Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Biomedical Engineering

Supervisor

Charles A. McKenzie

2nd Supervisor

Terry M. Peters

Joint Supervisor

Abstract

Quantifying the tissue fat concentration is important for several diseases in various organs including liver, heart, skeletal muscle and kidney. Uniquely, MRI can separate the signal from water and fat in-vivo, rendering it the most suitable imaging modality for non-invasive fat quantification. Chemical-shift-encoded MRI is commonly used for quantitative fat measurement due to its unique ability to generate a separate image for water and fat. The tissue fat concentration can be consequently estimated from the two images. However, several confounding factors can hinder the water/fat separation process, leading to incorrect estimation of fat concentration.

The inhomogeneities of the main magnetic field represent the main obstacle to water/fat separation. Most existing techniques rely mainly on imposing spatial smoothness constraints to address this problem; however, these often fail to resolve large and abrupt variations in the magnetic field. A novel convex relaxation approach to water/fat separation is proposed. The technique is compared to existing methods, demonstrating its robustness to resolve abrupt magnetic field inhomogeneities.

Water/fat separation requires the acquisition of multiple images with different echo-times, which prolongs the acquisition time. Bipolar acquisitions can efficiently acquire the required data in shorter time. However, they induce phase errors that significantly distort the fat measurements. A new bipolar acquisition strategy that overcomes the phase errors and provides accurate fat measurements is proposed. The technique is compared to the current clinical sequence, demonstrating its efficiency in phantoms and in-vivo experiments. The proposed acquisition technique is also applied on animal models to achieve higher spatial resolution than the current sequence.

In conclusion, this dissertation describes a complete framework for accurate and precise MRI fat quantification. Novel acquisitions and reconstruction techniques that address the current challenges for fat quantification are proposed.

Share

COinS