Fully automated surgical robots are becoming a reality

Researchers have made a real breakthrough by introducing a surgical robot training system based on imitation learning. This achievement marks an important milestone, opening up limitless possibilities for the application of robots in the medical field, promising to bring about a revolution in surgery.

A robot that has undergone extensive training, learning from thousands of hours of video footage of surgeries performed by leading experts. Thanks to advanced machine learning algorithms, the robot is able to simulate and perform surgical operations with incredible precision and dexterity, gradually getting closer to the level of a real surgeon.

The successful application of imitation learning in training surgical robots has created a significant breakthrough. Instead of having to program each meticulous movement, robots are now able to self-learn and simulate surgical skills from leading experts, bringing higher efficiency and accuracy in surgeries.

The researchers say the achievement brings us closer to a future where surgical robots can perform complex operations without the need for direct intervention from a doctor. It is a major step forward, opening up limitless possibilities for the application of robotics in the medical field.

The breakthrough findings by a team from Johns Hopkins University and Stanford University in the US will be presented at the Robot Learning Conference in Munich, Germany, this weekend. The event is considered one of the leading forums on artificial intelligence (AI), and their research results promise to attract great interest from the international scientific community.

"What's really impressive is that we just need to provide the model with image data from the camera, and it can automatically predict the entire sequence of complex movements that the robot needs to perform during surgery," shared Professor Axel Krieger, Associate Professor of Mechanical Engineering at Johns Hopkins University (USA).

"We believe this is a significant step towards a new area of ​​medical robotics," added Professor Krieger. "The model is very good at learning things that we haven't taught it. For example, if it drops a needle, it will automatically pick it up and continue. This is not something I taught it to do."

The team successfully trained the Da Vinci robotic surgical system to perform three basic surgical skills, including needle manipulation, tissue lifting, and suturing. Surprisingly, the robot achieved performance comparable to that of professional surgeons.

The Da Vinci Surgical System, with more than 7,000 units in use globally, has become a popular robotic surgery platform. With more than 50,000 surgeons trained to use the system, it has created a huge database of surgical cases, providing an invaluable resource for robots to learn and improve their skills.

Unlike ChatGPT which uses natural language, this model leverages a machine learning architecture to analyze and predict robot movements based on kinematic language. Instead of processing text, this model works with data about the robot's position, velocity, acceleration, and other technical parameters.

The researchers collected a large amount of video data from hundreds of real surgeries, recorded directly from the perspective of the Da Vinci robotic arm. This dataset includes detailed images of complex surgical procedures, providing a valuable source of information to train the model.

Despite the widespread use of the Da Vinci surgical system, previous studies have shown that it often suffers from accuracy issues when performing surgical maneuvers. The research team has found a new approach to address this issue by focusing on training the model to perform relative movements rather than absolute actions. This helps to increase the robot's accuracy and flexibility during surgery.

“All we need is the image data and then this AI system will figure out the appropriate action,” said lead author Ji Woong “Brian” Kim, a postdoctoral fellow at Johns Hopkins University.

“We found that even with a few hundred simulations, the model could still learn the process and generalize to new environments it had never encountered before,” added Ji Woong “Brian” Kim.

The researchers believe that this model could revolutionize the training of surgical robots. Instead of having to manually program each surgery, the model has the potential to quickly train a robot to perform any procedure, from simple to complex. By applying imitation learning, the robot can learn from expert surgeons and perform the entire surgery autonomously.