Key Points
- Janus AI, led by MBZUAI researchers, took first prize at Europe's robotic urology innovation challenge.
- The system matches live operating-room video to 30-second clips of similar past procedures in milliseconds.
- Its backers position it as a training tool for residents and less experienced surgeons.
The latest:
An AI system that feeds surgeons video from past robotic operations while they operate won first prize at the Robotic Urology Breakthrough Innovation Challenge, held at ERUS26, described as Europe’s largest robotic urology meeting. The system, Janus AI, was built by a team led by researchers from the Mohamed bin Zayed University of Artificial Intelligence in Abu Dhabi, working with two partner universities.
Details:
- How it works: According to MBZUAI, Janus AI is a video-to-video retrieval system. It takes footage captured live in the operating room and matches it against a library of 30-second clips drawn from similar surgeries, putting reference material from earlier procedures in front of the surgeon as the operation is underway.
- The speed question: Features from every clip in the library are precomputed and stored in advance, so only the clip from the ongoing surgery has to be processed live. The similarity search runs in RAM and typically takes milliseconds, according to the university, which is what makes real-time use possible.
- The technical method: Each library clip is represented by features extracted using a neural network, MBZUAI said. Operating-room video is then matched to the library through an exact similarity search across those representations, rather than an approximate one.
- Beyond the clip: Surgeons are not limited to the matched 30-second segment. They can keep watching the full-length reference video to see how the rest of a comparable procedure played out, according to the university.
- The teaching case: The researchers say Janus AI could serve as an educational tool for residents and less experienced surgeons. Cesare Stefanini, Professor of Robotics at MBZUAI, who supervised the work, framed the value in curation: “For a junior surgeon, this is gold.”
- Who built it: The project is a collaboration between MBZUAI, the University of Illinois Chicago and the University of Turin. Edoardo Fazzari, a postdoctoral researcher at MBZUAI, was the main contributor, under Stefanini’s supervision.
- The clinical partners: The team also included Simone Crivellaro, Luca Morgantini and Chiara Gumiero from the Department of Urology at the University of Illinois Chicago, alongside Alberto Vignola of the University of Turin’s Medical School, according to MBZUAI.
- What comes next: Future work will expand the video library with hundreds of additional surgical videos, bring graduate students into development and validation, and move toward clinical translation. MBZUAI did not set a timeline for clinical deployment or name the hospitals where it would be trialed.
Background:
MBZUAI is Abu Dhabi’s graduate university dedicated to artificial intelligence. ERUS is the robotic urology section meeting of the European Association of Urology, and RUBIC is the innovation competition run alongside it.
Between the lines:
The design choice carries the argument. By precomputing library features and keeping the search in RAM, the team traded storage and preparation for latency — the one variable that decides whether a tool is usable mid-operation or not. That, plus Stefanini’s point about restricting the library to the best procedures, suggests the value proposition is curation and timing rather than diagnosis, which is a lower regulatory bar to clear.
What’s next
Watch for the expanded video library, the validation work involving graduate students, and any move from competition prize to a clinical setting — the step MBZUAI has named but not dated.
