Development Of Spatial ERP Component Extraction Method And Its Multisensory ERP Analysis Application

For cognitive neurophysiology, one major question of interest is how and where the brain information processing process takes place. So far, existing brain function evaluation techniques includes electroencephalogram (EEC),magnetoencephalogram (MEG), positron emission tomography (PET), functional magnetic resonance imaging (fMRI) etc. Among them, the technique using event-related potential (ERP) contained in EEG features high temporal resolution, low costs and convenient application. It's undoubtedly a very helpful technique for exploring the ever-changing human brain and its cognitive mechanism. ERP reflects the electrical activities happened during the process of outside stimulus (event) inputting into the brain through the primary sensory receptor, transmitted to the primary cortex via specific neural pathway, then integrated in associated cortex. Thus, ERP is a summation of several electro-physiological subprocesses (components), which can be mainly divided into two categories, stimulus-related exogenous potential and psychological process related endogenous potential. Taking advantage of statistical analysis, spatiotemporal overlapped ERP components can be extracted.In ERP researches, compared with once prevailed unisensory ERP research, multisensory ERP research is much more meaningful to the exploration of human cognitive mechanism since it's closer to real life. Previous research findings of multisensory ERP indicate that, there is divergence between auditory and visual stimuli simultaneously functioned ERP ( 'simultaneous' ERP) and the sum of auditory and visual stimuli respectively functioned ERP('sum' ERP). And it can be speculated that, this extra component could reveal the trail of human brain's work in multisensory integration and cognition. So if this kind of component could be extracted and its spatiotemporal topography be made, it would facilitate the data analysis of multisensory ERP greatly.Conventionally, due to the huge volume of data, the analysis of ERP data, viewed as a multivariate problem, usually adopts principle component analysis (PCA) to reduce data dimension. PCA can project complicated related multi-dimensional data into an orthogonal system to extract independent components, and discard unimportant components to reduce data dimension. However, PCA has a main defect that its analysis result component has neither anatomical nor physiologicalsignificance. Existing temporal component extraction method uses multiple regression analysis (MRA) after the application of temporal PCA to estimate the amplitude of resultant components and thus made it up. In spite of its great advance, this method is a one-dimensional method working on single electrode site. It didn't take the correlativity between electrodes into account. In this case, spatial topography could only be pictured through manual combination of electrodes, but not topography of statistical spatial component. With the increasing number of electrodes being used, this kind of manual combination will become practically infeasible. To solve this problem, it's necessary to develop a spatial component extraction method so that the spatial component topography mapping sequence along the time course could be easily got into hand, even in a dense electrode array study.This study is backed by the research project of stimulus onset asynchrony (SOA) detection, a newly devised multisensory ERP experimental method. On the basis of spatiotemporal PCA and existing temporal component extraction method, a spatial component extraction method was designed and materialized. Detailed algorithm description, software development introduction and application example discussion were all presented. As demonstrated in the example, this method proved to be effective and useful.