Teams from the University of California, Berkeley, Google Brain and Intel Corporation have developed a robot capable of simulating suturing.
Through a project called Motion2Vec, the researchers used 78 instructional videos to train the two-armed DaVinci robot’s artificial intelligence (AI) to insert needles into a cloth device, thereby mimicking a suturing motion.
The team used Siamese networks — a deep learning configuration wherein two or more networks share the same data — to enable the robot to mimic the suturing motions gleaned from the instructional videos. The Siamese networks are typically used in facial detection, and language and signature verification applications.
Although the technology’s use in an operating room is a long way off, researchers believe that eventually, repetitive tasks during surgery, such as suturing, could one day be accomplished by robots, thereby freeing up surgeons to focus on more complex surgical tasks.
To see the robot simulate suturing, watch the accompanying video that appears courtesy of Google.
