Calibration process for improving Crystal Identification rate in the LabPET™ phoswich detectors

The LabPET™ is a small animal APD-based PET scanner using LYSO-LGSO phoswich detectors. A digital Crystal Identification (CI) method based on a Wiener filter, chosen for its low computation burden and low error rate, was implemented to localize the position of interaction in the phoswich. To facilitate its implementation, a unique Data AcQuisition (DAQ) model, in the Z-domain, was applied to every channel in the scanner, resulting in uneven CI error rates due to slight variations in detector channel characteristics. To improve the overall scanner identification performance, we propose to compute personalized DAQ model reference for every individual channel. Four individual DAQ models based on different minimization of cost functions were investigated. The first DAQ model uses a standard Wiener filter. The second DAQ model maximizes similarity to DAQ behavior. The third DAQ model attempts to minimize crystal decay time bias relative to the reference values and finally, the fourth DAQ model maximizes the CI itself. Results obtained on all detector channels of a 4 cm LabPET™ located at Sherbrooke show mean error rates of >;10%, 3.66%, 2.96% and 1.45% for the four models, respectively, and are to be compared to the current unique model implementation achieving an error rate of 3.53% for an energy window of 350-650 keV.