Use of Independent Component Analysis in the extraction of Artifacts of the Mismatch Negativity Response

Abstract

Mismatch Negativity is a cortical response elicited by the automatic detection of stimuli which have different characteristics, allowing exploration of neuropsychological processes. However, the analysis of this signal can be difficult by a low SNR due to artifacts present when the signal is recorded. Different publications propose to use the approach given by the Blind Source Separation Technique by means of the Independent Component Analysis (ICA)to preprocess and eliminate these artifacts. Nevertheless, it has not been studied which of the ICA algorithms found in the literature will be optimal for improving the quality of MMN. Therefore the aim of this study is to determinewhether there are significant differences in the responses obtained by using FastICA, Infomax and SOBI to remove artifacts typically present in such signals. In addition, some features of the Independent Components related toartifacts are given in order to systematize the identification and elimination of those. In addition, MMN responses obtained with and without data preprocessing, as well as topographic maps before and after the elimination of artifacts were compared. Thus, Infomax is the best ICA algorithm to calculate Independent Components associated with artifacts, resulting in high amplitude MMN and a topographic map with a clear fronto-central distribution with left-hemisphere predominance.