Speaker
Description
A scintillator-based Compton camera (CC) is a γ-ray detector, often utilized for medical imaging, astronomy, and homeland security. In these applications, high spatial resolution is vital for accurately identifying tumors, resolving distant celestial objects, and guiding decontamination efforts following nuclear incidents. Degradation of the spatial resolution is significantly caused by uncertainty in the interaction positions within the camera. The accuracy of the interaction positioning depends on the size of each scintillator, often referred to as a voxel. Decreasing the voxel’s size reduces the positioning uncertainty, which in turn, improves the spatial resolution of the reconstructed image. Decreasing voxels size will decrease the CC volume, which in turn, will reduce its sensitivity. Therefore, we suggest not changing the volume of the CC during this process. This results in the addition of voxels, which in turn requires more readout channels. The addition of readout channels increases power consumption, required electronics, and acquired data to process.
In this study, we introduce a novel approach to divide the volume of a CC into more voxels without adding more readout channels. This is achieved by connecting each readout channel to groups of voxels. When acquiring events from groups of voxels, it is impossible to determine in which of the group’s voxels the event occurred. Thus, in addition to the true event, many “pseudo” events are considered in the data set. This concept is illustrated in Figure 1. The addition of the “pseudo” events implies that the assumptions on the data acquisition procedure based on which the widely-in-use Maximum likelihood expectation maximization (MLEM) algorithm is designed, no longer hold, necessitating its adaptation. We utilize knowledge about the groups’ formation to develop a new image reconstruction algorithm that extends the MLEM for voxels-grouped CC. We evaluated the resolution enhancement achieved with voxel grouping through simulations. Using the GEANT4-based Architecture for Medicine-Oriented Simulations (GAMOS) software, we simulated setups with single and two adjacent 137Cs γ sources. Various voxel grouping sizes were tested to assess spatial resolution improvement with increased voxel count through grouping. The resulting Compton images are shown in Figure 2 and Figure 3 for grouping sizes of G = {1, 4, 16} and a constant number of readout channels R = 50. The results systematically demonstrate that voxels grouping method with our extended MLEM leads to better γ-ray sources localization and separation without adding more readout channels. This novel approach holds the potential for developing high spatial resolution, cost-effective, and mobile CCs.
