Region of Interest Image Reconstruction for Compton Imaging Using 3-D Position Sensing CdZnTe

Iterative methods, such as list-mode maximum likelihood expectation maximization (LM-MLEM) provide high signal-to-noise ratio reconstructions for Compton imaging. However, MLEM reconstructions can be computationally expensive when reconstructing a high-resolution image over the full field of view (FOV). When employing MLEM for high-resolution imaging, a vast number of data pixels would be needed. This work proposes a region of interest MLEM algorithm (ROI-MLEM) for Compton imaging, using 3-D position sensing CdZnTe, that allows for image reconstruction to take place within a fixed ROI based on prior knowledge of the approximate location of the source. ROI-MLEM is demonstrated by reconstructing simulated and experimental
¹³⁷
Cs source measurements. A preliminary study of the ROI-MLEM performance in high-energy imaging is shown by reconstructing the Compton image of the 4.44 MeV photopeak from a PuBe source. ROI-MLEM shows a 44% improvement in the signal-to-noise ratio (SNR) compared to standard MLEM for simulated source reconstructions in a truncated FOV. The experimental results show the capabilities of estimating the source location with submillimeter pixel resolution at a 30-cm source-to-detector distance, resulting in an average error of 1.4 mm in source location estimation.