An Efficient Hybrid Wavelet-ICA algorithm for Analyzing Simulated fMRI Data in Noisy Environment

The performance of ICA algorithms in correct separation of independent sources can be highly affected by existence of noises in the observation data. In this paper a hybrid Wavelet-ICA method for improving the functionality of noise free ICA algorithms in noisy environment is proposed. At first the robustness of two most frequent ICA algorithms, named Fast ICA and Information maximization ICA, for extracting true activated spatial and temporal sources off MRI signals in the presence of different noise levels are evaluated These algorithms are applied on simulated fMRI datasets consisting of different activated sources with various temporal patterns, different levels of activation, trend and noise. Then, a hybrid wavelet-Fast ICA model to transform the signals into a domain, allowing for simultaneous un-mixing and wavelet based de-noising is proposed. As the results show this combination has significantly improved the sensitivity of extracted sources in different SNR levels, in particular in low SNR's. To measure the accuracy of source separation, the correlation coefficients between extracted activation signals and simulated temporal patterns are also measured. As the results suggest the proposed hybrid method is more robust in comparison with noise free ICA for noisy observation in extracting more accurate independent sources.