Random Segmentation Based Principal Component Analysis to Remove Residual MR Gradient Artifact in the Simultaneous EEG/fMRI: A Preliminary Study

In the electroencephalography (EEG) data simultaneously acquired with the functional magnetic resonance imaging (fMRI) data, the removal of the residual magnetic resonance (MR) gradient artifacts has been a challenging issue. To remove gradient artifacts generated from switching MR gradient field, average artifact subtraction (AAS) has been widely used. After applying the AAS method, however, residual MR gradient artifacts still remained in corrected EEG data. In this study, we proposed a novel method to remove the residual MR gradient artifacts (GAs) using random segmentation based principal component analysis (rsPCA). The performance of rsPCA was compared to that of the independent component analysis (ICA) method using data acquired from a motor imagery task. The results indicated that rsPCA could suppress further the residual MR gradient artifacts remained from the AAS step compared to the ICA method.