Analýza EEG metodou ICA

Abstract

Each EEG recording consists of not only brain activity of the patient but also different types of artifacts superimposed to the measured signal during the measurement. Some of the artifacts can be eliminated by filtration, but it is necessary to use fine filters. If we choose the wrong filtration, we can remove some of the pathological graphoelements, such as high altitude waves of epileptic activity. Therefore, we can use independent component analysis, which finds the most important components from recorded brain activity; these components are further processed and evaluated. The main principal of this method is separation of measured data into independent components which are non-Gaussian. After removal of certain components that contain artifacts we are able to reconstruct a new EEG recording without these interfering components. My thesis is mainly focused on spectral analysis of the data processed by this method and on detection of epileptiform activity using the tests of extreme values occurring in records.