List Mode Reconstruction in X-Ray CT and SPECT Imaging

Abstract

Since the introduction of tomography imaging in the 1970s, it has become an indispensable tool in medical imaging, driving continuous advancements in pursuit of higher image quality, reduced scan time, lower radiation doses, and more recorded attributes. List mode data, which records all attributes in a list format, offers distinct advantages over binned data, including reduced storage requirements, improved data accuracy, and enhanced flexibility in reconstruction techniques. This thesis explores the application of list mode data in two tomography systems. Firstly, we present a modified filtered back projection (FBP) utilizing list mode data, which allows continuous to continuous reconstruction from imaging space to object space. To mitigate the challenges posed by the singularity of the ideal ramp filter in the spatial domain, we introduce an exponentially decaying filter as an approximation. This modified filter not only emulates the characteristics of an ideal ramp filter but also functions as an apodizing filter. We qualify the performance of the novel FBP with two imaging quality metrics by a simulated single photon emission computed tomography (SPECT) system. The results demonstrate superior reconstruction quality when using list mode data compared to binned data, with the difference being particularly pronounced under low-dose conditions. The second part of this thesis extends the application of list mode data to a computed tomography (CT) system. To accommodate three-dimensional scan geometries, we develop a reconstruction method based on maximum likelihood estimation, subsequently transforming it into an iterative algorithm to obtain an analytical solution. We validate the algorithm's effectiveness using both a 3D Shepp-Logan phantom and a 3D patient phantom in a helical CT system. In the reconstruction process, graphics processing units (GPU) were applied to accelerate the computational speed. To present comprehensive results, the results under different noise levels and the impact of applied filters were demonstrated.