If you’re a robot, you can become a surgeon by binge-watching videos.
Researchers from Johns Hopkins and Stanford universities did just that with the da Vinci Surgical System, a robotic platform widely used for minimally invasive surgeries. Instead of traditional, painstaking programming, where each movement must be coded step-by-step, the team used imitation learning.
They fed the robot data from hundreds of recorded surgeries, allowing it to analyze and mimic essential surgical tasks like needle handling, tissue manipulation, and suturing. The method, known as imitation learning, takes a vastly different approach from traditional programming.
Rather than being programmed for each precise movement, the robot observes real surgical procedures and learns to perform these tasks by copying what it “sees.”
The breakthrough allows the robot to adjust and refine its actions based on what it has learned, making it both more precise and adaptable than manually coded robots. The results showed the power of imitation learning: the da Vinci Surgical System, trained with this method, achieved high success rates across tasks and outperformed traditional programming in consistency and accuracy.
The researchers used a hybrid-relative action model, allowing the robot to use its own positioning and “wrist cameras” near its tools to improve depth perception and spatial awareness. The setup helps the robot make finer adjustments and succeed even in scenarios where the environment or setup changes, something fixed, camera-centric programming struggles with.
The researchers say the breakthrough brings the field of robotic surgery closer to true autonomy. They envision a future with robots performing complex surgeries without human assistance.
The team detailed their findings in the paper “Surgical Robot Transformer (SRT): Imitation Learning for Surgical Tasks,” which can be accessed here.