Cutting-edge Biomedical Research Network Launched in San Diego
7-31-02 - Academics and computer scientists at leading research universities around the country have begun to deploy a large-scale network to share various types of brain images and related data likely to speed up research on disorders ranging from Alzheimer's disease to schizophrenia. On Oct. 25, they attended the inaugural workshop of the Biomedical Informatics Research Network (BIRN) at the San Diego Supercomputer Center on the campus of the University of California, San Diego. The three-year project received a $20-million award in early October from the National Center for Research Resources (NCRR) at the National Institutes of Health (NIH). Participating labs will begin installing standardized hardware and software in January and, within a year, researchers expect to begin accessing brain images hundreds or thousands of miles away.
"We have been able to pull people and technology together in such a short period of time because all our institutions recognize the importance of this effort to brain-related research," said UCSD neuroscientist Mark Ellisman, one of the principal investigators for BIRN and an academic participant in Calit². "With the advanced networking and data-mining technology we are putting in place, we expect to demonstrate new ways to make exciting advances in the study of brain disorders."
Ellisman says Calit² director Larry Smarr was influential in bringing the BIRN project to UCSD: "As a pioneer in the area of supercomputing and the Internet, Larry knows that future medical research will increasingly rely on technology-enabled, multi-institutional sharing of huge data-sets." Smarr, a member of the Advisory Committee to the Director of NIH, says the BIRN project is important to Calit²'s mandate to build networked prototype systems which harness cutting-edge telecommunications and information technology for real-world solutions.
"BIRN will be a showcase of the Calit² program in digitally enabled genomic medicine," said Smarr. "We also plan to work with BIRN to develop a high-speed wireless capability to enable anywhere/anytime access to the data and images for researchers on this network."
At the Oct. 25 meeting, officials confirmed that the BIRN Network Operations Center-NOC or "BIRN Central"-is located at the San Diego Supercomputer Center (SDSC), also at UCSD. It will be linked to the research centers that shared in the NCRR award: Duke University in Durham, North Carolina; Harvard's Massachusetts General Hospital and Brigham and Women's Hospital in Boston; Caltech in Pasadena, California; UCSD's School of Medicine; and the University of California-Los Angeles. Participants will collaborate on sub-projects involving mouse and human brain images. The so-called Mouse BIRN will address a neurological disorder similar to multiple sclerosis, as well as changes in brain dopamine levels (like those found in Parkinson's disease and schizophrenia). The initial clinical focus in the Brain Morphology BIRN project (human subjects) will be on depression and Alzheimer's disease.
Sharing high-resolution brain images across institutions presents huge challenges, not least the size of the data-sets involved. "One three-dimensional brain image can contain as much as one terabyte of data, so when you're talking about sharing images through the network, you're dealing with a massive bandwidth challenge," said Phil Papadopoulos, Program Director, Grid and Cluster Computing at SDSC. Papadopoulos, a member of Calit²'s Interfaces and Software Systems group, is in charge of implementing middleware for BIRN sites. "As scanning technologies become more powerful, the data increases exponentially. If you double the resolution of a brain scan, the amount of data increases four times; and if it's a 3-D scan, the data expands by a factor of eight."
To meet that challenge, the network infrastructure for BIRN will piggyback on existing high-speed Internet2 networks-the Abilene high-speed backbone nationally, and CalREN-2 within California. "Even with the huge amount of data involved, this infrastructure will allow a researcher at UCLA to get quicker response from a site in Boston than they would if they were using most local-area networks," said SDSC network expert Jay Dombrowski. "So the network is not an issue; our primary concern will be the 'last mile'-from the network to each end point." Dombrowski predicts the first pre-configured systems will be installed by January 15, with several sites operational a month later, and most sites up and running by April 1.
UCLA's Art Toga predicts researchers will begin sharing images over the network within a year. "My lab is already hooked up to Abilene, so we are working with BIRN to solve the compatibility issue in data structures between sites," said the director of the university's Laboratory of Neuro Imaging. "Fortunately, this project is getting a jump-start because we all have worked together. There is also a tremendous amount of pressure from funding agencies and constituencies to make this happen. The genome folks did the same thing, and their teamwork expedited those discoveries. We expect BIRN will allow brain researchers to benefit from these economies of scale and compatibility, so this solution could become a model for other medical research areas."
On top of the network infrastructure, BIRN will develop and deploy a "virtual data grid" across the equipment, which will be a state-of-the-art configuration like several already demonstrated for particle physics. The grid will allow researchers affiliated with the project to share the large volume of data being developed.
Apart from the huge size of the data-sets and establishing compatibility across disparate data-sets, the biggest challenge for scientists setting up BIRN will be in data mining and interpretation. "How do you link together data stored not only in different sites, but also where data is not related in a straightforward way?" asked Maryann Martone, a researcher at UCSD's National Center for Microscopy and Imaging Research. "Researchers have data taken at different scales, at different resolutions, across different species, using different laboratory instrumentation and techniques. We are developing a knowledge retrieval environment so a scientist can retrieve scans from disparate information sources and types of imaging, taking those very differences into account."
Also supporting BIRN are three Calit² industrial partners. Oracle Corporation, Compaq Computer Corporation, and Sun Microsystems are providing database, storage, server, and computer-cluster technologies.
Participating researchers expect to begin accessing brain scans from distributed sites this time next year. "We will be able to standardize and cross-correlate data from many different imaging systems," said UCSD's Ellisman. "That will enable us to pose new, more meaningful questions and to get better answers sooner."
For more information on BIRN, go to http://birn.ncrr.nih.gov.
An article about BIRN will also appear in the upcoming Winter 2001 edition of EnVision; to read the online version of the SDSC/NPACI publication, go to http://www.npaci.edu/online/
Related Links
BIRN Home
DeGeM Layer
Living Labs