Trend Analysis for Brain Source Localization Techniques Using EEG Signals

Abstract

Brain source localization is a growing field of research in neuroscience as it has diversified applications for diagnosing of various brain disorders. It encompasses 2segments: forward problem and inverse problem. Various numerical techniques such as finite element method (FEM) and Boundary element method (BEM) are used for head modelling to solve forward problem. However, the inverse problem is evaluated utilizing optimization techniques which include minimum norm estimation (MNE), low resolution brain electromagnetic tomography (LORETA) and Bayesian framework based multiple sparse priors (MSP). This research work uses EEG signals for localizing active sources of brain. EEG data is generated at SNR value of -5dB synthetically. Thus, BEM is used for forward modelling and classical MNE, LORETA and MSP are used for inverse problem. However, novel concept of increasing number of patches to improve localization accuracy is implemented within Bayesian framework. Hence, this technique is termed as modified MSP (M-MSP). Thus, the data is subjected to multiple trials to validate results statistically. The trends are plotted between various parameters of localization. According to results, MMSP has improved accuracy related to free energy and localization error.