A Linux Cluster Computational Facility for Neuroimaging Research

Song, Allen

Abstract

Neuroimaging research using parallel acquired functional MRI and high-resolution structural MRI, and dense array electrical recording involves the acquisition of very large data sets that then must be transformed through a series of computation-intensive processing tasks into meaningful information. This project, entitled """"""""A Linux Cluster Computational Facility for Neuroimaging Research"""""""", is jointly submitted by NIH-supported investigators at Duke University and the University of North Carolina who primarily conduct neuroimaging research at the Duke-UNC Brain Imaging and Analysis Center at Duke University Medical School. The collaborating scientists have found that the sheer volume of data resulting from improved acquisition methods and improved hardware has made it increasingly difficult to fully analyze data from all the labs in a timely manner. With our existing computational infrastructure, some processes such as the iterative reconstruction of time-course spiral images acquired through eight to sixteen parallel receiver channels would take over 20 hrs for a typical ten-run functional exam, and a typical brain image segmentation and flattening procedure would take more than 24 hrs. Thus, we are requesting funds to purchase a high-performance Linux cluster to meet these and other processing requirements at a relatively low cost. Our application builds upon our past experience administering a multiprocessor, though quite dated, supercomputer (IBM p670 supported by NCRR in 2002) and our recent pilot project using an eight-node Linux cluster (Dell Power Edge 1950, supported by Duke internal funds), with high-speed fiber channel adaptors and connections to our multiple data servers totaling over 40TB online storage. We anticipate that this much-improved computational resource will greatly enhance our ability to meet the much-increased MRI acquisition capability and analysis demand in the many NIH-sponsored research projects, and gain maximum efficiency at a low overall cost. In addition to meeting the scientific needs of our neuroimaging scientists, this application also benefits from the close collaboration that already exists among us in the development of analytical software and in the management of large research instruments.

Public Health Relevance

We propose in this S10 application to purchase a 50-node Linux cluster to meet the ever-increasing computational demands from our NIH-funded neuroimaging investigators, as the results of much improved MR imaging hardware and analysis software in the past five years. Because most of our projects share common pre- and post-processing steps such as parallel image reconstructions for fast spiral acquisitions, brain surface flattening and dynamic causal network modeling, this shared instrumentation will greatly improve the productivity and efficiency of our research projects. We anticipate that this cluster will help us gain maximum computational power at the lowest cost taking into consideration the large amount of projects it will support.