Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Biomedical Engineering


Dr. Maria Drangova


Magnetic resonance imaging (MRI) enables tissue characterization using the intrinsic tissue properties. By manipulating a number of imaging parameters different image contrasts can be achieved. Multi-echo gradient recalled echo (ME-GRE) enables acquisition of two distinct images, magnitude and phase, in a relatively short time. Phase image, specifically, contains a wealth of information for generation of quantitative maps, with the local tissue susceptibility differences as a source of contrast. GRE imaging is sensitive to field-inhomogeneities. This challenge presents more strongly in multi-channel acquisition, where the coil sensitivity variations impose complications in extracting information from the underlying anatomy. This calls for a phase-sensitive coil-combination approach. While many approaches have been presented to-date, ME-GRE remains an unpopular clinical tool due to the commonly observed susceptibility artifacts in phase and magnitude images.

This thesis presents ME-GRE acquisition considerations and post-processing tools for tissue characterization using readily available clinical acquisition protocols. The main contribution of the work presented here is in the proposed post-processing techniques that enable extraction of quantitative maps from these image data. These post-processing techniques are designed, optimized and validated for the first time through the work done in the present dissertation. The proposed approaches demonstrate the benefits of extracting information from each channel in the array of coils prior to combining the images.

The proposed techniques were applied in neurological and cardiac imaging. The former allowed for development of a robust approach, which enabled the extraction of tissue susceptibility information. The latter allowed for translation of these techniques to account for region-specific phase biases as well as different chemical environments. Quantitative maps of the brain and multi-parametric quantitative cardiac maps were generated for healthy participants as well as for cohorts of patients with multiple sclerosis and heart failure. The work of this thesis can easily be translated into clinic as it does not change the routine image acquisitions and has a focus on the post-processing workflow. With the techniques developed, non-contrast tissue mapping is made possible, especially benefiting patients with poor renal function.