Temporal And Spatial Group-level Analysis For Event-related Potentials

Event-related potential(ERP)is an important tool in cognitive neuroscience research.When data is decomposed in group-level,two algorithms are usually used: the algorithm of principal component analysis followed by Promax rotation in temporal domain(the channel data of each subject are concataneted together in spatial domain,namely,temporal PCA plus Promax)and the algorithm of information-maximisation based on independent component analysis in the spatial domain(the waveforms of each subject are concataneted together in the time domain,namely,spatial Infomax ICA).Currently,four problems exist in these methods: ⑴ In previous ERP simulations,the datasets were far away from ERP in temporal-spatial domain although they met the requirement of the basic assumptions;⑵ The indeterminacy of components’ variance and polarity in temporal PCA plus Promax prevented the comparison of different components;⑶ The validity of wavelet filter in group-level analysis has rarely been studied;⑷ The order of difference wave(DW)and data decomposition in group-level lack criteria of assessing its rationality.In this study,the simulated data to validate mentioned methods proposed by Dipole Simulator(V3.3.0.2)matches the ERP data’s scalp maps and waveforms;For temporal PCA plus Promax,we applied back-projection to correct the indeterrminacy of components’ variance and polarity,allowing the comparions of different components in the electrodes’ field;Meanwhile,the significance of real ERP data was further promoted by wavelet filter;For spatial Infomax ICA,the stability combined with the effective temporal-spatial distribution of components were proposed as the assessment standard of effectiveness of wavelet filter and the order of DW decomposition.The above solution was also applied to study the real ERP datasets of an oddball experiment.The optimal results were obtained when the wavelet filter was combined with the temporal PCA plus Promax decomposition in group-level;However,the currently used wavelet filter didn’t facilitate the spatial Infomax ICA for group-level decomposition;When DW data were decomposed in group-level directly,temporal PCA plus Promax didn’t produce the significant results and spatial Infomax ICA failed to extract reasonable ERP waveforms.