FFT- and DWT-based FPGA realization of pulse shape discrimination in PET system

Abstract

In positron emission tomography (PET) scanners, the depth of interaction (DOI) is a great interest research area. To gain the DOI information, different pulse shape discrimination (PSD) techniques are used for crystal-identification of phoswich detectors. This paper evaluates two recent PSD algorithms due to their discrimination performance of two different pulses types. These pulses recorded from LSO and LuYAP crystals. The two evaluated PSD algorithms are based on the Fast Fourier Transform (FFT) and the discrete wavelet transform (DWT). The designs and FPGA implementations of the two algorithms are proposed and compared. For PET events measured with LSO-LuYAP phoswich crystals, the FFT-based and the DWT-based achieve a successful discrimination rate of 97.18%, and 99.3%, respectively. Based on Xilinx XCV800 chip, the DWT-based algorithm realization saves 26% of slices, and 10% of computation time with respect to the FFT-based realization.