Artifactual component classification from MEG data using support vector machine

Recently, an independent component analysis (ICA) has been proven to be an effective method for removing artifacts and noise in multi-channel physiological measures. ICA can extract independent component (IC) which was directly regarded as artifacts. In this paper, we propose an automatic method for classifying physiological artifacts from magnetoencephalogram (MEG) data. The artifactual ICs were classified based on support vector machine (SVM) algorithm. The following parameters: kurtosis (K), probability density (PD), central moment of frequency (CMoF), spectral entropy (SpecEn), and fractal dimension (FD) were used as input vector of SVM. The proposed method showed the average classification rates of 99.18%, 92.33%, and 98.15% for cardiac (EKG), ocular (EOG), and high-amplitude changes (HAM), respectively.