The Study Of Methology For Eeg Artifacts Removal And Its Applications

The study of brain has been one of the hottest topics in the21st century.Electroencephalography (EEG) acts as a noninvasive detection method in the field ofbrain function study, and the millisecond temporal resolution of EEG determines itsspecial position in brain fucntion study.With the fast development of neuroscience, the single modality EEG cannot satisfythe research requirement, and the combination with fMRI involving high spatialresolution, or with TMS providing casuality relationship has attracts much attention andbeen widely adopted in the research and clinical fields. However, because EEG isusually of low uv amplitude, it is very easy to be seriously contaminated by otherphysiologycal signal or electromagnetic noise due to the ineractions between humanbody and electromagnetism. The noise will influence the following analysis. Therefore,it is very important to remove those artifacts in EEG while keep the useful EEGinformation as possible.Based on the inherent characteristics of noise, this dissertation developed thespecific artifacts removal methods to revover EEG from the strong background noise,which may provide the potential tool for research and clinical application. Thecontribution of this dissertation is as below:1. As for EOG contaminated in EEG, the widely used independent componentanalysis (ICA) method may not be able to completely remove EOG, or it may lose toomuch useful EEG information. By combining ICA and system identification technique(autoRegressive eXogenous, ARX), the possible distortation introduced by ICA can becomposented to result in a relatively SNR EEG estimation.2. As for the Gradient Artifacts induced in EEG-fMRI, we also used ARX as thepost-processing technique to further correct the possible bias introduced by thetraditional MRI artifacts removal approach fMRI artifact slice template removal(FASTR). The preliminary result proves the reliable EEG can be recovered from thestrong MRI artifacts.3. As for the maganetic artifacts in EEG-TMS experiment, the transient magnetic field artifact following TMS stimulus will be observed. The transient magnetic artifactsin TMS experiment are of high consistency in the waveform across stimuli. Based onthis, we develop the two stage methods consisting of template estimation and temporalprincipal component analysis (PCA) to removal TMS artifacts. Compared to theexisting spatial decomposition based approach needing relatively large number of EEGelectrodes, the proposed approach is of high value for the clinical application whichusually uses a small number of EEG electrodes.4. Psychogenic nonepileptic seizures (PNES) are paroxysmal episodes of alteredbehavior, and it is different from the conventional epileptic seizures due to the lack ofboth the expected electroencephalographical epileptic changes and the association todysfunction of central nervous system. Diagnosing PNES remains challenging. Incurrent work, we used the developed EOG artifact removal technique to remove theartifacts existed in the scalp resting EEG, and then PNES is recognized by using thenetwork properties as LDA input, which may provide a feasible tool for clinicaldiagnosis of PNES.