Effects of incorrect interaction identification on image resolution in HPGe Compton cameras

The performance of Compton imaging systems is limited by the angular uncertainty arising from the detector geometry and spatial resolution [1,2]. When closely spaced multiple interactions are incorrectly recorded as a single event [3], termed "interaction packing", the system response has considerable angular and directional uncertainty. In this situation, we test two methods for assigning the event location; binning the combined interaction to a central location, and subsampling to an energy weighted centroid. We considered a dual layer camera geometry containing two SmartPET detectors [4], and employed Geant4 [5] to simulate the response to 662 keV photons. A cone-intersection algorithm [6] is utilised to reconstruct the activity distribution. The major findings are as follows. Interaction packing in the scatter detector leads to an increase in reconstructed background levels. Interaction packing in the absorption detector leads to a broader point spread function, but this is only observed using the energy centroid approximation. While both of these effects are small for the SmartPET based geometry, they may lead to image degradation when using different detection geometries, such as more closely spaced detection volumes.