Evaluation of easily implementable inter-crystal scatter recovery schemes in high-resolution PET imaging

The detection efficiency of high-resolution PET systems based on arrays of pixelated detectors with individual crystal readout can be increased by lowering the energy threshold to recover low-energy first Compton events. However, allowing such low-energy events also results in inter-crystal scatter processes that generate triple coincidences (or "triplets") with ambiguous line-of-response (LORs). Whereas the problem has been investigated extensively by Monte Carlo simulations, little experimental data exist to confirm findings. Taking advantage of the fully parallel data processing and acquisition system of the LabPET scanner, every hits belonging to multiple events, which are normally discarded at an early stage in the data processing, were recorded in the research list mode with the relevant information (crystal position, energy and timestamp). Four different algorithms that can be readily implemented in the real-time coincidence processor were then used to select the LORs and build the corresponding acquisition sinograms. Images of a NEMA phantom were then reconstructed to assess the consequences of the four different recovery algorithms on detection efficiency and image quality. One of the main goals of the study was to determine whether simple triplets recovery schemes could be used as a solution to increase system detection efficiency while preserving image accuracy with real data sets. Results indicate that the addition of triple coincidences in the PET image formation process has an interesting potential to increase detector efficiency without significantly degrading image quality.