Toronto surgeons test AI that flags safer and riskier cutting areas

In short: Surgeons at Toronto General Hospital are developing an AI tool that analyzes live surgery video and highlights areas that look safer or riskier to cut.

What's going on

A New York Times report describes work led by Dr. Amin Madani at Toronto General Hospital, which is part of University Health Network. The team is building an AI system that watches the same video feed surgeons see during an operation and then marks parts of the image in color.

In a demonstration, the tool overlaid green and red regions on the surgical view. You can think of it like a GPS map that marks some roads as low risk and others as higher risk, while the driver still decides where to go. The hospital describes the system as real-time guidance, not an autonomous surgeon.

The technology relies on computer vision, which means software that learns to recognize what it is looking at in images and video. It is trained on past recordings of operations that expert surgeons have labeled, showing where it is appropriate to cut and where it could be dangerous. An earlier prototype was trained on hundreds of hours of gallbladder surgery video, and a study of 290 videos from 153 surgeons reported 93 to 95 percent accuracy at identifying structures and “go” and “no-go” zones.

What to watch

Researchers still need to show that the tool helps in everyday hospital use, including whether it actually reduces mistakes and complications. It also has to work when the camera view is messy, or when a patient’s anatomy looks different than usual. Another open question is how to show uncertainty, since a colleague of Madani’s, Dr. Timothy Jackson, noted some areas may need a “caution” label instead of a simple red or green choice.

Source: NYTimes