The Study Of EEG Denoising Algorithm Based On Simultaneous EEG-fMRI Measure

With the progress of science and technology,simultaneous recording of electroencephalogram(EEG)-functional magnetic resonance imaging(fMRI)plays an important role in scientific research and clinical field.The time accuracy of EEG is high,and the spatial accuracy is low,while functional magnetic resonance is exactly the opposite.The combination of them makes up for their respective shortcomings.However,EEG is easily influenced by gradient artifcats(GA)and ballistocardiogram(BCG)artifacts under complex MRI environment.Gradient artifacts is caused by switching of magnetic field.Ballistocardiogram artifacts is the result of the heartbeat related movement of subjects due to static magnetic field.The existence of two kinds of artifacts seriously affected the follow-up fusion study.In this paper,catch the time-varying characteristics of ECG signals,two novel method clustering-constrained ICA(ccICA)and real-time constrained ICA(rt-cICA)based on constrained ICA are proposed to remove the BCG artifacts;The constrained Independent Vector Analysis(cIVA)method based on the Independent Vector Analysis(IVA)is proposed to remove the gradient artifacts.Three algorithms proposed in this paper are used to remove the artifacts in the simulated data and the real EEG-fMRI data.The results show that:(1)For simulated data analysis in gradient artifacts removal,the value of error in signal amplitude(Er)obtained by cIVA method was lower than the Average Artifact Subtraction(AAS)and IVA(P<0.005).In real EEG data analysis,the computational cost of cIVA algorithm is significantly lower than that of AAS and IVA and the Improvement of Normalized Power Spectrum(INPS)calculated by cIVA method was higher than other methods(P<0.005);(2)For simulated data analysis In ballistocardiogram artifacts removal,the value of Er obtained by ccICA and rt-cICA method are lower than AAS,Optimal Basis Sets(OBS)and cICA(P<0.005).In real EEG data analysis,the Improvement of Normalized Power Spectrum(INPS)calculated by ccICA and rt-cICA method are higher than other methods(P<0.005);(3)In addition,the results of the spectrum diagram,the proportion of residual artifacts and event related potential(ERP)extraction with its signal-to-noise ratio(SNR)are shown that ccICA and rt-cICA algorithm are superior to other algorithms.In conclusion,three artifacts removal algorithms were proposed in this paper and used in the simulated data and real data.The experiment results are better than the traditional methods.The novel method proposed by this paper lays the earlier stage technical foundation for further research on the fusion model of EEG-fMRI.