A novel method for sampling the active and noisy areas is proposed by using the purification of gray and non-gray matter areas of fMRI data. Also, a data-driven network is proposed in a parallel, multi-step and integrated manner for optimal noise reduction of t-fMRI data.
