A Neural Network Based Algorithm for Building Crystal Look-up Table of PET Block Detector

Crystal Look-up table (CLT) used in the Siemens Inveon dedicated PET scanner defines the matching relation between signal position of a detected event to a corresponding detector pixel location. It is the result of the first stage scanner calibration and brings significant influence to the gantry overall performance. The currently used method involves a lot of human interaction for CLT corrections, and can not be implemented as an on-line process due to its complexity. This paper introduces a neural network based algorithm for crystal identification. A modified unsupervised self-organizing feature map (SOFM) is trained by the incoming events to construct a CLT. The algorithm is implemented in a Field Programmable Gate Array (FPGA) chip in the Inveon Event Processing Module (EPM) electronics, which significantly reduces the training time and brings feasibility to detector on-line monitoring. The preliminary training result shows that SOFM can be used effectively in CLT construction with excellent accuracy.