De-Noising Functional Magnetic Resonance Imaging (fMRI) Data Using an Exponential Gradient Filter

Unlike Gaussian noise, Rician noise filtering is more challenging, since this type of noise exists in Functional Magnetic Resonance Imaging (fMRI) data which makes the analysis of fMRI data very difficult for experimental and clinical purposes. To cope with the situation, normally (smoothing) de-noising is done before the analysis of the data using conventional methods like Gaussian filtering as done in Statistical Parametric Mapping (SPM). However results of conventional filtering are not satisfactory. In this work a specially designed filter for fMRI data de-noising is presented. Weights of this filter are calculated by taking the difference of spatial neighboring pixels in each direction of a predefined window. The proposed filter outperforms when applied to synthetic and actual fMRI data. Results of the proposed and state of the art filters are compared using correlation, PSNR and Mean Square Error (MSE). Furthermore, proposed filter can be used not only for fMRI de-noising but also on other problems where Rician noise exists.