Experimental Surgery Performed by AI-Powered Surgical Robot

A major breakthrough in the field of robotic surgery has been achieved by researchers at Johns Hopkins University, who have developed an autonomous surgical robot, the Surgical Robot Transformer-H (SRT-H), capable of performing a cholecystectomy (gallbladder removal) with precision comparable to that of a human surgeon. Leveraging artificial intelligence and training based on surgical videos, this robot marks a decisive step toward surgical autonomy.

A revolution in robotic surgery

Since the introduction of surgical robots like the Da Vinci in the 1990s, robotic surgery has transformed medicine by offering unmatched precision and less invasive interventions. However, these systems relied until recently on a “master-slave” relationship, where a surgeon controls every movement via a console. The SRT-H, developed by Johns Hopkins’ team in collaboration with Stanford and Columbia, pushes these boundaries by integrating advanced artificial intelligence, based on a transformer architecture similar to ChatGPT’s.

Indeed, the SRT-H successfully performed cholecystectomies on eight pig gallbladder samples without any human intervention. This procedure, which consists of clipping and cutting the cystic duct and cystic artery, is one of the most common in the United States, with approximately 700,000 procedures per year. The robot demonstrated 100% accuracy, even when faced with anatomical variations or unforeseen conditions, such as tissues stained with dyes simulating blood.

How does the SRT-H work?

The SRT-H stands out for its language-guided learning by imitation approach. Unlike previous surgical robots, which followed preprogrammed instructions, this system was trained on approximately 18,000 video demonstrations of surgeries performed by experts on pig organs. These videos, captured via the endoscopic cameras and robotic arms of the Da Vinci, allowed the robot to learn the 17 steps of cholecystectomy, from identifying anatomical structures to precise cutting.

The system relies on two complementary artificial intelligence modules:

• High-level module: Plans tasks in natural language, such as “clip the cystic duct” or “adjust the left arm position,” adapting to unforeseen circumstances.
• Low-level module: Translates these instructions into precise trajectories for the robotic arms, ensuring smooth movements and real-time error correction.

Furthermore, the SRT-H can respond to voice commands, such as “grasp the gallbladder head” or “move the arm to the left,” and learn from these corrections to improve its performance. This adaptive capacity, tested against unforeseen scenarios like altered initial positions or damaged tissues, makes it comparable to a surgical resident guided by a mentor.

Comparable precision, but slower

Although the SRT-H achieved 100% accuracy in its trials, it is slower than an experienced human surgeon. This slowness reflects a tradeoff in favor of precision, with smoother movements and shorter trajectories than those of human surgeons, reducing the risk of errors. According to Axel Krieger, a medical roboticist at Johns Hopkins, “this work shows that it is possible to automate a complex surgical procedure with high robustness.”

Moreover, the tests were conducted on ex vivo organs, avoiding the complications of surgeries on living organisms, such as bleeding or respiratory movements. This limitation indicates that further work is necessary before application to human patients.

Implications for the future of surgery

This breakthrough opens promising perspectives for robotic surgery. By making robots capable of understanding and executing complex procedures autonomously, the SRT-H could:

• Improve access to care: By replicating the skills of the best surgeons, these systems could be deployed in remote areas or those lacking medical personnel.
• Reduce human error: The robot’s mechanical precision and adaptive capacity minimize risks associated with surgeon fatigue or tremors.
• Optimize surgical training: By integrating real-time feedback and artificial intelligence-based analytics, these systems could train novice surgeons.

However, challenges remain. Nuha Yassin, of the Royal College of Surgeons of England, emphasizes the need for careful exploration to ensure a safe transition to human trials. John McGrath, chairman of the NHS England robotics committee, estimates that clinical adoption could take several years, due to the complexities of real surgical environments.

Ethical and technical challenges

The integration of artificial intelligence in robotic surgery raises ethical and regulatory questions. Who would be responsible in case of error: the robot manufacturer, the AI developers, or the hospital? Moreover, dependence on training data quality poses a risk, as biases in videos could affect the robot’s performance. Finally, the high costs of development and integration into hospital workflows remain a major obstacle.

Furthermore, the transition to full autonomy (LoA Level V) will require overcoming challenges such as managing bleeding, unforeseen anatomical variations, or real-time technical failures. For now, the SRT-H operates at LoA Level IV, with minimal human supervision, but represents a significant leap forward compared to current systems without autonomy (LoA Level 0).

Toward an autonomous future in surgery

This breakthrough marks a turning point in robotic surgery, shifting from robots executing specific tasks to systems capable of understanding and adapting complex procedures. As Ji Woong “Brian” Kim, lead author of the study, explains, “our work shows that AI models can be reliable enough for surgical autonomy, an idea that seemed distant but is now achievable.”

In conclusion, the SRT-H illustrates the potential of artificial intelligence to transform surgery, offering mechanical precision and human adaptability. While tests on human patients are still necessary, this advancement brings the medical field closer to a future where autonomous robots could revolutionize surgical care, reducing complications and improving outcomes for patients.

Sources

https://arstechnica.com/science/2025/07/experimental-surgery-performed-by-ai-driven-surgical-robot