Sechenov University: software for assessing the risk of blood clots

The main articles are:

• Artificial Intelligence in Medicine
• Artificial intelligence in radiology

2025: Presentation of Blood Clot Risk Assessment Software

Young scientists First Moscow State Medical University named after I.M. Sechenov are developing a system based on mathematical modeling and that algorithms artificial intelligence will help doctors assess the risk of thrombus detachment in patients and prevent the development of heart attacks, strokes and other complications. With a help ON capable of detecting thrombus on - CT plots and building 3D models of images, the doctor will be able to predict the risk of thrombus detachment and prescribe a more accurate personalized treatment to the patient. The prototype of the system is planned to be created by the end of 2025. The university announced this on June 30, 2025.

Thrombosis is one of the leading causes of death and disability in people around the world. In Russia, according to various estimates, 1.5 thousand people per 100 thousand population die from cardiovascular diseases directly related to thrombosis. Among the most dangerous complications of thrombosis are heart attacks, strokes and pulmonary embolism. This life-threatening state develops when blood clots, torn off, spread with the current of blood through the body and enter the heart and vessels of the lungs. As a result, a complete or partial closure of the lumen of the pulmonary artery occurs.

This method of assessing the risk of developing these conditions was proposed by young scientists at Sechenov University. They develop software that can not only detect blood clots, but also simulate scenarios in which the risk of their separation increases many times over. The data obtained will enable the doctor to predict under what conditions and when exactly this can happen and prescribe preventive personalized treatment to the patient.

As of June 2025, there are already Russian and foreign IT solutions that are able to detect blood clots on CT scans - patient images. However, none of them can assess the risks of their separation from the vessel wall, because this factor is very difficult to predict, "said Karina Urazova, author and project manager at Thromb.AI, master's student at the Advanced Engineering School of Sechenov University, winner of season 6 of the Sechenov Tech acceleration program. - Our project is aimed at solving this problem. We develop software based on machine and deep learning algorithms, which will not only detect blood clots on CT images, but also design their 3D models. According to these models, neural networks will build calculated grids, calculate the hemodynamics of blood flow, as well as, depending on the shape of the thrombus, its size and a number of other indicators, compile various scenarios for the development of events. And among them find the one that with a high probability can lead to the separation of a blood clot.

The development team includes programmers, specialists in mathematical modeling, machine and deep learning, and clinician physicians. As of June 2025, an initial dataset of about 100 real CT images of venous blood clots of patients has already been created, and an algorithm has been developed that detects a blood clot in the image. For June 2025, the team continues to replenish the dataset with various types of blood clots and is refining the algorithm for calculating hemodynamics. By the end of 2025, it is planned to create a prototype of the system and conduct its pilot testing in clinics. In the future, the functionality of the system will expand and add integration with wearable devices. The finished solution is planned to be developed by 2027. Use software to assess the risk of blood clots will be in public and private clinics. The system can also be used to train students of medical universities and in DPO programs.