High End Computing for PET/SPECT

描述（由申请人提供）：脑和其他器官的高分辨率PET成像继续经历巨大的技术进步。高分辨率研究层析成像仪（HRRT）就是一个例子，它是最高分辨率和灵敏度的人脑专用扫描仪，最近在美国国立卫生研究院SIG的赞助下安装在约翰霍普金斯大学，编号1S10RR017219-01A1。目前的重建软件和潜在的替代方案的计算需求是具有挑战性的，但可行的。我们正在请求计算机硬件来支持NIH资助的几个使用发射断层扫描（PET/SPECT）的项目。部分是将目前分配给HRRT常规重建的32个节点集群大小增加一倍，这将大大提高吞吐量，而不会牺牲大量（超过26个）动物和临床方案的定量准确性。随着我们的NIH研究从其他NIH NCRR支持的扫描仪转移，这些需求将变得更大，GE Advance变得过时（现在已经8岁了）。其次，一些NIH项目需要更多的计算时间。这包括HRRT量化的进一步改进，包括头部运动校正（HRRT有超过207个切片）和适当的校正，包括扫描仪等，使用基于Linux的程序进行更准确的重建需要在仅用于开发的集群上进行并行工作。此外，需要4个NIH SPECT/PET项目，包括密集的模拟和SPECT重建。因此，我们要求将其增加到24节点集群。这也将包括在心脏和癌症领域研究类似算法的合作者的算法开发。所有这些项目都由总共30个NIH资助，正在进行或试点研究需要这些计算设备用于人类，动物或计算断层扫描研究。通过提供这些额外的设备，这将允许充分利用已经资助的NCRR PET扫描仪，同时继续进行基础以及应用大脑和其他NIH相关研究，并将这种高分辨率扫描的限制扩展到以前未实现的应用中。

DESCRIPTION (provided by applicant): High resolution PET imaging for Brain and other organs continues to undergo enormous technological advances. One example is the High Resolution Research Tomograph (HRRT) the highest resolution and sensitivity human brain dedicated scanner which was recently installed under the auspices of an NIH SIG, 1S10RR017219-01A1, at Johns Hopkins University. The computational demands for the current reconstruction software and potential alternative are challenging but feasible. We are requesting the computer hardware to support several NIH funded projects employing emission tomography (PET/SPECT). Part is to double our current 32 node cluster size, allocated for HRRT routine reconstruction, which will substantially improve throughput without sacrificing quantitative accuracy for the large number (over 26) of animal and clinical protocols to be supported. These demands will become even greater as our NIH research studies shift from the other NIH NCRR supported scanner, and the GE Advance becomes obsolete (now 8 years old). Secondly, a number of NIH projects require more computational time. This includes further improvements in HRRT quantitation including head movement correction (the HRRT has over 207 slices) and appropriate corrections including scanner, etc., and more accurate reconstructions using Linux based program require a parallel effort on a development only cluster. Furthermore, 4 NIH SPECT/PET projects involving intensive simulations and SPECT reconstructions are needed. Hence we are requesting increasing this to a 24 node cluster. This will also incorporate algorithm development by collaborators working on similar algorithms in heart and cancer. All of these projects are supported by a total of 30 NIH funded, pending or pilot studies needing these computing facilities for human, animal or calculation tomography research. By providing this additional equipment this will allow full utilization of an already funded NCRR PET scanner while continued work by both basic as well as applied brain and other NIH related research and extend the limits of this high resolution scanning into applications not realized before.