DTI Denoising by Directional Filtering of k-space Inspired by the Pattern of Gradient Influence

Motivation: DTI suffers from intrinsic low SNR compared to conventional MRI, making denoising crucial. Goal(s): Our study aims to introduce a novel filtration method based on an estimated pattern of areas most directionally affected in the spatial frequency domain. Approach: A pattern was suggested through DTI theory and observations of simulation data. This was then used to propose a filter for noise reduction. The application of this filter was quantitatively and qualitatively evaluated on simulated MR and DTI images, considering added noise. Results: The results showed not only the method’s increased robustness to noise but also a clearer representation of white matter tracts. Impact: By integrating the fundamental concepts of diffusion within the white matter into our filter design, we present a promising approach to denoising DTI data, potentially yielding more reliable and biologically meaningful results, and benefiting researchers investigating brain connectivity.