Identification and elimination of cardiac contribution in single-trial magnetoencephalographic signals

A two-step method for identification and elimination of the cardiac contribution in single-trial magnetoencephalographic (MEG) signals is proposed. In the first step, the mean interfering signal (MIS) in one period is estimated by QRS-synchronous averaging of the raw MEG data. In the second step, a QRS-synchronous segmentation of the MEG signals is performed and each signal segment is Gram-Schmidt orthogonalized with the MIS. The above method is applied both to artificial and real MEG data. In each case the heart interference is all but eliminated whereas the components of interest, generated by the brain, remain almost unaffected.