The University of California at Berkeley (United States) has designed, in collaboration with Intel, artificial intelligence with partial supervision allowing a robot to make stitches. Thanks to deep learning technology, "Motion2Vec" learns to do the right things by analyzing videos of surgeons stitching up their patients. The system then succeeds in reproducing the gesture with precision, explains Engagdget .
"Five hundred hours of images are uploaded to YouTube every minute, it's an incredible fund, a database," comments Ken Goldberg of the University of California. A human is able to watch a video and understand what he is seeing, but robots are not able to see just a series of pixels. The goal of this project is to try to make sense of these pixels. "
Any help for surgeons?
After watching 78 videos, the system was already making stitches with an accuracy of 85.5% and an average margin of error of 94 millimeters, says a study published on May 31. "Learning through visual representations is likely to simplify downstream tasks, including segmentation and imitation actions", further summarizes a video presentation of the device posted on YouTube.
We are still far from seeing "Motion2Vec" in action in a surgical unit. But for one of its inventors, Ajay Tanwani, the system "could help surgeons use their time more productively to deal with the most complex tasks and to use technology to assist them in more action. basic and repetitive ”.
Health
Montpellier: CHU announces a world first with the operation of a patient with a liver tumor thanks to a robot
High-Tech
Chinese start-up Elephant Robotics designs first bionic cat
- Sciences
- High-Tech
- Artificial intelligence
- Robot
- Surgery