Making advanced characterisation of tissue microstructure clinically practical: a data-driven approach to efficient microstructural MRI (360G-Wellcome-215944_Z_19_Z)

Diffusion MRI (dMRI) is the preferred tool for quantifying tissue microstructure, but current technology prohibits comprehensive assessment. Even in the extensively-studied brain white matter (WM), state-of-the-art measurements invalidate commonly-applied biophysical model constraints, and parameter estimates are unreliable. Moreover, models disregard myelin, a key WM-component typically considered ‘off-limits’ with dMRI. We have reached a hiatus in advancing tissue characterisation by dMRI alone, motivating multi-contrast MRI including myelin-sensitive contrasts. Simultaneously varying multiple experimental variables is now possible through ultra-strong gradients. However, as the dimensionality of the accessible MRI acquisition space increases, efficient data acquisition and representation become challenging, hampering the clinical translation of microstructural MRI. I will develop methods to optimise the execution and data representation of multi-contrast MRI experiments, with the goal of comprehensive tissue-characterisation in a clinically-applicable time. I will employ 1) a top-down approach that considers that the tissue properties most important to characterise are known, and assumes that analytical models exist that can be optimised to maximise precision per unit acquisition time; 2) a bottom-up approach that considers that the dimensionality of the analysis-space is unknown. Starting with rich multi-contrast data, I will develop data-driven approaches to characterise the measurement information-content and to select the most relevant measurements.