Improved image reconstruction in small animal PET using a priori estimates of single-pixel events

Most small animal PET scanners are based on arrays of pixelated scintillators crystals. As the read-out of individual pixels would be too expensive, identification of the crystal of interaction is usually made by center of energy methods based for instance on Anger logic. This allows for a reduction in the number of signals to be acquired, but prevents the identification of multi-hit events, that is, events in that one (or several) photon produces several hits in the detector, thus blurring the correct positioning of the interaction. Improving the identification of the pixel of interaction is pursued in this work by combining all the information acquired by the scanner without increasing the number of signals. The probability for every individual event for being single or multi-pixel is estimated from the XY positioning and energy information. This probability is fed into a 3D-OSEM iterative statistical reconstruction method. Every coincidence event detected may be analyzed combining information such as deposited energy, PMT XY location, time difference between both singles of the coincidence and coincidence and single rates, if available. With the proposed method, improved peak/noise ratio and better resolution are obtained without the introduction of additional hardware.