We may have made one of the biggest breakthroughs in robot-assisted surgery. Researchers at Johns Hopkins University (JHU) have achieved a milestone in surgical robotics after they trained a robot to perform complex medical procedures. Interestingly, the robot was solely trained by watching videos of human doctors at work. The robot has carried out some surgical procedures with precision similar to human doctors.
The researchers at JHU have termed this ‘the successful use of imitation learning’. This, according to them, ends the need to program robots with each individual move that is required during medical procedures. This brings robotic surgery closer to true autonomy, where robots would be able to perform complex surgeries without human intervention. The findings by the JHU researchers were recently showcased at the Conference on Robot Learning, a prestigious event for robotics and machine learning, held in Munich, Germany.
What do we know?
The da Vinci Surgical System is a robotic system that allows surgeons to perform minimally invasive surgeries—a technique that uses small incisions or no incisions at all for surgeries. The da Vinci robot imitates the hand movements of a surgeon in real time. The multiple arms of the robot are controlled by a surgeon using a hand-operated console. It comes with a robotic arm that manages the camera, lighting and a vision cart that supports a 3D high-definition vision system.
Based on the latest development, the da Vinci Surgical System robot learned and performed critical surgical tasks like needle manipulation, suturing, and lifting tissue. With the new imitation learning approach, the robot has been trained using hundreds of surgical videos captured by da Vinci robot wrist cameras.
Reportedly, the model combines a ChatGPT-like architecture, which teaches the robot to’speak surgery’ through mathematical movements. According to the researchers, the system also showed unexpected adaptability, such as autonomously collecting dropped needles—something that the robot was not programmed to perform.
The advancement in robotic capabilities, both in terms of training and dexterity, is bringing forth new use cases. This video learning approach would allow robots to learn and adapt rapidly to any new procedure, reducing the reliance on hand-coding for individual movement.
© IE Online Media Services Pvt Ltd
