Digital signal processing applied to crystal identification in Positron Emission Tomography dedicated to small animals

The recent introduction of all-digital electronic architecture in Positron Emission Tomography (PET) scanners, enables new paradigms to be explored for extracting relevant information from the detector signals, such as energy, time and crystal identification. The LabPET™ small animal scanner, which implements free-running 45-MHz sampling directly at the output of the charge sensitive preamplifiers, provides an excellent platform to test such advanced digital algorithms. A real-time identification method, based on an Auto-Regressive Moving-Average (ARMA) scheme, was tested for discriminating between LYSO (tr∼40 ns) and LGSO (tr∼65 ns) scintillators in phoswich detectors, coupled to a single Avalanche Photodiode (APD). Even with a low energy threshold of 250 keV applied individually, error rates<4% can be achieved, as compared to >10%, typically with conventional analog pulse shape discrimination techniques. Such digital crystal identification techniques can be readily implemented with phoswich detectors for improving spatial resolution in PET, either by increasing crystal pixellization or by mitigating parallax errors through depth-of-interaction determination. It also allows to reduce the event rate presented to the real-time coincidence engine by applying a low energy limit at the crystal granularity and rejecting more Compton photons.