基于FPGA的PET快速符合系统设计

FPGA-based rapid coincidence design for PET system

符合探测是PET成像技术的本质,其设计好坏直接影响着整个系统的性能。针对PET系统的多通道、高速、大数据处理的特性,首先引入基于视野的自适应时间窗,并将其窗值预存在FPGA RAM中以供符合查找,这在一定程度上降低了PET系统随机散射的概率,保证了符合数据的有效性;其次针对系统对后端处理速度的需求,结合FPGA并行处理的优点,在子板FPGA中提出利用并行全比较排序算法对模块时间数据进行一次排序,待数据汇总到母板FPGA中时进一步利用FIFIO缓存,完成所有子板数据的二次排序;最后为了保证数据不被丢失,提出基于多时间片的并行处理的架构,将多个时间片内的数据缓存进行乒乓处理,达到系统对实时性的要求。最终经测试表明,这种基于FPGA的以空间换时间符合策略满足人体PET成像装置对后端处理速度及实时性的要求。

Coincidence detection is the essence of PET imaging system,and its design directly affects the performance of the whole system.Focusing on the characteristics of multi-channel,high-speed and bulk data handling in PET systems,the paper first introduced the adaptive time window based on field of view and pre-stored their values in FPGA RAM for table look-up,which could reduce the random scattering probability of PET system and ensure the validity of coincidence data.Secondly,to meet the processing speed requirements of the system on back-end circuit,the first sorting was performed in childboard FPGA using comparsion sort algorithm based on the advantages of parallel processing of FPGA.And the second sorting was excuted in motherboard FPGA when all the data was cached to FIFO.Lastly,the time slice-based parallel processing architecture was proposed to ensure that system data was not lost.Multiple time-slice data was stored for "ping-pong structure" post-processing to obtain the system real-time performance.Finally,this FPGA-based coincidence strategy trading space for time was proven to fulfill the back-end processing speed and real-time requirements of human PET system through tests.