Compton imaging system development and performance assessment.

Abstract

Current single-photon imaging systems in nuclear medicine employ mechanical (absorptive) collimation to restrict the directions of gamma-ray photons reaching the detector in order to form an image of the source distribution. These mechanical collimators couple detection efficiency and spatial resolution in an inverse relationship, thus compromising the system performance. This undesirable property places severe limitation on the ability to detect small tumors and quantify local radiotracer concentration accurately. Electronic collimation, an alternative photon collimation technique, uses the positions of interaction and energies deposited in scatter and absorption detectors to compute the direction of the original photon by applying the physics of Compton scattering. Without the use of mechanical collimators, spatial resolution and detection efficiency are no longer bounded by the inverse relationship. This dissertation reports investigations into the applicability of Compton imaging using electronic collimation to nuclear medicine. To demonstrate the applicability and superiority, limiting performance of Compton cameras has been computed theoretically and compared with that of conventional mechanically-collimated cameras. Lesion detection performance predicted by ideal observer theory and parametric estimation performance measured by uniform Cramer-Rao bound both show that Compton camera performance is object-dependent and improves as incident photon energy increases from 140.5 keV (<super>99</super><italic><super>m</super></italic>Tc) to 364.4 keV (<super>131</super>I). Compton cameras generally perform significantly worse than parallel-hole collimated cameras at 140.5 keV for equal number of detected photons (1.3&sim;5.5 times worse for signal-to-noise ratio or 1.6&sim;30 times worse for signal-to-noise ratio squared which is proportional to the number of counts and 17&sim;91 times worse for variance) and slightly worse than or comparable to their counterparts at 364.4 keV (1.1&sim;3.4 times worse for signal-to-noise ratio or 1.3&sim;11.5 times worse for signal-to-noise ratio squared and 0.5&sim;5.5 times worse for variance) depending on the object size. However, it has been shown that for a reasonable volume scatter detector and an efficient absorption detector, the efficiency of a ring Compton camera modeled in this study can be more than 30 times better than general-purpose collimator efficiency. This advantage allows the overall Compton camera performance to significantly outperform conventional camera performance for equal imaging time, especially for photon energies above 140.5 keV. In addition to theoretical analysis, a prototype Compton camera has been constructed in order to experimentally assess the performance. This camera consists of a silicon pad detector as the scatter detector and a ring sodium iodide detector as the absorption detector. Timing-coincidence detection circuits and data-acquisition software have been designed and implemented. Attributes of each coincidence event were stored in a list and images were reconstructed using the list-mode likelihood method. Two- and three-dimensional phantom studies show encouraging results. The prototype camera clearly resolves two vertical <super>99</super><italic><super>m</super></italic>Tc line sources 5 mm apart at 3.5 cm.