Posterior probability estimation for actual and artifactual components from MEG data

The presence of physiological artifacts from magnetoencephalogram (MEG) data, e.g., eye movements, muscular contractions, cardiac signals, sudden high-amplitude changes, and environmental noise reduce the correctness of interpretation. Therefore, the automatic artifact removal is needed. In this paper, we present a posterior probabilities of actual and artifactual components in order to determine optimal threshold values for each parameter. The results showed that the actual and artifactual MEG components were classified clearly by using optimal threshold values of 1.352, 0.017, 0.443, 0.949, and 0.963 for kurtosis (K), probability density (PD), central moment of frequency (CMoF), spectral entropy (SpecEn), and fractal dimension (FD), respectively.