Digital Signal Processing Techniques to Improve Time Resolution in Positron Emission Tomography

Coincidence time resolution is one of the most important issues in PET detectors. Improving this resolution is required to increase the noise equivalent count rate (NECR) that reduces the noise in the reconstructed images. The aim of this work is to evaluate the behavior and time resolution of different proposed time pick-off algorithms in order to select the best configuration for our PET system. The experimental setup used for this research is composed by two monolithic LSO crystals+PSPMT detectors and an FPGA based PET data acquisition system (DAQ). The acquired signals are sampled using a 12-bit 70 MHz analog to digital converter (ADC) per channel. The setup has no centralized electronics for trigger and event time extraction. Consequently, events for each detector head are processed independently and all the signals are acquired in the same way. Time resolution in this kind of systems can be improved by means of digital processing techniques and using different shapings for the last dynode signals. Four digital algorithms extracting time information from the acquired pulses have been evaluated: (1) Amplitude bipolar digital constant fraction discriminator (BCFD), (2) charge BCFD, (3) interpolated amplitude BCFD and (4) interpolated charge BCFD. Two different architectures for the interpolation algorithm have been used (one-sample and two-sample interpolation), which allow us to work with two different FPGA internal sampling frequencies: 140 MHz and 210 MHz. The results show the importance of selecting the right algorithm and parameters. Time coincidence resolution in our hardware system can be improved by up to 6.9 ns FWHM depending on the chosen digital algorithm programmed on the FPGA. The measurements with our setup reveal that charge based algorithms are less sensitive to signal noise and generate better results than amplitude algorithms. The best configuration achieves a FWHM resolution close to 1.8 ns.