Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Characterization of epileptic activity based on integrated functional network from MEG data on a realistic head model

Dimakopoulos Vasileios

Full record

---

Summary

Epilepsy is a complex brain disorder which affects millions of people worldwide. A significant percentage of the cases is described by drug resistance increasing this way the need for exploiting different approaches for the treatment. Such techniques incorporate invasive methods to resect the cortical tissues that are responsible for the epileptic seizures. However, these approaches require spatial accuracy in order to localize the epileptic focus as well as to avoid damaging the equivalent brain areas. An important step prior to the operative treatment is the presurgical evaluation which aims in accurate detection of the epileptic focus utilizing the Electrocorticography (ECoG). This thesis addresses the source localization/reconstruction problem from interictal epileptic spikes to improve presurgical epilepsy diagnosis with the ultimate goal to save patients from multiple repetitions of these invasive techniques. Specifically, it aims to detect the epileptic activity at source level as derived from combined electroencephalography (EEG) and magnetoencephalography (MEG) Data on a realistic head model. The source reconstruction problem is focusing on disentangling the brain sources from the activity recorded non-invasively by the sensors of the neuroimaging modalities by simulating brain anatomy and conductivities. The proposed approach includes unsupervised learning methods to sort epileptic activity using adaptive features for the spikes and comparison of algorithms such as sLORETA, eLORETA and Minimum Norm Estimate (MNE) for solving the localization problem. In the clustering model of our method we consider the problem of describing the interictal spikes with adequate features, which could be used for sorting purposes. In the source reconstruction, we solve the forward problem using a 6-compartment head model constructed with Finite Element Method (FEM). The inverse solution of the problem is being performed mainly with sLORETA algorithm but MNE and other inverse methods were also evaluated utilizing the FEM headmodel. Finally, the results obtained achieve exceptional accuracy in detecting the epileptic foci in a patient with multifocal epilepsy with the activated areas being in the vicinity of patient’s focal cortical dysplasias.

Available Files

Services