Novel approaches to characterize neural responses to standard and non-standard sounds in humans: A single-trial MEG/EEG study of the auditory mismatch negativity.
Molecular Biology and Physiology of Neurons and Glial Cells
Final Report Abstract
The auditory mismatch negativity (MMN) is the result of a simple mathematical operation performed off-line, namely the subtraction of the trial-averaged responses to standard stimuli from those to deviant stimuli. In order to be able to address some fundamental methodological aspects concerning the MMN, we acquired MEG and EEG data from three different experimental paradigms, a classical MMN paradigm and two modifications of it. The data constituted the input to our methodological work which we pursued in three different directions. First, and most fundamentally, we emphasize the importance of a stable variance as an essential prerequisite for computing and comparing grand mean waveforms. Such operations are common practice in MEG and EEG studies. In our publication on that topic, we argue that variance stabilization by means of the asinhtransformation is an essential step to put the findings of statistical analysis and the interpretation of the comparison process on firm ground. Second, we developed a novel two-stage algorithm to disentangle the generators of the N100(m) and the MMN(m) components, and successfully tested the performance of this algorithm which is clearly superior to standard algorithms, on simulated data. Third, we approached the single-trial source localization problem of responses to standard and deviant tones by using several different variants of the multivariate matching pursuit adaptive approximation as a preprocessing stage to achieve a noise-free input to the inverse problem. This latter topic remains the most challenging one among those proposed in our application.
