Yandex B2B Tech and ShAD will help improve the accuracy of cerebral palsy diagnosis and neonatal therapy

2025: Creation of a neural network for the diagnosis of cerebral palsy and therapy for newborns

Yandex B2B Tech, students of the School of Data Analysis (SHAD) and specialists of the St. Petersburg State Pediatric Medical University have created a neural network that helps doctors assess the development of the brain of babies in the first months of life. The solution can be used as an auxiliary tool for suspected cerebral palsy (cerebral palsy) and other diseases of the central nervous system to choose the best tactics for patient rehabilitation. Now, instead of several days, it takes a specialist in radiation diagnostics a few minutes to decipher the MRI results. The neural network was tested at the St. Petersburg State Pediatric Medical University, specialists are ready to share their work with other medical institutions. Yandex Cloud announced this on August 26, 2025.

𝓲

Фото: Yandex Cloud

The solution is a service deployed on the Yandex Cloud platform. Any doctor can use it for free on a special project page. The specialist of radiation diagnostics loads the results of MRI of the baby into it. The data is anonymized - the patient's name and surname and other confidential information are hidden. The system with an accuracy of over 90% produces an image with delineated contours and a percentage of gray and white matter in the child's brain. In the future, the decision will allow you to assess the development of the brain in dynamics and significantly speed up the receipt of study results in order to decide on therapy. This is particularly important when central nervous system pathology, including cerebral palsy, is suspected. Pathology develops in 2-3 cases per 1000 newborns and is one of the main causes of childhood disability. If the disease can be detected in the first months of life, then the effectiveness of therapy increases, and the patient improves the prognosis for recovery.

The human brain is a complex system that requires attention from the first days of life. There are violations associated with both too slow and too fast its development. Magnetic resonance imaging (MRI) helps to identify such pathologies. But MRI of the brain in infants is a complex and responsible procedure. One such study lasts an average of 20-40 minutes, and it can take from several hours to several days to analyze images and write a medical report even with an experienced specialist. The Yandex neural network will help doctors significantly speed up the diagnosis and choice of therapy for small patients, "said Alexander Pozdnyakov, Head of the Department of Medical Biophysics, Doctor of Medical Sciences, Professor of St. Petersburg State Pediatric Medical University.

We strive to ensure that the latest Yandex developments are as accessible as possible to doctors and help them make accurate and timely diagnoses, provide assistance to patients, choose the optimal treatment methods and develop drugs. Despite the fact that there are many commercial solutions for radiation diagnostics, not a single one has previously solved the task set by the university experts. The main complexity of this project was a limited set of data. Thanks to well-coordinated work with experts, we managed to create a solution that helps doctors examine more patients in the same time and promptly offer therapy where necessary, "said Anna Lemyakina, head of the Yandex Cloud Center for Technology for Society.

Experts trained the neural network on 1,500 impersonal MRI images of university patients and on an open-source dataset provided as part of the MICCAI Grand Challenge, an international competition for segmentation of MRI images of the brain of infants. The BIBSNet (Baby Intensity-Based Segmentation Network) model was used for automatic markup. Two neural networks were used to segment images: ResNet and U-Net. Yandex Cloud offered a solution architecture, tested and configured a web service. Later, the development is planned to be made publicly available so that it can be used in medical and other projects around the world.