Flexible, ultrahigh-throughput and easy-implementing distributed coincidence processor for improving PET imaging performance

本项目旨在开发一种新的和创新的符合处理器，以克服 集中式符合处理器（CCP），已被许多商业和研究用正电子 发射断层扫描（PET）扫描仪自PET的最初发展。而CCP处理整个 与计算机中央处理器类似，使用单个复杂中央处理器进行PET符合事件选择的任务 处理单元（CPU），其具有有限的处理符合数据的计数率，因为单个 它是太复杂了，在线实现现场可编程门阵列（FPGA）的许多 研究小组没有广泛的专业知识和资源，我们建议使用分布式网络， 重合处理器（DCP）独立且并行工作，以处理每个重合数据 探测器与其专用的符合处理器配对，类似于图形处理单元（GPU）。通过打破 将单个复杂的系统级符合过程转化为许多简单的探测器对级过程，DCP可以 显著减少每个CP级别的处理延迟，从而增加整体数据吞吐量， 即使具有大量的检测器对。单探测器符合事件选择算法 一对探测器结构简单，易于实现，用一对探测器进行测试，并可直接复制 (or其余的人）。本项目的目标是设计、实施、评估、增强和 推广DCP技术。我们将努力实现三个具体目标，以实现这一项目的目标：（1） 设计DCP技术，包括DCP组件和功能的硬件架构， 固件程序，在FPGA上实现设计的DCP组件和功能。(2)要在 具有400个符合处理器的单个FPGA板，作为中小型PET的实用解决方案 数量探测器对，并在两个FPGA板上，每个板上有50个符合处理器作为示例 一种可扩展的解决方案，用于具有大量探测器对的PET;用于使用脉冲信号评估DCP， PET探测器，并通过迭代设计和开发增强DCP能力和性能 过程3)通过出版物和网站记录和传播DCP技术， 下载技术文档和固件/软件代码。如果开发成功，DCP将 为符合处理提供了一个新的、不同的技术平台，以解决CCP的问题。 作为一个游戏规则的改变者，DCP可以产生一个非常高的计数率PET在线符合数据采集远远超过 CCP可以提供的限制，可以在FPGA上实现，技术挑战比 实施CCP。通过解决CCP的问题并为研究界提供解决方案， 该项目将对提高PET成像的能力和性能产生变革性的影响 加快PET新系统和新技术的开发。

This project aims to develop a new and innovative coincidence processor to overcome the limit and drawback of centralized coincidence processor (CCP) that has been used by many commercial and research Positron Emission Tomography (PET) scanners since the initial development of PET. Whereas CCP handles the entire task of PET coincidence event selection with a single complex central processor, similarly to a computer central processing unit (CPU), that it has the limited count-rate of processing coincidence data because of a single processor and it is too complex to implement online on a field-programmable-gate-array (FPGA) for many research groups without extensive expertise and resources, we propose to use a network of distributed coincidence processors (DCP) that work independently and in parallel to process coincidence data for each detector pair with its dedicated coincidence processor, similar to a graphics processing unit (GPU). By breaking a single complex system-level coincidence process into many simple detector-pair-level processes, DCP can significantly reduce the processing delay at each CP level and therefore increase the overall data throughput, even with a large number of detector pairs. The algorithm for coincidence event selection with a single detector pair is simple and can be easily implemented, tested with one detector pair and be straightforwardly replicated (or populated) to the rest. The goal of this proposed project is to design, implement, evaluate, enhance, and disseminate the proposed DCP technology. We will pursue three specific aims to achieve this project goal: (1) To design DCP technology, including the hardware infrastructure of DCP components and functions and firmware program to realize the design DCP components and functions on FPGA. (2) To implement DCP on a single FPGA board with 400 coincidence processors as a practical solution to a PET with small to medium number detector pairs, and on two FPGA boards with 50 coincidence processors on each board as an example of an expandable solution to a PET with a large number detector pairs; to evaluate DCP with pulsed signals and PET detectors, and enhance the DCP capability and performance with an iterative design and development process. 3) To document and disseminate DCP technology through publications and a website with downloadable technical documentations and firmware/software code. If successfully developed, DCP will provide a novel and different technology platform for coincidence processing to solve the problems with CCP. As a game changer, DCP can yield a very high count-rate PET online coincidence data acquisition far beyond the limit of what CCP can provide and can be implemented on FPGA with much less technical challenging than implementing CCP. By addressing the problems with CCP and providing the solutions to the research community, this project would have a transformative impact on improving the capability and performance of PET imaging and accelerating the development of new PET systems and technologies.