Abstract
Image quality for PET imaging is an essential issue in measuring signal associated with new probes, tumor targets, and assays under preclinical and clinical study. One of the determining factors on image quality is the PET sensitivity. The loss of true coincidence due to multiple coincidences is an important factor for PET sensitivity, particularly at high radioactive concentration. Multiple coincidence events are discarded in most scanners as no proper ways to identify the true line of response (LOR). However, the majority of the triple coincidence contains true coincidence that can be recovered. As a result, the possibility of recovering true coincidences from triple coincidences could help enhance image quality advance. The aim of this dissertation was to improve the sensitivity of PET via the recycling of the triple coincidence for both pure and nonpure positron emitters. For pure positron emitters, a novel method is proposed to recycle the triple events based on the probability of each LOR that is truly formed by the annihilated pair. LOR with the maximum probability will be further tested to determine its origin.. If the LOR is inside the FOV it will be retrieved, otherwise the triple event is discarded. Results using simulated data showed the proposed method can retaining about 92% of the true coincidences from the triple events when the administrated activity is higher than 1 mCi. The percentage gain of noise equivalent count rate ranges from about 3% at activity of 1 mCi to 55% at 15 mCi. The proposed method poses a significant improvement in sensitivity for high activity PET imaging. For nonpure positron emitters, most of these isotopes emit high-energy gamma rays in addition to the positron decay that can be detected and create a triple coincidence with the annihilated photons. The triple events of non-pure positron emitter in PET imaging are more numerous and complicated than those of 18F. To this end, we proposed a strategy of recycling triple coincidence events for non-pure positron emitters. To identify the LOR from a triple coincidence, a framework utilizing the geometrical, energy and temporal information is proposed. The geometrical criterion is based on the assumption that the LOR with the largest radial offset among the three sub pairs of triple coincidence is least likely to be a true event. Then, a confidence time window is used to test the valid LOR among the LOR’s with triple coincidence. Finally, a likelihood ratio discriminant rule based on the energy probability density distribution of associated and annihilated gammas is established to identify the true coincidence. After recycling, the noise equivalent count rate (NEC) was found to be increased by from 11% to 30% and 22% to 45% for 124I and 76Br, respectively, when 0.2 to 5 mCi of activity were used. We conclude that ours proposed methods can improve the counting statistics for pure and nonpure positron emitters and are readily applicable to clinical studies as no hardware modification is needed.