MRI: Acquisition of a High-Performance Instrument for Heterogeneous and Biologically Inspired Architectures Research at CUA

提案号：11-26120PI(s): El-Araby, Esam；Chang Lin-Ching;机构：美国天主教大学（CUA）职称：MRI/Acq。研究项目建议：该项目获取SRC-7系统和GPU集群并将其集成到一个高性能异构计算仪器中，旨在研究、建模、设计和潜在原型化受生物学启发的新型异构架构。该仪器将有助于扩展一些可用的硬件加速器，如可重构现场可编程门阵列（fpga）和图形（GPGPU）处理器，目前用于高性能计算、计算神经科学和医学信息学、电磁和遥感等领域的研究。它将促进对设计/发现易于使用且高效的新型异构架构的研究。此外，该仪器还将允许通过与机构内外其他研究小组的跨学科合作来调查新处理器技术的使用。在计算神经科学和医学信息学中，高性能集群（HPC）将有助于设计和开发新的计算方法来处理和分析医学成像数据，如扩散张量磁成像（DT-MRI）。尽管DT-MRI因其独特的描绘白质束的能力以及对脑组织微结构和结构特征的敏感性而越来越多地用于临床研究，但它存在许多图像伪影，这些伪影会破坏扩散加权图像（dwi），从而影响DTI衍生量。虽然并行计算是理想的方法，但DT-MRI数据的后处理非常困难且耗时。新仪器将有助于测试新的后处理算法，探索现有技术的局限性，并开发解决方案。在电磁和遥感应用中，新仪器将有助于提高商业和军事应用研究界所面临的数字密集型问题的计算时间。超材料的设计与优化、平台天线的性能分析与优化以及杂波（如土壤、树叶、雾）后的目标检测是后者的一些例子。更广泛的影响：该仪器使现实的计算问题能够整合研究和教学，并通过发现促进学习。因此，预计它将有助于培养新一代的学生和博士后，他们有能力面对当前计算机技术的变化。此外，该仪器能够创建一个新的跨学科异质和生物学启发的建筑研究实验室（HEBA）。该项目包括一个外展计划，包括社区大学、高中、初中学生和K-12教师参加的开放日活动。增加妇女和少数民族也是目标。结果将被分享和传播。

Proposal #: 11-26120PI(s): El-Araby, Esam; Chang, Lin-Ching; Kilic, OzlemInstitution: Catholic University of America (CUA)Title: MRI/Acq.: High Performance Instrument for Heterogeneous and Biologically Inspired Architectures Research Project Proposed:This project, acquiring and integrating an SRC-7 system and a GPU cluster into one High Performance Heterogeneous Computing instrument, aims to investigate, model, design, and potentially prototype new heterogeneous architectures that are inspired from biology. The instrument will contribute to expand some of the available hardware accelerators such as reconfigurable Field Programmable Gate Arrays (FPGAs) and graphical (GPGPU) processors currently utilized for research in the fields of high-performance computing, computational neuroscience and medical informatics, and electromagnetic and remote sensing. It will facilitate research towards designing/discovering new heterogeneous architectures that are easy-to-use and highly productive. Moreover, the instrument will also allow investigating the use of new processor technologies through interdisciplinary collaborations with other research groups inside and outside the institution. In computational neuroscience and medical informatics, the high-performance cluster (HPC) will help in designing and developing new computational methods to process and analyze medical imaging data such as diffusion tensor magnetic imaging (DT-MRI). Although DT-MRI is used increasingly in clinical research for its unique ability to depict white matter tracts and for its sensitivity to microstructural and architectural features of brain tissue, it suffers a number of image artifacts that can corrupt diffusion weighted images (DWIs) which in turn affect DTI derived quantities. Although ideal for parallel computation, the post-processing DT-MRI data is very difficult and time consuming. The new instrument will help test new post-processing algorithms, explore the limitations of the existing technology, and develop solutions. In electromagnetic and remote sensing applications, the new instrument will help enhance the computation times of the numerically intensive problems faced by the research community for commercial and military applications. Design and optimization of metamaterials, performance analysis and optimization of antennas on platforms, and target detection behind clutter (e.g., soil, foliage, fog) represent some examples of the latter.Broader Impacts: This instrumentation enables realistic computational problems that can integrate research and teaching as well as promote learning through discovery. Hence, it is expected to contribute in producing a new generation of students and postdocs with the ability to face the current changes in computing technology. Moreover, the instrument enables the creation of a new interdisciplinary HEterogenous and Biologically Inspired Architectures research lab (HEBA). The project includes an outreach program, open house events that include community colleges, high school, and middle school students, and K-12 teachers. Increasing women and minorities is also targeted. Results will be shared and disseminated.