Compressed sensing for accelerated microstructure imaging

Magnetic Resonance Imaging (MRI) interpretation is primarily based on qualitative visual assessment of differences in image contrast. New developments in quantitative MRI (qMRI) offer the possibility for objective tissue characterization by leveraging the rich multi-dimensional signal of images acquired with different parameter settings to estimate fundamental underlying tissue properties using a generative model. Among these qMRI techniques, microstructure imaging focuses on probing features of the tissue microstructure using diffusion-weighted MRI (dMRI). In the brain, dMRI can probe characteristics of neural tissue on a sub-voxel scale and trace neuronal connectivity pathways. High-precision microstructure imaging with dMRI is currently not feasible in clinically acceptable scan times due to the high dimensionality of the dMRI measurements that must be acquired. The aim of this PhD research is to accelerate qMRI of neural tissue microstructure by an order of magnitude by exploiting a previously developed and validated rank-1 representation of the dMRI signal. This representation could impose a very effective constraint on dMRI reconstruction techniques for strongly undersampled data, akin to compressed sensing methods, reducing the number of parameter combinations that need to be sampled and hence the scan time needed for robust microstructure modelling.