Russian Researchers Develop Platform for Generating Synthetic MRI Images
The project is being developed by master’s students at Samara Polytechnic University.
Master’s students Ivan Belov and Yegor Bekish from the Institute of Engineering, Economics, and Humanities at Samara Polytechnic University have created an AI-based software and research platform called DataSculpt.
The platform generates and analyzes synthetic MRI brain images and is designed for specialists in medical information technologies. It also enables the training of neural network models. The project is supervised by Andrey Pensky, a senior lecturer at the Department of Higher Mathematics. The development was reported by NIA Samara, citing the university’s press service.
A Critical Constraint
Why does this matter? Modern AI developments in medicine face a major constraint: legal and ethical regulations restrict access to MRI data for training and validating diagnostic systems. Real-world data are often unavailable, while standard generative models, although visually convincing, lack clinical accuracy. At present, there is no tool that combines precise labeling, ethical safety, and the ability to define the type, location, and severity of pathologies.
That gap is now being addressed. The Samara Polytechnic team has developed a software prototype based on generative neural networks. The models are trained on real medical images and then generate new synthetic MRI scans. Specialists from Samara State Medical University provided the students with data for the project — specifically, sequential two-dimensional brain slices.
Addressing Data Scarcity and Limited Access
Foreign analogues of such platforms are either closed or narrowly specialized. By contrast, the Samara-developed solution combines image generation with automatic quality control and structural consistency checks. Users can manage the generation process themselves, obtain ready-to-use datasets, and apply them to train and test other AI models for medical image analysis.
The team is currently refining the platform and aims to develop it into a full-fledged software product and secure a patent.
Earlier, we reported that Sevastopol State University is developing a neural network system designed to help diagnose rare types of cancer, including adrenal cancer.
