Enhancing stability and speed in multidimensional MRI processing: a kernel-driven signal dictionary with pattern matching approach

Motivation: Diffusion-relaxation multidimensional-MRI (MD-MRI) provides valuable sub-voxel information.However, it suffers from estimation instability and high computational costs. Goal(s): To reduce high computational cost for MD-MRI parameter estimation via a kernel-driven signal dictionary in conjunction with pattern matching, while preserving the quality of parameter maps. Approach: Generate a synthesized signal dictionary using a kernel function, perform pattern matching, and compare the results with state-of-the-art Monte Carlo (MC) inversion. Analyze the results. Results: We propose a novel framework for processing MD-MRI data and estimating diffusion-relaxation parameters. Applied to data from diverse age groups, it enhances parameter maps quality and reduces computational time. Impact: We introduce a novel parameter estimation framework that reduces computational time by up to 12 times while preserving parameter map quality, allowing for efficient processing of large data in clinical settings.