Non-invasive EEG Functional Neuroimaging for Localizing Epileptic Brain Activity
Doctoral thesis, 2013
Epilepsy is one of the most common neurologic diseases in the world, and is present in up to 1% of the world’s population. Many patients with epilepsy never receive the treatment which make them seizure free. Surgical therapy has become an important therapeutic alternative for patients with drug resistant epilepsy. Correct and anatomically precise localization of the epileptic focus, preferably with non-invasive methods, is the main goal of the pre-surgical epilepsy diagnosis to decide if resection of brain tissue could be a successful treatment option. The most important diagnosis tool used at epilepsy surgery centers is electroencephalography (EEG), which is used to find the source of activities inside the brain by measuring the potential on the scalp with EEG electrodes at different locations. One major advantage of EEG source localization over other brain imaging modalities is its high temporal
resolution. The procedure of EEG source localization deals with solving the forward problem to find the scalp potentials for a given current dipole(s) inside the brain and the inverse problem to estimate the source(s) that fits with the given potential distribution at the scalp electrodes. Realistic models of the human head are geometrically complex and the tissue conductivity is inhomogeneous as well as anisotropic. A critical issue for the forward problem is how to handle the computational complexity in the numerical approaches with regard to the inverse problem. There is still a lack of sufficiently
powerful methods and algorithms that would satisfy the time-restrictions for the solution of the inverse problem. The overall goal in this thesis is to develop a non-invasive, clinically-viable, time-efficient method for localization of epileptic brain activity based on EEG source localization. For the forward problem two methods are proposed for modeling the dipole source which can handle the head model complexity; a modified subtraction method and a method based on the reciprocity theorem. For the inverse problem we propose a new global optimization method based on particle swarm optimization (PSO) to solve the multi-dipole EEG source localization. The techniques of multimodal magnetic resonance imaging (MRI) are used in order to generate a high-resolution realistically shaped volume conductor model. The anisotropic white matter conductivity tensor is determined by diffusion tensor
MRI (DT-MRI) measurements and isotropic conductivities are assigned to the other tissues in the model. The new proposed methods are tested for synthetic and real EEG data. The results are compared with state-of-the-art and other existing methods. In the synthetic data both spherical
head models and realistic head models with anisotropic tissues are used for validation. In the real EEG test, measured somatosensory evoked potentials (SEPs) for a healthy subject are used for EEG source localization. A realistic 1 mm patient-specific, anisotropic finite element model of the subject’s head, with special consideration of precise modeling the two compartments, skull and cerebrospinal fluid (CSF), generated from T1-weighted MRI data is used. Source localization results are validated
against a clinical expert source localization as well as functional MRI palm-brushing measurements and the proposed method typically finds the source location within 10 millisecond. The EEG source localization results agree well with both the clinical expert and fMRI results. The finite element method (FEM) in combination with the reciprocity theorem and the modified PSO is a highly efficient
and robust solution methodology for EEG source localization.
Somatosensory Evoked Potentials
Forward Problem
Inverse Problem
Mathematical Dipole
Epileptic Spike
Functional Neuroimaging
Structural Magnetic Resonance Imaging
Particle Swarm Optimization
Source localization
Finite Element Method
Non-invasive EEG