We’ve been seeing how AI has been used rather extensively in the medical industry, where AI has been trained to spot certain things that a doctor might have missed. This is because as humans, there are many things that can affect our judgement, like being tired, being stressed, and so on.
This becomes especially more critical during surgery, where certain procedures could run for hours and could leave surgeons tired. However, thanks to the efforts of researchers, it seems that in the future, robots could take over certain more mundane (but still critical) roles in surgery, such as stitching a patient up after an operation.
This collaboration is between Intel and the University of California, Berkeley, where researchers have developed an AI deep learning system dubbed Motion2Vec which teaches a robot how to suture. This is done by feeding the AI publicly available surgery videos where it then tries to break down the movements of surgeons, like where to insert the needle, extraction, and hand-off.
Speaking to Engadget, lead researcher Dr. Ajay Tanwani said, “There’s a lot of appeal in learning from visual observations, compared to traditional interfaces for learning in a static way or learning from [mimicking] trajectories, because of the huge amount of information content available in existing videos.”
So far based on early tests, the robot is capable of performing a suture with 85.5% segmentation accuracy with an average of 0.94cm error in targeting accuracy. It will probably be years before we see robots take over the suturing role in the operating theater, but for now it certainly looks promising.