Date of Award
Spring 2018
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Mathematics, Statistics and Computer Science
First Advisor
Clough, Anne V.
Second Advisor
Audi, Said H.
Third Advisor
Hamilton, Sarah J.
Abstract
Single Photon Emission Computed Tomography (SPECT) can be used to identify and quantify changes in molecular and cellular targets involved in disease. A radiopharmaceutical that targets a specific metabolic function is administered to a subject and planar projections are formed by imaging emissions at different view angles around the subject. The reconstruction task is to determine the distribution of radioactivity within the subject from the projections. We present a reconstruction approach that utilizes only a few view angles, resulting in a highly underdetermined system, which could be used for dynamic imaging applications designed to quantify physiologic processes altered with disease. We developed an approach to solving the underdetermined problem that incorporates a fast matrix- based multi-pinhole projection model into a primal-dual algorithm (Chambolle-Pock), tailored to perform penalized data fidelity minimization using the reconstruction’s total variation as a sparse regularizer. The resulting algorithm was implemented on a Graphics Processing Unit (GPU), and validated by solving an idealized quadratic problem. Simulated noisy data from a digital rat thorax phantom was reconstructed using a range of regularizing parameters and primal-dual scale factors to control the smoothness of the reconstruction and the step-size in the iterative algorithm, respectively. The approach was characterized by evaluating data fidelity, convergence, and noise properties. The proposed approach was then applied to few-view experimental data obtained in a preclinical SPECT study. 99mTc-labeled macroaggregated albumin (MAA) that accumulates in the lung was administered to a rat and multi-pinhole image data was acquired and reconstructed. The results demonstrate MAA uptake in the lungs is accurately quantified over a wide range of activity levels using as few as three view angles. In a pilot experiment, we also showed using 15 and 60 view angles that uptake of 99mTc-hexamethylpropyleneamineoxime in hyperoxia-exposed rats is higher than that in control rats, consistent with previous studies in our laboratory. Overall these experiments demonstrate the potential utility of the proposed method for few-view three-dimensional reconstruction of SPECT data for dynamic preclinical studies.