Design of a Segmented Multi-Level Real-Time Coincidence Algorithm for In-Beam PET Readout Systems

摘要: In-beam positron emission tomography (In-beam PET) can serve as a non-invasive monitoring technique for heavy-ion tumor therapy. By detecting and identifying coincidence event pairs generated from positron-electron annihilation, it enables rapid and accurate imaging of the incident beam position and dose distribution. In this work, we propose a timestamp-based segmented multi-level pipelined real-time coincidence algorithm. This algorithm integrates the advantages of a merging tree structure and parallel insertion sorting, allowing for complete multi-channel data sorting within five clock cycles (T = 40 ns). Once the sorting is completed, the data are forwarded to the coincidence discrimination module for coincidence event selection and real-time imaging. The proposed scheme achieves an extremely low dead time. After preprocessing, events are first processed through a merging tree structure sorting stage, with the final stage employing parallel comparison. This design significantly reduces hardware resource consumption while maintaining stable operation at high event rates of up to 8.5 Mcps, demonstrating excellent feasibility and scalability. The algorithm was integrated into an In-beam PET prototype and tested using both background irradiation and 22Na radioactive source, while maintaining a disorder rate below 0.5%. In beam experiments using a 190 MeV carbon ion beam to irradiate a PMMA target, real-time Bragg peak imaging with a spatial resolution at the 2 mm was successfully achieved.