Can Artificial Intelligence Learn the Art of Surgery?

Canada wants to leverage A.I. to improve its health care system. At a Toronto hospital, surgeons are training a machine to distinguish a safe incision from a catastrophic one.Can Artificial Intelligence Learn the Art of Surgery?Canada wants to leverage A.I. to improve its health care system. At a Toronto hospital, surgeons are training a machine to distinguish a safe incision from a catastrophic one.Under shadowless operating room lights at Toronto General Hospital, two surgeons clipped away at a pear-shaped gallbladder they were removing from a woman on a spring morning.“Can we get the A.I. on, please?” called Dr. Amin Madani, an endocrine and acute care surgeon.On a large monitor, the view from a camera inside the patient bloomed with green and red splotches showing where it was and was not safe to cut. The overlay was powered by artificial intelligence, and Dr. Madani has been training the model to think like the best surgeons.His colleague, Dr. Timothy Jackson, pointed out that the system might need to show another color: caution yellow, to represent the zones always requiring human judgment that no algorithm has learned to mimic.The surgical field — what doctors call the area of a body under operation — is among the latest frontiers for artificial intelligence in Canada, a country that has earned a global reputation for the early development of A.I. and is home to some of its pioneers.ImageDr. Amin Madani alongside the computer system that powers his A.I. surgical guide program at Toronto General Hospital.Canada spends more on health care than most developed countries, yet its universal, primarily tax-funded system is beset by long waits for family doctors and a shortage of hospital beds, making the promise of A.I. to improve care while cutting costs enticing to the government.Prime Minister Mark Carney unveiled Canada’s national A.I. strategy at an event at Toronto General, ranked among the world’s best, describing the plan as a way “to harness A.I. to improve lives and, indeed, to save them.”The push for Canada to step out of Silicon Valley’s shadow is driven partly by the rupture in the country’s trade relationship with the United States and a desire to assert its sovereignty, even in the digital world. It also builds on Canada’s history of punching above its weight in clinical research, including inventing the world’s first surgical robot and performing the first successful lung transplant.Now, with a scalpel, Canadian researchers are testing the bounds of what A.I. can do.‘That’s the Art of Surgery’Surgeons have long trained on an apprenticeship model: they observe others at work before operating themselves, and eventually master the techniques well enough to instruct a peer. The American surgical pioneer Dr. William Stewart Halsted called it “see one, do one, teach one.”Over the last few years, Dr. Madani, 40, became fascinated by how surgeons learn — their cognitive processes, their behavioral psychology, what it takes to become elite in the operating room. During the pandemic, when operating rooms were largely closed, he designed tools to train residents to read the surgical field, visually, like putting pilots in a flight simulator.Then A.I. arrived in medicine.“All the work that I had done was like a perfect setup to train A.I. algorithms to think and behave like expert surgeons,” said Dr. Madani, who founded the Surgical Artificial Intelligence Research Academy at the University Health Network, a public research and teaching hospital group in Toronto known as U.H.N. “It was all very serendipitous.”Laparoscopic operations are recorded by a camera inserted through a small incision, generating the video that trains Dr. Madani’s A.I. model. One of the first algorithms out of his lab was for the common gallbladder removal, and was aimed at improving precision to avoid bile duct injuries. The rare complication is the leading cause of medical malpractice lawsuits against general surgeons in Canada, the United States and the United Kingdom, according to a 2025 peer-reviewed journal article.Though gallbladder operations are routine, surgeons can misread the anatomy from inattention or overconfidence, a gap A.I. tools are meant to fill, said Nicolas Padoy, the scientific director of the Institute of Image-Guided Surgery in Strasbourg, France, which is conducting similar research. ImageA.I. guided “go” and “no-go” zones and a live view inside a patient during gallbladder removal surgery.ImageAn iPad is used to control a monitor showing A.I. determined zones that guide surgeons on where to make incisions.Dr. Madani’s tool — called GoNoGoNet — functions like a GPS for the operating table — recognizing anatomical patterns and predicting “no-go zones,” places where an incision would cause a major complication. Like a stoplight, green indicates cut and red means don’t.But, “there’s a yellow zone there, and that’s the art of surgery that can’t sort of be calculated in zeros and ones yet in an A.I. model,” said Dr. Jackson, the operating room medical director of general surgery at U.H.N.In the operating room, the 44-year-old patient’s heart monitor beeped steadily during the roughly 40-minute procedure. (The patient agreed to let The Times observe her surgery.) After the surgeons inflated her abdomen with carbon dioxide to create room for their instruments, Dr. Jackson took his place beside Dr. Madani.Two monitors — one with the A.I. overlay and one without — showed the same view from inside the patient’s abdomen via the laparoscope, a tubular surgical instrument with a tiny light and camera. The surgeons exposed, clipped and cauterized the duct and artery connected to the gallbladder while watching the monitor without A.I., because they said they sometimes find the red and green overlay distracting. Both options are always available to surgeons. Once the organ was free, they slipped it into a bag to keep it from rupturing.“I didn’t personally see it coming, but I find myself in a world where I think interacting with A.I. is unavoidable,” Dr. Jackson said.Trust IssuesA.I. has spread through nearly every corner of North American health care, automating clerical work like clinical notes and patient correspondence while speeding diagnoses and identifying personalized treatments. In the United States, the Trump administration is pushing to integrate it into medical care at a pace that some in the field find unnerving.The Toronto surgeons said that GoNoGoNet has been quite accurate.“I was very proud of my A.I. model,” Dr. Madani said, beaming after the procedure. He performs about two gallbladder removals a month using the tool for guidance as part of a clinical trial that requires consent from participants.How much patients need to know about A.I. before agreeing to its use in their care — and how they consent to their data being used, particularly if the models are later commercialized — is one of the field’s major ethical questions, said Dr. Jay Shaw, acting director of the University of Toronto Joint Centre for Bioethics.ImageDr. Madani, left, and Dr. Timothy Jackson, perform laparoscopic cholecystectomy surgery on a patient at Toronto General Hospital, with the help of visual A.I. guides.Another concern is whether A.I. in the operating room will make surgeons more dependent on their tools than their skill. Even automating clinical notes can erode critical thinking, said Dr. Benjamin Chin-Yee, a hematologist and professor at Western University in London, Ontario, who studies how new technologies affect patient care.“I worry that if we are so quick to outsource that to A.I., a lot of those important steps in the reasoning process will be lost,” he said.It may be a while before that becomes pressing. A 2023 peer-reviewed study found emergency surgeons divided over whether to let A.I. guide surgical decisions. Dr. Madani still sees the gap between early adopters and skeptics. But, “most of us are pretty excited about the digital age of surgery,” he said.Vjosa Isai is a reporter for The Times based in Toronto, where she covers news from across Canada.Related ContentAdvertisementSKIP ADVERTISEMENT