Project Description

Focal cortical dysplasias (FCDs) are examples of surgically remediable drug-resistant epilepsy in children. Other locations of epileptogenic foci are possible. Surgical resection can result in reduced need for anti-epileptic medication, reduced frequency or most commonly complete absence of seizures. There is evidence too that it can even improve developmental outcome. The challenge in many cases is to accurately locate the area of responsible tissue.

Surgical outcome is significantly improved when lesions are identified on MRI scans pre-surgically. However between 50 and 80% of FCDs are too subtle to detect by conventional radiological analysis of MRI scans. Therefore an automated tool capable of improving the detection of FCD and other possible foci in the paediatric and adult population would represent an important step in improving the quality and consistency of presurgical evaluation with implications for surgical outcome [1,2].

The purpose of current research project (performed with medical partner - V.I. Kulakov Research Center for Obstetrics, Gynecology and Perinatology) is to build a highly automated and precise deep learning system for potentially epileptogenic foci localization built on multimodal MRI data (T1, T2, FLAIR, DTI modalities data fusion).


  • Check the applicability of already developed epilepsy diagnostic algorithm on new provided data: extract morphometric features from structural MR images, teach the classifier and etc.
  • 2D/3D image segmentation based on available labelled and segmented data, here conventional ML as well as deep learning algorithms could be applied
  • “Healthy” and “Ill” brain pattern creation based on GANs and available labelled data


  1. Hong S. et al. Multimodal MRI profiling of focal cortical dysplasia type II // Neurology. 2017. Vol. 88. P. 734–742
  2. Adler S. et al. Novel surface features for automated detection of focal cortical dysplasias in paediatric epilepsy // NeuroImage Clin. The Authors, 2017. Vol. 14. P. 18–27.