ITR: Fine-Grain Data Management in Computational Grids and Applications in Network-Enabled Medical Imaging for Early Cancer Detection

在过去的几十年里，高性能计算推动了现在广泛应用的实际医疗应用的发展，例如磁共振成像和计算机断层扫描。近年来，信息处理正在经历由使用通过全球网络连接的分布式计算系统驱动的快速发展。类似于电力网，“计算网格”有可能提供从无处不在的网络设备无缝访问高性能计算机的能力。这种模式所实现的前所未有的计算水平可能会促进新的医疗应用的开发，从而大大改善医疗保健。在一类新兴的医学应用中发现了一个例子，这些应用使用光散射光谱（LSS）成像来允许体内检测人类上皮中的癌前变化。LSS映像的有效部署将取决于高级别性能的可用性，并需要访问远程资源。该项目旨在提高网格数据管理技术的最新水平，以实现医疗仪器设备生成的数据与分布式计算机的无缝和高性能集成。长期目标是实现基于网络计算的医疗应用的部署，用于早期癌症检测，这些应用需要超出医疗机构可用的处理能力。特别是，该项目将开发计算技术，以便能够准确和快速地分析LSS图像。第一个目标是开发高性能的LSS分析算法的并行实现。第二个目标是数据管理技术的发展，允许按需访问的数据所产生的网络使能的仪器设备从场外计算资源，这些技术是基于每个用户的虚拟文件系统代理的概念，由网格中间件控制，并允许latencyhiding性能增强从操作系统和应用程序实现解耦。第三个目标是将所提出的解决方案与计算网格基础设施相结合，以便向研究界传播。该项目的结果将被使用，并导致与西北大学医学院和西北纪念医院合作开发和实施这种LSS成像系统用于临床。

In the past few decades, high-performance computing has driven the development of practical medical applications that are now widely available, such as magnetic resonance imaging and computerized tomography. In recent years, information processing is undergoing rapid advances driven by the use of distributed computing systems connected by world-wide networks. Analogous to power grids, "computational grids" have the potential to provide seamless access to high-performance computers from ubiquitous, network-enabled devices. The unprecedent levels of computation enabled by this model may foster the development of new medical applications that can substantially improve healthcare. An example is found in a class of emerging medical applications that use Light-Scattering Spectroscopy (LSS) imaging to allow in-vivo detection of pre-cancerous changes in human epithelium. Effective deployments of LSS imaging will depend on the availability of high levels of performance and require access to remote resources. This project aims to improve the state-of-the-art in data management techniques for grids to allow seamless and high-performance integration of data generated by medical instrumentation devices with distributed computers. The long-term objective is to enable deployments of network-computing based medical applications for early cancer detection that require processing capabilities beyond those available in healthcare facilities. In particular, this project will develop computing techniques to enable accurate and fast analyses of LSS images.To this end, this project will focus on three specific aims. The first aim is the development ofhigh-performance, parallel implementations of LSS analysis algorithms. The second aim is the development of data management techniques that allow on-demand access to data generated by a network-enabled instrumentation device from off-site computing resources; these techniques are based on the notion of per-user virtual file system proxies that are controlled by grid middleware and allow latencyhiding performance enhancements to be decoupled from operating system and application implementations. The third aim consists of the integration of the proposed solutions with computational grid infrastructures to enable dissemination to the research community. The results of this project will be used and lead to the development and implementation of such LSS imaging system for clinical use in collaboration with the Northwestern University Medical School and Northwestern Memorial Hospital.