REMOVING ARTIFACTS FROM EEG RECORDED WITHIN MR SCANNER BY DYNAMICAL TEMPLATE STATE SPACE APPROACH

摘要

We propose a novel state space modelling approach toremoving scanner-related artifacts from electroencephalogramsrecorded inside MR scanners. For this purpose,dynamical templates for the actual brain activity and theballistocardiogram are obtained from a short piece of datarecorded without fMRI scanning; dynamical templates forthe scanner artifacts are obtained from data recorded duringfMRI scanning. Finally the two sets of dynamical templatesare merged. We compare our approach with IndependentComponent Analysis and find superior performance.