Wavelets-based crystal identification of phoswich detectors for small-animal PET

Abstract--The advent of new all-digital electronic architectures in PET scanners enables the development and investigation of novel crystal identification algorithms for phoswich detectors used for parallax mitigation or higher detector pixelization. The high flexibility and real-time signal processing capability of FPGA/DSP-based digital electronics, such as the one developed for the LabPET scanner, provide an excellent platform to test enhanced digital methods. A novel approach based on the wavelet analysis theory has been investigated for crystal identification in phoswich detectors with crystals having similar scintillation characteristics such as LYSO ([t.sub.r] ~40 ns) and LGSO ([t.sub.r] ~65 ns). The proposed algorithm uses Stationary Wavelet Transform to clean the digitized signal and Discrete Wavelet Transform for crystal identification. Such a process can achieve a successful discrimination rate of ~95% for PET events measured with an LYSO-LGSO phaswich crystal combination readout by an avalanche photodiode. Index Terms--Crystal identification, LabPET scanner, phoswich detector, small-animal PET, wavelet transform.