Position reconstruction in monolithic block detectors

In high resolution PET systems the detector generally uses a scintillator which consists of individual pixel elements. The scintillation light of such a pixel element will be identified and thus the interaction is localized by the pixel position. Consequently, the delivered position of such a detector can only take discrete values. A different approach is the monolithic scintillator detector. A continuous scintillator block spans over an area of several photodetector pixels and the position is reconstructed from the recorded light distribution. Manufacturing of this detector is easier and the sensitivity is generally higher as no scintillating material is wasted for optical isolation between the pixels. But the challenge is to find a dedicated algorithm in order to identify the interaction position with sufficient resolution. We will present measurements of a monolithic scintillator detector (21×18×10mm3 LYSO) and compare different reconstruction methods. Already a Least Square Optimization algorithm based on a rather simple model delivers a resolution similar to an Artificial Neural Network approach but which requires pre-registered data for training. The comparison of the resolution to that of a pixelated detector of similar size and 2×2×10mm3 pixels shows the superior performance of the continuous block.