Meeting Banner
Abstract #3248

Removing Ballistocardiographic Noise in Combined EEG-fMRI Using Soft Constrained PLS Method

Siamak Salari Sharif1, Michael D. Noseworthy2, James P. Reilly3, Daniel J. Bosnyak4

1MR Innovations Inc., Detroit, MI, United States; 2Electrical and computer engineering, school of biomedical engineering, McMaster University, Hamilton, ON, Canada; 3Electrical and computer engineering, McMaster University, Hamilton, ON, Canada; 4Department of Psychology, Neuroscience and Behaviour, McMaster University, Hamilton, ON, Canada

Here we present a novel approach to removal of ballistocardiographic noise that contaminates EEG during combined EEG/fMRI. By using a soft constrained PLS algorithm (SC-PLS) we were able to extract EEG components that are essentially noise free. Our method is effective at removing BCG especially when ECG data is corrupt or not available. Furthermore, our method is advantageous as it does not require the finding of ECG QRS peaks, nor requires collection of ECG data.

Abstract PDF Presentation Video

Keywords

accommodating addition additional affecting allowing appears approved artifact artifacts available basis better biomedical black blood blue board body brain built calculating channel channels checker cleaning coefficient collected collection combination combined compatible component components computational computer constrained constraint contaminated corr correlation corrupt covariance dashed determine effective eigenvalue eigenvalues eigenvector electrical engineering epoch error especially essential ethics event experiment exposition extract extracted field find finding free furthermore general generated gradient healthy identification incorporates individual induced inherits inside involved iterative knowledge known largest latent linear lobe local major makes many matrix maximization middle missing movement noise noisy occipital often onto optimal original outside overall patient paused peaks penalizes plotted prior problem problematic processing products projecting property proposed psychology quality regularizing related removal removed removing require requires reusing running scanning school selection sets severely sign simple slices soft software squared static step structures subject subspace subspaces system temporal thesis toolbox true type variable volunteers whether