A robot, trained for the first time by observing videos of experienced surgeons, performed surgical procedures with the same proficiency as its human counterparts.
By leveraging imitation learning to train surgical robots, healthcare professionals can bypass the need to painstakingly programme each step of a medical procedure, thereby bringing the field of robotic surgery closer to true autonomy, where robots can proficiently perform complex surgeries independently.
Here is the rewritten text:
According to senior writer Axel Krieger, having a mannequin like this is truly magical: by feeding it digital input, it can accurately predict the robotic actions required in surgery. “We envision this milestone as a significant leap forward in pioneering a novel frontierspace in medical robotics.”
Researchers at Johns Hopkins University are showcasing their discoveries this week at the Convention on Robotic Learning in Munich, a premier event for robotics and artificial intelligence.
Researchers from Stanford College collaborated with the workforce to utilize imitation learning in training the da Vinci Surgical System robot to perform basic surgical tasks, including manipulating a needle, lifting bodily tissue, and suturing. The mannequin integrated imitation learning with a comparable machine learning architecture that serves as the foundation for ChatGPT. Notwithstanding its proficiency in processing phrases and textual data, this model communicates in a “robotic” tone, translating human kinematics into mathematical formulas.
Researchers trained a simulator by feeding it vast quantities of film footage taken from wrist-mounted cameras attached to Da Vinci robots during various surgical operations. Captured by surgeons globally, these films serve as a valuable resource for postoperative assessment and subsequent archival purposes. With over 7,000 Da Vinci robots deployed globally, and more than 50,000 surgeons trained on the platform, a vast repository of expertise has been amassed, allowing the robots to learn from these collective experiences.
While the Da Vinci surgical robot has widespread adoption, researchers criticise its notorious lack of precision. Despite the initial difficulties, the team managed to develop a workaround that successfully addressed the problematic input. What’s crucial here is teaching the mannequin to execute actions in relation to its environment, rather than fixating on precise, absolute movements that may not be accurate in every situation.
Ji Woong “Brian” Kim, the lead writer, succinctly expressed the team’s desire: “We simply want the AI system to identify the correct movement after capturing a single frame.” As demonstrated by only a handful of examples, the model is capable of being trained and adaptably applying the learned process to novel situations it has not previously experienced.
The workforce trained the robot to perform three tasks: manipulate a needle, lift physical tissue, and sew. The trained robot model replicated surgical procedures with precision and dexterity equivalent to those of human surgeons.
According to Krieger, the doll excels at learning topics we haven’t yet instructed it on. When the record drops, it will consistently pick itself back up and continue. This isn’t something I was ever taught to do.
Researchers suggest that a mannequin could be utilised to expedite the development of a robot capable of performing various surgical procedures with precision. The workforce has begun leveraging imitation learning techniques to train a robot to perform not only minor surgical tasks but a full-fledged surgical procedure.
Prior to this breakthrough, scripting every aspect of a robotic-assisted surgery from scratch was the norm, requiring extensive manual coding for even the most straightforward procedures. Someone may spend a decade futilely attempting to master suturing, Krieger noted. Suturing is a fundamental technique used in various types of surgical procedures?
“It’s extremely confining,” Krieger said.
“What’s unique about this process is that we can quickly train a robot to mimic various procedures through imitation learning, which can be achieved in just a couple of days.” It enables accelerated pursuit of autonomy while reducing medical errors and achieving more precise surgical outcomes.
Researchers from Johns Hopkins University are represented by PhD student Samuel Schmidgall, Affiliate Research Engineer Anton Deguet, and Associate Professor of Mechanical Engineering Marin Kobilarov. The research on Stanford College, authored by Dr. Tony Z., a respected scholar with a PhD in his field. Zhao
