology at the Autonomous University of Madrid, welcomes the study but points out that it is “a very simple surgery.” Furthermore, robots are much slower than humans who perform the same task.
Mario Fernandez, head of Head and Neck Surgery at Gregorio Malanun General University Hospital in Madrid, finds advance payments interesting, but he believes replacing human surgeons with machines is “a long journey.” He also pays attention to being fascinated by technology without fully understanding its true benefits, and points out that its high cost means that no one can access it.
“For example, I know a hospital in India. For example, I have a robot, I can do two surgical sessions a month, and I can operate two patients. A total of 48 people a year. Robotic surgery may be a way to practice and learn, but it's not a reality for patients,” says Fernandez. He believes we should be grateful for the advancement of technology, but surgery must be judged by what actually reaches the patient. In contrast, he points out that “a technique known as oral ultrasound surgery, developed in Madrid and available worldwide, is performed on six patients every day.”
Krieger shows that their proof of concept can be performed autonomously in complex surgical procedures, and that mimic learning systems can be applied to more types of surgery.
Granel noted that not only overcomes technical challenges, but also slows adoption of surgical robots as they are “very conservative about patient safety” in surgery.
He also raises philosophical questions such as overcoming the first and second laws of robotics of Isaac Asimov. “Robots may not injure humans or allow humans to do harm through inaction.” The expert highlights the obvious contradiction in the fact that human surgeons “causing harm, but pursue the benefit of the patient. This is a dilemma. [for a robot] It needs to be resolved. ”
Sign up Weekly Newsletter To get more English news coverage from ElPaís USA Edition
