ULTRA-Fast wiener filter based crystal identification algorithm applied to the LabPETTM phoswich detectors

A Wiener filter based crystal identification (CI) algorithm achieving excellent discrimination accuracy (~98%) was recently proposed for identifying LYSO and LGSO crystals in phoswich detectors coupled to an avalanche photodiode. Such detectors are used in the LabPETtrade, an all-digital positron emission tomography (PET) scanner for small animal imaging recently developed in Sherbrooke. This algorithm was based on evaluating the scintillation decay constant and gain (a₁ and b₀ parameters, respectively) of events sampled at 45 MSPS and discriminating crystals by thresholding the a₁spectra. The input gain was not considered in the CI process even if it must be computed. We propose a 2-fold faster CI approach which also takes into consideration the input gain coefficient of each crystal. The new algorithm incorporates the DAQ chain model -in the Z domain- to each individual crystal model evaluated in a Wiener filter calibration process. The identification is performed by evaluating a single parameter -compared to two parameters in previous Wiener CI algorithm- characterizing the percentage of each crystal gain contribution in the event signal. The CI algorithm demonstrated a discrimination rate accuracy >98.5% for LYSO-LGSO and >99% for LSO-GSO crystals in phoswich arrangement for 511 keV photopeak Although a calibration is required, initial implementation of the algorithm in a 400 MHz clocked FPGA can process up to 15 Mevents/sec.