Undersampled Brain Tissue Microstructure Parameter Estimation Based on an Accelerated Soft Threshold Iterative Network

Diffusion Tensor Imaging (DTI) is currently the standard clinical detection method and plays a crucial role in diagnosing brain diseases. However, its interpretation reliability for specific brain tissue structures remains low. Neurite Orientation Dispersion and Density Imaging (NODDI), as a multi-compartment model, provides an alternative explanation. Nevertheless, the Maximum Likelihood Estimation (MLE) method it employs suffers from inefficiency, susceptibility to artifacts and noise interference, and reliance on extensive gradient data. Building on prior research, this paper further addresses the basis pursuit problem using an Accelerated Soft Threshold Iterative Shrinkage Network (AFITSA-net). Our approach maintains fitting accuracy under conditions of high noise and reduced diffusion gradient numbers.