Functional magnetic resonance imaging based on Wiener filter over wavelet domain

Magnetic resonance images acquired with temporal high resolution often exhibit large noise artifacts, which could be generated by inspected objects or by instruments. The distribution of temporal noises is uncertain in most cases. Many existent de-noising techniques assume that the noises have uncorrelated Gaussian distributions. This paper discusses de-noising Wiener filter incorporating wavelet domain and applies it for functional magnetic resonance imaging (fMRI) data analysis. We incorporate multi-resolution analysis based on wavelet domains. We transform the uncertain distributed noises into the space with a series of orthonormal basis functions using different spatial scales, and analyze data using uncorrelated Gaussian distribution in the subspace of each spatial scale. The fMRI data can be de-noised in different spatial scales using wavelet weakening algorithm. In each spatial scale, we use the robust median method to approximate the signals. The experiment results prove that our de-noising method is effective.