A cross-correlation method for crystal identification in APD-based phoswich detectors used in small animal PET scanner

PET images are based on extracting pertinent information from detector signals, such as energy, time stamp and crystal identification. The arrival of all-digital PET architectures coupled to recent improvement of digital devices such as FPGA enables the exploration of new real-time digital algorithms able to extract such information. The LabPET™, an avalanche photodiode-based small animal PET scanner dedicated to molecular imaging, allows such developments. Its detectors are composed of a dual 2×2×∼12 mm3 LYSO (tr∼40 ns) and LGSO (tr∼65 ns) crystal juxtaposed along their long side for higher pixelization. A novel approach based on the cross-correlation theory has been investigated for crystal identification in these phoswich detectors. The proposed method consists of interpolating the signal sampled at 45 MSPS by cubic spline technique followed by a de-noising process based on stationary wavelet transform and finally using cross-correlation as a crystal identification sorter. This method can achieve a successful discrimination rate of ∼96% for PET events measured with an LYSO–LGSO phoswich crystal combination read out by an avalanche photodiode.