Artifact reduction for EEG/fMRI recording part 1: nonlinear reduction of ballistocardiogram artifact

Objective: We present a new method to effectively remove ballistocardiogram artifacts (BAs) of electroencephalography (EEG), recorded inside a 1.5 T static magnetic field scanner, and conserve the time and frequency features of EEG activity. Methods: The BAs are approximated as deterministically chaotic dynamics. A wavelet-based nonlinear noise reduction (WNNR) method consisting of wavelet transform, nonlinear noise reduction and spatial average subtraction, is developed to effectively reduce the BAs to be smaller than the EEG activity. Results: The effectiveness of the WNNR method to conserve the temporal EEG signals is evaluated by simulations and real human experiments inside a 1.5 T static magnetic field, both visual evoked EEG dynamics and alpha waves during subject’s eyes closed. Significance: The current method has ability to conserve the time-frequency features of EEG activity, specifically event-related EEG dynamics. Moreover, it might effectively work in the higher field strength as well.