Analysis of the ballistocardiographic artifact removal in simultaneous EEG-fMRI recording using independent component analysis and coherence function

The simultaneous acquisition of EEG and fMRI has becoming increasingly important for neuroimaging research. However, this integration comes with artifacts that directly interfere in the signal quality. Independent component analysis has been one of the most used tools to remove biological artifacts and more recently it has been used to the removal of BCG artifact present when simultaneous EEG and fMRI is acquired. Although widely adopted some questions regarding the reliability of the estimated sources are still unanswered. The efficiency of the removal depends on the correct identification and selection of the artifact-related independent components. This selection is not obvious since different strategies give different results. Considering that no real BCG artifact-free data can be collected inside an MR scanner it is a difficult task to select the best criteria or best algorithm. In this paper we contribute to the solution of this problem proposing an automatic component selection method based on maximum absolute magnitude squared coherence function between the independent components and the ECG signal.