Johns Hopkins and Stanford researchers have pulled off a significant milestone in robotic surgery—they've taught a surgical robot to handle basic procedures just by watching videos of human surgeons doing their thing.
The team managed to program the da Vinci Surgical System to pull off three basic but critical surgical tasks: handling needles, lifting tissue, and stitching up wounds. And it's doing it with the same level of skill you'd see in experienced human surgeons. This is a massive step toward fully autonomous robotic surgery.
"It's remarkable to have this model where we simply provide camera input and it predicts the necessary robotic movements for surgery," says Dr. Axel Krieger, assistant professor at JHU's Department of Mechanical Engineering and senior author of the study. "We believe this represents a significant leap forward in medical robotics."
For this, the research team combined imitation learning with architecture similar to the language model used in ChatGPT. But instead of focusing on words, their model breaks down surgical movements mathematically, turning them into exact robotic actions.
To get the robot trained, they fed it hundreds of surgical videos taken from wrist-mounted cameras on da Vinci robots during actual surgeries. This hefty dataset comes from the nearly 7,000 da Vinci systems worldwide, used by over 50,000 trained surgeons.
While the da Vinci system is widely used, precision has been a weak spot. The research team tackled this by having the robot perform movements in relation to its surroundings rather than making exact, preset movements, and it made a big difference.
One of the most incredible things is how adaptable the systems become. "The model is so good at learning things we haven't taught it," says lead author Ji Woong "Brian" Kim, a postdoc at Johns Hopkins. "Like if it drops the needle, it will automatically pick it up and continue. This isn't something I taught it to do."
This development means there's no longer a need to painstakingly program each specific movement for a surgical task, which used to take years to get right for just one type of procedure. With this new approach, the robot can learn a full procedure by watching in just a few days.
They presented their findings at the Conference on Robot Learning in Munich, showing off how AI and medical technology are coming together in surgical robotics.
Are you a techie who knows how to write? Then join our Team! Wanted:
- News Writer (Romania based)
Details here
Source(s)
Johns Hopkins University (in English)