ABSTRACT
Objective To design a fast method based on wavelet analysis for fMRI data. Methods Lifting wavelet decomposition instead of stationary wavelet decomposition was utilized to separate paradigm responsive signal and confound ones in fMRI data, while frequency analysis was used to find out the wavelet scales in which paradigm responsive signal existed, then reconstructed signal from these scales was subjected to correlation analysis for actived pixels. Results Analyzing visual fMRI data revealed that when the significant level was α＜10-6, the proposed method gave more sensitive results than correlation analysis, but process time decreased on a large scale compared with the one based on the stationary wavelet transform. At the mean time, the proposed method only used 24 timepoints of data for wavelet reconstruction while one based on stationary wavelet transform used 256 timepoints of data. Conclusion The proposed method is the fast one based on wavelet transform for analyzing fMRI data, which also gives an effective technique for compressing fMRI data.