Soft tissue moves, stretches and slides the moment a surgeon cuts it, which is exactly why handing a scalpel to a machine is so hard. A Johns Hopkins team now reports a robot that kept track of those changes and removed tumors from eight lab-made kidneys, making 77 cuts without a single error.
The result, announced by the Department of Mechanical Engineering on November 9, 2026, is a research milestone in a controlled setting. The kidneys were synthetic gel models, not patients, and the work is news about robotics research, not medical guidance.
What the team built
According to the Johns Hopkins release, the system uses two robotic arms and existing robotic surgery tools. One arm carries a vacuum gripper that gently pulls tissue away. The other holds an electrosurgical tool that does the cutting. The team was led by Axel Krieger, associate professor of mechanical engineering, and the details appeared online in October in IEEE Transactions on Robotics.
The robot was trained with AI modeling for two jobs: simulating how kidney tissue reacts to incisions, and estimating where the tumor sits relative to the kidney. As the operation proceeds, the system checks its own picture of the tissue against real-time data and recalibrates its approach.
How the test worked
The researchers made eight synthetic kidneys from hydrogels, which are polymer-based gels, modeled on CT scans of people with localized kidney tumors. PhD student Ethan Kilmer noted the most visible difference from a real kidney: the synthetic one does not bleed.
The goal mirrored common surgical practice. The robot was asked to remove the tumor plus a 6 millimeter border of surrounding tissue, so that all cancerous material comes out while as much healthy tissue as possible stays in. Across all eight trials the robot stayed within 1.7 millimeters of that margin goal and removed all cancerous tissue, as checked by CT scans taken before and after.
- Eight synthetic kidneys, each built from a real patient’s scan
- 77 cuts in total, with no errors reported
- Margin goal of 6 mm, met within 1.7 mm every time
- Pre- and postoperative CT scans used to confirm complete removal
Why tracking mattered
The most telling experiment was a comparison. Postdoctoral scholar Pit Henrich, who had earlier built the AI system for different research and adapted it to the kidney, said the team also ran surgeries where the robot did not update its understanding of how the tissue changed. Those all failed. With updating switched on, the robot could follow the tumor as the model deformed and was cut, and use that tracking to decide where to cut next.
Krieger framed it this way: removing tumors embedded deep in tissue is challenging because tissue deforms during surgery and surgeons must mentally connect preoperative images with a changing scene. Showing that a robot can do that is, in his words, a major milestone toward greater autonomy in soft tissue surgery.
What it means
For surgical robotics, the interesting part is not the cutting but the continuous re-evaluation. Pre-operative scans are only a starting map, and the tissue stops matching that map as soon as the work begins. A system that rebuilds its understanding every few moments is closer to how a surgeon actually works.
For hospitals and device makers, this is an early signal rather than a purchase decision. The release is clear that robotic surgery is not fully autonomous or ready for widespread use. The same group has previously worked on training robots from surgical videos and on a gallbladder procedure performed without human help, so this kidney study extends a steady line of research.
The team’s next steps are plain. Jiawei Ge, a co-author now working as a roboticist at Inner Logic, said the next test is on organic tissue. Longer term, the group wants to shrink the approach into minimally invasive surgery, operating through small incisions with an endoscopic camera rather than large open incisions.
- Watch for tests on organic tissue, which will bring bleeding and more complex motion
- Watch for a move from open surgery to endoscopic, small-incision methods
- Remember the current evidence is eight gel models, not clinical results
The bottom line
A robot that can find a tumor, cut around it with a precise margin and keep correcting itself as the tissue shifts is a meaningful step for autonomous surgery research. It is also a lab demonstration with a long road ahead. The Johns Hopkins team has shown the hard part, tracking deforming tissue, can be solved in a controlled test.