Automated Detection of Focal Cortical Dysplasia Lesions from MRI Scans of Epilepsy Patients

Abstract

Focal Cortical Dysplasia (FCD) is a predominant cause of medically intractable, partial and focal epilepsy. It is a medical condition that results due to disorganization of cortical development. Certain epilepsy patients have shown resistance to medicines. MRI scans of these patients have also appeared normal. This is because FCD lesions cannot be distinguished clearly from healthy cortex tissue by just looking visually. Even experienced radiologists find it difficult to identify the subtle characteristics with which FCD lesions are displayed in MR images. Unfortunately, because of this reason, many patients do not even receive proper treatment for FCD as doctors declare their MRIs as negative (normal). A computerized approach using the MRI image can reliably detect these lesions. In earlier approaches it was shown that an analysis of MRI data has the potential to detect FCD and hence whether medically intractable partial epilepsy is present or not. In this paper, a novel and hybrid algorithm MCS (a combination of Morphological Operations, weighted K-Means Clustering and Support Vector Machine) has been proposed to detect and analyze the lesional structures in MR images. FCD detection from MRI of partial epilepsy patients has been carried out successfully and the results have shown considerable improvement over the earlier approaches.