Research On Ballistocardiogram Artifact Removal

Brain science is one of the frontier in life science in the 21 st century. The ultimate goal of brain science is to clarify the working principle and mechanism of the brain and nervous system. It is increasingly recognized that the scientific significance and the role in promoting the development of human society of brain science. In recent years, the application of non-invasive neuroimaging techniques has gradually become an important trend in the development of brain science. Electroencephalogram(EEG) and functional magnetic resonance imaging(fMRI) are able to observe and record the whole brain activity noninvasively. EEG and fMRI are two types of the most important neuroimaging techniques which are used in brain function research. The simultaneous EEG–fMRI has gained attention due to enabling study of brain functions with high temporal and spatial resolution.Despite its numerous benefits, the major drawback of simultaneous EEG and fMRI recording is the compromised EEG data quality, which is contaminated by two main sources of artifacts, i.e. gradient artifact(GA) and ballistocardiogram(BCG) artifact. The characteristics of EEG signal recorded inside the MR scanner are low signal to noise ratio, high complexity, high randomness and nonlinear mixed. It is the most difficult task to suppress the ballistocardiogram artifact effectively. Therefore, a new method based on adaptive comb filter combined with blind source separation using time-frequency distributions(TFBSS) was proposed in this paper. Firstly, the location of J peak of BCG artifact in EEG signal recorded inside the MR scanner is estimated using the ECG recorded simultaneously as reference signal. Then the BCG artifact is preliminarily suppressed utilizing adaptive comb filter with adaptively adjusting the parameters of comb filter according to the location of J peak. At last, the clearer EEG is obtained by removing BCG artifact based on TFBSS.The proposed method is validated by the experiments on both synthetic and real EEG signal recorded inside the MR scanner. The visual results, the peak-to-peak value of artifact-related epochs and the improved normalized power spectrum ratio(INPS) are used to evaluate the performance of the BCG artifact removal method. The experimental results indicated that the proposed method is able to suppress the BCG artifact effectively and better than the principal component analysis(PCA) method and the fast independent component analysis(FastICA) method.