ÃÛ¶¹ÊÓƵ

Diffusion magnetic resonance imaging (dMRI) is a powerful non-invasive technique for probing tissuemicrostructure. While highly sensitive, conventional dMRI methods suffer from low specificity, due in part toexperimental designs that conflate multiple contrast mechanisms. This thesis aims to address these limitations bydeveloping advanced time-dependent dMRI techniques to disentangle competing sources of contrast—specificallyrestricted diffusion, intercompartmental exchange, anisotropy, and intra-compartmental kurtosis.In Paper I, we developed a theoretical framework and experimental design using free gradient waveforms, termedRestriction-Exchange (ResEx), to disentangle restricted diffusion from exchange, and validated the approach usingMonte Carlo simulations. In Paper II, we applied ResEx in the healthy human brain on a high-performance scanner(300 mT/m), and isolated distinct exchange- and restriction-driven signal contrasts. Furthermore, in Paper VI, weapplied the ResEx framework in gliomas on a clinical scanner (80 mT/m) and observed substantial and spatiallyheterogeneous exchange-driven contrasts in tumour tissue, with potential implications for non-invasive tumourcharacterisation. Additionally, in Paper III, we used the framework to correct crusher gradient-induced bias in filterexchange imaging (FEXI), enabling more accurate estimation of the apparent exchange rate (AXR).To enhance specificity further, we also developed a tensor-valued encoding approach in Paper IV to enablemeasurement of exchange in the presence of anisotropy and intra-compartmental kurtosis, using variable-mixing-time double diffusion encoding acquisitions.Finally, in Paper V, we studied the interplay between dendritic spine-driven geometric exchange and permeativeexchange in simulations, and used the framework from Paper IV to separate the two, offering a potential route formapping spine density in vivo with dMRI.Together, these contributions establish a set of theoretical and experimental tools to improve the specificity andinterpretability of dMRI. They position advanced diffusion encoding—particularly when combined with ultra-strong gradient systems—as a promising avenue for non-invasive microstructure imaging in both clinical andresearch settings.