Acquisition, estimation, and interpretation of diffusion- and relaxation-based cerebral MRI contrasts

Abstract

From the past few years, Magnetic Resonance Imaging has significantly enhanced our understanding of brain structure and function. In particular, developments in Diffusion MRI are providing unique contrast mechanisms heretofore unavailable from other imaging modalities. Although predominantly used for structural imaging, current research is hinting at their suitability for functional imaging too. The aim of research conducted as part of this thesis was to further our understanding of some of the recently proposed strucutral and functional diffusion contrast mechanisms. This work involved aspects of MRI sequence development, data modeling and interpretation. The possible role of diffusion MRI in functional imaging was explored in the first study described in this thesis. The main contribution of this work was to provide an alternative explanation for an important observation that formed the basis of one class of functional diffusion contrasts. This work highlights the possible confounds (statistical and physiological) that can affect potential functional diffusion contrasts. In the second study, we report about two alternative MRI contrast mechanisms for visualizing spreading depolarizations, a pathophysiological condition implicated in acute ischemic stroke, migraine with aura, and in delayed cerebral ischemia after subarachnoid hemorrhage. Apart from cross-comparing three MR contrast mechanisms, the key contribution of this work was to highlight the possibility of simultaneously obtaining complementary MRI contrasts (T2 and apparent diffusion coefficient) for furthering our understanding of spreading depolarizations. In the third study, we sought to gain deeper insight into the contrast mechanisms underlying the recently proposed Diffusion Kurtosis Imaging (DKI), an extension to widely used Diffusion Tensor imaging (DTI), in the framework of chronic experimental stroke. For this, comparisons between DKI and DTI changes following stroke were performed not only on select few regions of interest, as commonly pursued, but also using a machine-learning approach. Moreover, to understand the underlying histopathological changes driving post-stroke diffusion parameter changes, we compared various immunohistochemical data with diffusion parameter changes in two quantitative ways. Results from this study indicate that the combined range of microstructural sensitivity provided by DKI parameters forms a superset of that provided by DTI parameters and thus offer greater insight into tissue (re)organization after stroke. However, the non-specificity of diffusion-based parameters vis-a-vis the underlying biological processes was highlighted by aspecific correlation between MRI diffusion contrasts and histopathology. This calls for greater caution in interpreting diffusion imaging results. In the final part of the thesis we report about the advantages of using complex-valued MR data for estimation of basic MRI parameters (T1, T2 and T2*), unlike the commonly used strategy of using magnitude MRI data. Through simulations and with acquired data we demonstrate that fitting to complex-valued data can yield unbiased estimates and lower variance in comparison with estimates obtained with even the most accurate data models that use magnitude data. Similar extensions can perhaps be made for other MR estimation problems too, including mapping diffusion parameters.