A machine that learned by observing real surgical procedures for the first time successfully performed surgical steps with human-like precision. The study was led by researchers at Johns Hopkins University in Maryland, USA, with contributions from Stanford University experts, and was featured at the Conference on Robot Learning in Munich.
The successful results of the robot’s learning model enable experts to avoid manually programming each individual movement, improve operational efficiency, and move robotic surgery closer to full autonomy.
“Having this model is remarkable—by simply providing camera input, it can determine the robotic actions required for surgery. We see this as an important milestone toward a new era in medical robotics,” said lead author Axel Krieger.
Imitation Learning for Surgical Manipulation
The researchers trained the da Vinci Surgical System–based robot using an imitation learning approach, enabling it to carry out basic tasks like handling a needle, lifting tissue, and stitching. The method blends imitation-based training with a ChatGPT-like framework, but instead of processing text, it relies on kinematics to generate precise physical movements.
The researchers trained the system using a large collection of surgical videos captured from procedures performed with da Vinci prototypes around the world. With nearly 7,000 da Vinci systems in use globally and more than 50,000 surgeons trained on the platform, the extensive dataset allowed the model to learn from a diverse range of surgical techniques and operations.
Although the system is already widely used and marks significant progress toward automated procedures, it still lacks full precision. The study proposes that overcoming this limitation may require training the model to perform relative movements instead of absolute ones, which tend to introduce more errors.
The model learns behaviors that developers never explicitly programmed. For example, if it drops a needle, it can automatically retrieve it and continue the task—something I never directly taught it to do,” Krieger said.
Imitation Learning Expands Toward Fully Autonomous Surgical Procedures
The team is now applying imitation learning not only to simple surgical actions but also to complete procedures, aiming for minimal or even no human intervention. Previously, programming such systems required extensive and rigid planning.
“That approach is very restrictive. What’s different now is that we can simply collect imitation data from various procedures and train a robot within days. This helps speed up autonomy, reduce medical errors, and improve surgical precision,” Krieger added.
The institution and its researchers published a video titled “Robotic Surgery Demonstration” showcasing the robot carrying out the tasks.
Read the original article on: Revista Planeta
Read more: New gel developed by scientist regenerates tooth enamel