Abstract
This thesis reports on the development of material analysis techniques which can be used to process data collected by the Medipix All Resolution System (MARS) multi-energy CT scanner. Two techniques are presented - a post-reconstruction material decomposition algorithm based on volume conservation and material sparsity constraints; and an algorithm for polychromatic material reconstruction based on a non-linear algebraic reconstruction formulation. I also show a fundamental limitation of K-edge subtraction imaging, which identifies a ``noise independent'' minimum detectable concentration for K-edge contrast pharmaceuticals. As I am the MARS research team's primary consultant for solving material analysis problems, part of the impact of my work is reflected in both my research outcomes and those of my colleagues. I show a sample of the results from pre-clinical projects that I have consulted for and in which my material analysis algorithms have been used.
The post-reconstruction material decomposition algorithm presented in this thesis, known as MARS-MD, has been the work-horse material analysis algorithm used by the MARS research team over the past several years. It has been used in projects involving imaging of soft tissues, atherosclerosis, osteoporosis arthritis, and contrast pharmaceutical imaging in small animals. Phantom studies show that this algorithm is capable of discriminating at least six different materials simultaneously excluding air (lipid, water, calcium, iodine, gadolinium, and gold) using multi-energy CT measurements in only four energy ranges.
The limitation with the current MARS data processing chain is that it is based on techniques which assume the polychromatic x-ray source used in MARS emits a monochromatic x-ray beam. This is also a common assumption that is made in both conventional and dual-energy CT systems and is the cause of image beam-hardening artifacts. To resolve this problem we have begun developing a polychromatic material reconstruction algorithm designed specifically for data obtained using MARS. This task is still progressing through the initial development phase, with various members of the MARS team working on different components of the problem. In this thesis I present the prototype material reconstruction algorithms that I have been developing as potential candidates to include in the new MARS data processing chain.