Quantitative Microstructural Imaging for Clinical Use

Today, Magnetic Resonance Imaging (MRI) is a well-established medical imaging modality thanks to its excellent soft tissue contrast. Conventional MR image contrasts can be weighted towards one or more physical properties that discriminate different tissues, which represents one of the main advantages of MRI compared to other imaging techniques. However, conventional contrasts always depend on the combination of different tissue properties, imaging parameters and employed hardware. Quantitative MRI (qMRI) techniques allow moving from such relative contrast information to a single, absolute measure of one or more separate tissue properties.

The measurement of water proton relaxation rates represents one major approach for quantitatively characterizing the human brain tissue microstructure with MRI. MR relaxation mechanisms depend on complex interactions among protons and between protons and their surrounding lattice. Hence, important determining factors of the physical parameters measured with qMRI are, for example, the concentration of water, macromolecules (for instance abundant in myelin) and paramagnetic atoms (e.g., iron). The exhibited sensitivity and specificity of qMRI acquisitions towards the microstructural properties of brain tissues drove an incremental investigation around qMRI metrics as biomarkers for alterations caused by disease. However, the practical use of qMRI in clinical settings is still hindered by difficulties in delivering precise quantification in an adequate resolution and within acquisition times comparable to conventional imaging. Moreover, to exploit the full potential of quantitative maps, normative values of physical parameters in healthy tissues are required, enabling the comparison of tissue properties from a single patient to normal values, potentially improving diagnosis and follow-up assessments.

With a focus on brain relaxometry and myelin water imaging, this thesis aims at bringing qMRI closer to clinical routine by tackling challenges ranging from the image acquisition to its practical application. More specifically, a validated T2 relaxometry sequence for myelin imaging - the multi-echo gradient and spin echo sequence - has been implemented and subsequently accelerated by combining it with parallel imaging to achieve whole brain coverage in a clinical compatible acquisition time. The direct estimation of microstructural features from relaxometry data has been also investigated using a fast protocol for T1 and T2 mapping and a data-driven approach that takes advantage of recent advances in the machine learning domain. Finally, the clinical value of qMRI in conjunction with normative atlases of physical properties has been explored to detect personalized changes in tissue parameters that reflect the underlying pathology.