Johns Hopkins Robot Performs Realistic Surgery Without Human Help for the First Time
Johns Hopkins researchers built SRT-H, a surgical robot that autonomously performed a complete gallbladder removal phase on a lifelike patient model with 100% accuracy. The system learned from surgical videos and adapted to unexpected anatomical variations in real time.
Johns Hopkins University researchers have achieved a milestone in autonomous surgery: their Surgical Robot Transformer-Hierarchy (SRT-H) system performed a complete gallbladder removal phase on a realistic patient model without human intervention, achieving 100% accuracy across the 17-task procedure.
What distinguishes SRT-H from previous autonomous surgical systems is its ability to learn from watching surgery videos — using a machine learning architecture similar to ChatGPT — rather than requiring predetermined plans or specially marked tissue. The robot adapts to individual anatomical variations in real time, makes decisions on the fly, and self-corrects when things don't go as expected.
Lead researcher Axel Krieger described the significance: 'This advancement moves us from robots that can execute specific surgical tasks to robots that truly understand surgical procedures.' The system demonstrated expert-level precision even when confronted with unexpected scenarios during the operation.
The research, published in Science Robotics and funded by ARPA-H, represents a leap toward Level 3 surgical autonomy — where robots can perform conditional task automation under supervision. While fully autonomous surgery remains years away due to safety, regulatory, and liability challenges, the work establishes that AI-powered surgical robots can comprehend and execute complex procedures rather than simply following scripts.