Denoising of functional MRI using ICA

This paper proposes a novel approach for elimination of various artifacts and noise from fMRI signals by using Independent Component Analysis (ICA). A comprehensive classification of different components in fMRI is first discribed and their methods of identification based on temporal/spatial characteristics are also discussed. The effect of the denoising scheme was explored both on a fMRI dataset collected from a visual task experiment and a synthetic one, where we applied the fast ICA algorithm for noise removal. The noisy dataset and the denoised one were both processed by a correlation technique to compare their capabilities of activation detection. From this study it can be concluded that ICA technique is possible for restoration of fMR images thus improving the efficacy of detection techniques of activation.