AI Navigator in the Operating Room: Could a Russian Neural Network Become a Surgeon’s Copilot?
RTU MIREA and the Kulakov National Medical Research Center for Obstetrics, Gynecology and Perinatology have developed a prototype AI service that analyzes video of laser-assisted urological surgeries.
In minimally invasive surgery, every millimeter matters, and the endoscope’s diameter limits how much room a surgeon has to maneuver. A new neural network is being trained to monitor surgical procedures, with the long-term goal of becoming an intelligent assistant for physicians.
For now, the system analyzes video footage, identifying the stone, the laser fiber and the treatment area. But the longer-term question is broader: could the algorithm eventually move beyond reviewing recorded procedures to become a reliable assistant to the surgeon in real time?
The Digital Eye: Inside the Prototype
The joint KulakMirea team developing the new product brings together IT specialists and practicing physicians. Their challenge is far from trivial: they must teach computer vision to work not with a static MRI or CT scan, but with a dynamic, noisy video stream. In the operating field, instruments shake, fluids obstruct the view, and tissue deforms.
The algorithm performs several tasks at once: it identifies the stone and laser fiber, classifies the targets, and determines exactly where the laser beam is pointing. The figures reported by the developers are promising: the system achieved 91% object-recognition accuracy in testing and correctly identified the laser’s direction in 92 out of 100 cases. The team also optimized the time needed to analyze a full surgical recording, cutting it from 102 seconds to 21 seconds.
For now, this remains a prototype, not a finished medical AI service. The neural network neither controls surgical instruments nor makes decisions on behalf of the physician. Its role is limited to visual assistance and analytics.
From Reviewing Procedures to Real-Time Guidance
At this stage, the main benefit of the technology is that it helps medical teams review complex episodes, assess the quality of surgical maneuvers and train early-career clinicians.
But the developers have more ambitious plans. One potential next step is to integrate the software module with an endoscopic monitor. Under the proposed design, the neural network would overlay the live image with a graphic marker highlighting the stone and a vector showing the laser’s direction. It would also alert the surgeon if the beam strayed into a potentially dangerous area.
Turning that concept into reality will require rigorous clinical testing. The algorithm must perform perfectly in the presence of bleeding, poor lighting and camera movement, without adding to the surgeon’s cognitive workload through false alarms.
Russia’s Effort and the Global Trend
MIREA’s project reflects a broader global trend at the intersection of IT and medicine. Russia has already seen its domestic AI assistant ArtInCol for colonoscopy progress from an initial concept to registration as a medical device by Roszdravnadzor, the country’s medical-device regulator, in 2024. Augmented-reality systems such as Medgital Vision are also being tested by neurosurgeons.
Internationally, the field is moving beyond simple algorithms toward more comprehensive solutions. Overseas counterparts, including the AiFURS system, are already undergoing external clinical evaluation. Meanwhile, regulators around the world, including the U.S. Food and Drug Administration (FDA), are establishing stringent requirements. Accuracy on a test dataset is only the first hurdle. What matters most is independent clinical validation and evidence that AI actually reduces the rate of complications.
The Long Road From Lab to Operating Room
The KulakMirea prototype represents a significant step forward for Russian medical computer vision and the development of specialized software. It demonstrates that Russian teams can build sophisticated algorithms for dynamic data in settings where safety requirements are exceptionally high.
Time will tell whether this system becomes a standard interface in clinics in Russia and abroad. Major clinical evaluations using independent datasets, adaptation to different equipment and the lengthy process of medical-device registration still lie ahead. Its export potential will also depend on demonstrating clinical benefits. But the direction is clear: the future of surgery is not about replacing physicians with machines. It is about a partnership in which human experience and intuition are enhanced by tireless digital vision.
