List-mode SPECT reconstruction using empirical likelihood [electronic resource]

by Lehovich, Andre.

This dissertation investigates three topics related to image reconstruction from listmode Anger camera data. Our main focus is the processing of photomultiplier-tube (PMT) signals directly into images. First we look at the use of list-mode calibration data to reconstruct a nonparametric likelihood model relating the object to the data list. The reconstructed model can then be combined with list-mode object data to produce a maximumlikelihood (ML) reconstruction, an approach we call double list-mode reconstruction. This trades off reduced prior assumptions about the properties of the imaging system for greatly increased processing time and increased uncertainty in the reconstruction. Second we use the list-mode expectation-maximization (EM) algorithm to reconstruct planar projection images directly from PMT data. Images reconstructed by EM are compared with images produced using the faster and more common technique of first producing ML position estimates, then histograming to form an image. A mathematical model of the human visual system, the channelized Hotelling observer, is used to compare the reconstructions by performing the Rayleigh task, a traditional measure of resolution. EM is found to produce higher resolution images than the histogram approach, suggesting that information is lost during the position estimation step. Finally we investigate which linear parameters of an object are estimable, in other words may be estimated without bias from list-mode data. We extend the notion of a linear system operator, familiar from binned-mode systems, to list-mode systems, and show the estimable parameters are determined by the range of the adjoint of the system operator. As in the binned-mode case, the list-mode sensitivity functions define “natural pixels” with which to reconstruct the object. 11