Researchers have trained a surgical robot using videos of experts, enabling it to learn tasks such as suturing and manipulating needles without the need for detailed manual programming. The robot demonstrated skills comparable to those of human surgeons, with the ability to adapt and correct mistakes, such as retrieving a dropped needle.
For the first time, scientists have trained a surgical robot to perform procedures simply by watching videos of expert surgeons, representing a revolutionary advance in medical robotics.
This method, called “learning by imitation,” allows robots to learn complex tasks by watching, without the need for detailed programming of each movement.
The study, led by researchers at Johns Hopkins University in collaboration with Stanford, used the same artificial intelligence framework that language models such as ChatGPT have adapted to control robotic movements.
While language models process words, this system translates the movements captured in the video into “kinematics,” or mathematical commands that guide the robotics.
Thousands of videos captured by cameras mounted on the robotic arms of the da Vinci surgical system were used for training. Nearly 7,000 da Vinci robots are in use worldwide, and more than 50,000 surgeons are trained on the system, creating a vast archive of data for the robots to “imitate.”
The team focused on three key surgical tasks: manipulating needles, lifting tissue, and suturing. The robot performed all of these actions with precision comparable to that of human surgeons and demonstrated unexpected abilities, such as retrieving a dropped needle without specific instructions.
The major innovation is that the system does not require manual programming, which can typically take years for simple tasks such as specific suturing. Now, with just a few hundred demonstrations, a robot can be trained in days to learn and perform procedures with high precision.
This paves the way for the development of fully autonomous surgeries, reducing medical errors and increasing patient safety. The results were presented at the 2024 Conference on Robot Learning in Munich in November, highlighting the importance of this advance.
The researchers are continuing to develop the model to train robots for complete surgeries, which promises to transform medical practice and make the technology more accessible and efficient.
READ MORE:
Commentaires