Semi-automatic position calibration for a dual-head small animal PET scanner

A dedicated small-animal positron emission tomography (PET) scanner that can provide ~30% detection sensitivity is under development at The University of Chicago. This scanner employs two HRRT (high resolution research tomography) detector heads in a compact configuration to achieve a high system sensitivity. The HRRT detector employs a quadrant-sharing configuration to substantially reduce the number of photomultipliers needed, but also creates particular challenges in calibrating the crystal position. To address these challenges, we develop a watershed segmentation method that also incorporates a curve-fitting forecast algorithm. Our results indicate that the watershed algorithm alone can correctly detect 92.77% of the positions for the 72x104 crystals of the HRRT detector head, when including the curve-fitting forecast algorithm the accuracy can be further improved to 97%.