AI Approach Improves Brain MRI Lesion Detection in Epilepsy and Multiple Sclerosis

Drug-resistant epilepsy can require precise lesion localization before surgery, yet some affected patients have brain MRI scans that appear negative to physicians. Multiple sclerosis also depends on MRI monitoring because new disease foci can guide treatment planning. Small or atypical abnormalities can be missed by existing computer vision models, limiting their usefulness in diagnostic radiology. Researchers have now developed an AI approach that improves detection of subtle signs of epilepsy and multiple sclerosis on brain MRI.

The technology, BioGuide, was developed by researchers from the joint Biomedically-Informed AI Laboratory of Skoltech and the University of Sharjah, with colleagues. It is designed to improve automated medical image analysis by making neural networks attend to biomedical features recognized by physicians. The approach also uses the model’s ability to detect subtle image details that may not be visible to the human eye.

BioGuide supplements standard segmentation with an auxiliary classification procedure. Instead of only separating healthy tissue from pathological tissue, the model also categorizes suspected abnormal regions by features such as brain location and lesion boundary characteristics. This added classification step directs the system toward clinically meaningful features and improves segmentation quality.

The approach was tested on clinical data for both diseases. Pathological area detection increased by 2% for MRI scans showing signs of multiple sclerosis. For scans showing signs of epilepsy, performance rose by 8 percentage points, from 72% to 80%, and the model matched prior performance with 30% less training data.

The study, titled “BioGuide: Biomedically-Guided Segmentation for Medical Tasks,” was published in the proceedings of the 29th International Conference on Medical Image Computing and Computer Assisted Intervention, held in Strasbourg, France, from Sept. 27 to Oct. 1, 2026. The findings point toward more interpretable computer vision outputs for radiology. Future versions could accompany detection results with explanatory comments for physicians.

“By integrating biological features into the neural network, we make it work more in line with the actual reasoning of physicians. Further development of this approach will make the results of computer vision analysis more interpretable. The model will justify its conclusions in a comment, making the findings more trustworthy in the eyes of a physician,” said Nadezhda Alsahanova, study lead author, research engineer at Skoltech, and Ph.D. student in the institute's Computational and Data Science and Engineering program.

“Searching for small and atypical pathologies on MRIs is a critically important task, since missing them can cost patients the right treatment. What makes this study's approach original is that when detecting a pathology, the model begins to rely on biomedical features, much like a human physician. Consequently, the output becomes interpretable and more suitable for real-world diagnostics. Importantly, Nadezhda, who came up with this idea and brought it all the way to MICCAI level, managed to confirm the findings on clinical data,” said assistant professor Maxim Sharaev, co-director of the joint Skoltech-University of Sharjah BIMAI-Lab and principal investigator in the study.

Related Links
MICCAI
Skoltech 
University of Sharjah