Individual Independent Component Analysis on EEG: Event-Related Responses Vs. Difference Wave of Deviant and Standard Responses

Independent component analysis (ICA) is often used to spatially filter event-related potentials (ERPs). When an oddball paradigm is applied to elicit ERPs, difference wave (DW, responses of deviant stimuli minus those of standard ones) is often used to remove the common responses between the deviant and the standard. Thus, DW can be produced first, and then ICA is used to decompose the DW. Or, ICA is performed on responses of the deviant and standard stimuli separately, and then DW is applied on the filtered responses. In this study, we compared the two approaches to analyzing mismatch negativity (MMN). We found that DW introduced noise in the time and space domains, resulting in more difficulty to obtain the spatial properties of MMN by ICA on DW. Thus, we suggest using ICA to spatially filter event-related responses of each stimulus; and then DW is produced by the filtered responses.