An Optimal Filter Based Algorithm for PET Detectors With Digital Sampling Front-End

With the development of fast sampling electronics, digital pulse processing techniques for PET signals are raising interest. The optimal filter (OF) algorithm reconstructs pulse amplitude and time by two weighted sums, making it compatible with real-time implementation. The filters are usually optimized for stationary noise. We developed and tested a method to optimize the filters for the nonstationary noise of scintillation pulses. It is based on offline statistical analysis of coincident waveforms that could be applied during the system initialization phase. Experimental tests were done on a coincidence setup with two detection blocks composed of a fast inorganic scintillator ( LaBr3 or LYSO) coupled to a photodetector (APD or PMT), preamplifiers and prefilters. The signals were sampled at high rate (250 MHz for APDs, 5 GHz for PMTs) and treated offline. The optimization of the filter coefficients for nonstationary noise yielded a significant improvement compared to those optimized for stationary noise, resp. 368 ps and 632 ps fwhm in coincidence for the LYSO-PMT setup. However, little improvement was achieved compared to leading-edge (DLED) and constant fraction (DCFD) discriminator algorithms (resp. 419 ps, 435 ps fwhm). Indeed, the adjustment of thresholds can be interpreted as an optimization for nonstationary noise. Yet, OF is more robust to white noise than DLED or DCFD. The applicability to PET is discussed.