Skip to Main content Skip to Navigation

Development of Differential Connectivity Graph for Characterization of Brain Regions Involved in Epilepsy

Abstract : Drug-resistant epileptic patients suffering from focal epilepsy are recommended for epilepsy surgery. The aim of this surgery is to remove the seizure onset zones (SOZ) without creating new neurological deficits. To identify the SOZs, the best way is to record the seizures from intracerebral electroencephalogram (iEEG) recordings. However recording seizures is complicated contrary to the recording of interictal epileptiform discharges (IED). Therefore prediction of SOZ by estimating the IED regions is very valuable. There are several studies wondering if the estimation of the IED regions can be useful to predict the SOZ and eventually for presurgery evaluations. Although the encouraging results of these studies, the problem is still an open issue. The main problem of the previous studies is the reliability of the results. The aim of this thesis is to estimate the leading IED (LIED) regions from interictal analysis of iEEG recordings through connectivity graph. The main originality of the proposed method refers to a new reliable graph analysis method called differential connectivity graph (DCG). This graph is designed to identify the significant discriminated connections between IED and non-IED brain states. The statistical reliability of DCG is obtained by using permutation-based multiple testing. In the proposed method, multiple DCGs associated with different frequency bands are constructed. Each DCG includes both source and sink nodes involved in IED events. To identify the source nodes related to LIED regions, the directions of the edges of DCG are estimated and a new measure called local information (LI) is proposed to measure the emittance contribution of each node. To estimate the LIED regions from the LI values related to multiple directed DCGs, a multi-objective optimization method is used. The proposed method is applied on five epileptic patients. These patients underwent resective surgery and they are seizure-free after the surgery. Estimated LIED regions are compared with SOZ detected visually by the epileptologist and SOZ detected by a method using induced ictal iEEG. The comparison reveals congruent results between estimated LIED regions and SOZs. Being all of the patients seizure free and inclusion of estimated LIED regions in the removed regions during surgery shows the reliability of estimated LIED regions for presurgery evaluations.
Complete list of metadatas

Cited literature [227 references]  Display  Hide  Download
Contributor : Ladan Amini <>
Submitted on : Wednesday, April 6, 2011 - 3:13:44 PM
Last modification on : Thursday, November 19, 2020 - 12:59:53 PM
Long-term archiving on: : Thursday, July 7, 2011 - 2:54:50 AM


  • HAL Id : tel-00559915, version 2



Ladan Amini. Development of Differential Connectivity Graph for Characterization of Brain Regions Involved in Epilepsy. Signal and Image processing. Institut National Polytechnique de Grenoble - INPG, 2010. English. ⟨tel-00559915v2⟩



Record views


Files downloads