Research On EEG-fMRI Hybrid Brain-computer Interface Denoise And Source Localization

Alone with the continuous improvement of research on BCI technology, many scholar pay more attention to the mechanisms of human brain. Although we have had a roughly sketch of brain by the way of anatomy, we can get more information of our brain form several dimensionality now, due to the advance of technology. Such as, now we can use electroencephalography(EEG) and magnetic resonance imaging(MRI) to survey brain. The main work of this article is to set up a system, working in MRI environment, as the base of research on EEG-MRI signal fusion. As a result, we can applying BCI experiment in MRI environment. Anther content of this article is to using EEG signal to localize active brain cell by approach of beamformer.This article begin with the analysis of the reason why EEG signal is contained during fMRI scanning. Then we have an insight into the character of fMRI noise. First, we have hypothesis that noise is approximate same in every slice data block. As to this hypothesis, we can use Imaging Artifact Reduction(IAR) to denoise raw data. The other assumption is that the same Slice data block in every Volume is approximate same. There is a solution, named Artifact Slice Template Removal(ASTR), pointed to this assumption. Because the two assumption above is too strict to exit in normal environment, this article use a new method, Cluster Template Removal(CTR), to denoise. CTR base on the precondition that fMRI noise is not approximate same in the certain slice position, but slice noise is approximate same in uncertain slice position. CTR use cluster to find template for every class. This operation can decrease error during get template. As a result, CTR will get a better performance than IAR and ASTR.The second part of this article, author use Brain Amp MR Series amplifier, made by Brain Products Corporation, and a set of software, written by author, to set up a system which can perform BCI experiment in MRI environment. This system contain signal acquisition and process module, stimulation production module, stimulation appearance module and classification module. During setting up this system, we need notice the different environment between laboratory and MRI. The main difference is signal acquisition and process module should denoise MRI noise in EEG signal. But the denoise algorithm is high-computational, so it result in that software can’t gain data block in amplifier buffer which may cover by next data block. As to this, author optimize software by using multi-thread and making denosie channel selectable. At last, using this system, we perform P300 experiment in MRI environment successful. And using EEG signal recorded during this experiment, we localize active brain cell by beamformer.At the end of this article is use EEG signal to localize active brain cell. The technology we use is beamformer. At begin of this part, author expound beamformer basic principle and then analysis the drawback of conventional beamformer. At last, use simulation to verify this method validity. What’s more, convex optimization is also a main part of this chapter. Author use convex optimization theory to improve conventional beamformer. And then, present simulation result.