Compton Imaging Algorithms for Position-Sensitive Gamma-Ray Detectors in the Presence of Motion.

Abstract

Position-sensitive gamma-ray spectrometers, like pixelated CdZnTe detectors, record the energies and 3-D positions of gamma-ray interactions. This information allows one to perform Compton-image reconstructions which can determine the source direction in the 4-pi space surrounding the detector. This work describes several image-reconstruction algorithms for use in scenarios where motion is involved. In the absence of motion, standard simple back projection and maximum-likelihood image-reconstruction techniques are appropriate. When the source is close to the detector or the detector is moving relative to the source, 3-D image reconstruction is possible. This reconstruction allows one to localize the source position in 3-D rather than just the source direction. The 3-D reconstruction method using a moving detector is demonstrated by mapping radiation sources in a room and provides a convenient way to view the results overlaid on an optical image. In contrast, when the source is moving relative to the detector, images produced by standard image reconstructions are blurred by the source motion. If the source motion is known, the reconstruction can account for the motion by adjusting the reference frame of the reconstruction to keep the source in the center of the field of view. This type of compensation works well when a single source is present, but it cannot simultaneously reconstruct multiple sources. Thus, this work introduces a new binning structure that allows multiple moving or stationary sources to be reconstructed with minimal blur related to source motion. This method is demonstrated using experimental data of multiple moving and stationary sources. Finally, when the source motion is unknown or there are too many sources to track, the time domain must be included to reconstruct radiation movies instead of radiation images. Simply adding time binning works well when the source is strong and there are many counts reconstructed in each time bin. However, when there are only a few counts per time bin, the maximum-likelihood reconstruction method is modified to enforce smoothness in the time domain which produces much improved results over the standard maximum-likelihood reconstruction. This method is also validated with experimental data.