Stage 1 breast cancer tumors are small—less than 2 centimeters in diameter. The mammogram images that can save a womans life are enormous—both in the sense of their critical role in diagnosis and in the fact that they take up a massive amount of storage and hospital floor space.
Thanks to NDMA (National Digital Medical Archive), the fruit of a federally funded project by the University of Pennsylvania, hospitals and clinics such as Murray Hill Radiology are managing to move those images off their own expensive storage systems.
That means getting them off their precious wards and onto NDMAs national network storage system of digital medical images and data, available to patients, physicians, hospitals and researchers through grid computing and a shared archive.
"It saves enormous amounts," said Jill Vincente, practice administrator at Murray Hill Radiology, in New York. "A new [8TB] system [would have been] $150,000 wed have to spend in hardware to replace what we have [now]."
And thats not even counting the impact of having to provide a new location for such a system. A new system would have had to be moved into the network room, which would have had to be piped with new air conditioning and new electricity.
Last but not least, the doctors can finally have their lounge back.
"[The 4TB] system was in the doctors reading room," Vincente said. "Now their legs wont bump against it"—it being a piece of hardware that Vincente described as half the size of a refrigerator and which hummed and radiated heat with fans whirring.
NDMA was created by i3 Archive with the help of IBM. Since the federal project launched in 1999, it has grown to become a nationally coordinated system of 24 hospitals with some 1.5 million digital mammography images that help about 300 doctors diagnose, treat and find a cure for breast cancer.
NDMA collects mammographies, CT (computerized tomography) scans and MRI (magnetic resonance imaging) images from doctors, hospitals and radiology centers such as Murray Hill Radiology. The service, which links constituents across the country, helps researchers identify and target potential links to breast cancer—and, someday, to other diseases—by providing access to analytical databases and reporting systems.
In addition, i3 Archive launched MyNDMA in November. MyNDMA is a portal that gives women access to and control of their own health records and is also the name of i3 Archives wholly owned subsidiary. One month after the portals launch, nearly 1,200 women had registered profiles to manage their mammography records through the portal.
Derek Danois, president and CEO of i3 Archive, explained that the original federal project was launched because of concern around the large-scale rollout of digital imaging, especially around mammography. People were getting worried, Danois said, about the burdens all this data would impose on private institutions.
That kind of impact can have a direct impact on patient care, Danois said, which is why the government was interested in funding research for alternative storage schemes.
At the core of the archive is IBMs DB2 database. Danois said that i3 used a parallel edition that runs on a grid of DB2 servers, aided by i3s load balancing architecture. IBMs WebSphere middleware is also a component of MyNDMA, and IBMs Rational suite of development tools is used as well.
NDMA is based exclusively on xSeries servers—both at the archive end and at point-of-presence devices employed at hospitals and physicians offices—and uses IBMs FAStT (Fibre Array Storage Technology) storage.
Its a "unique architectural design" that uses all off-the-shelf components, said Danois in Berwyn, Pa. The capability to stitch the pieces together freed up the minds behind the service to concentrate on making a difference in how medical images are handled, rather than getting bogged down with creating every piece of the solution from scratch.
"Knowing that IBM had enterprise-class tool sets, I think the intellectual property contribution had to do with those who wanted to take a different approach with computing architecture because of the challenges of large image data sets," Danois said.
Specifically, the challenges of handling such large data sets are quite distinct from the needs of computational data sets that require very fast rates to get pumped through. It has more to do with the need to construct a sophisticated storage structure that could scale to handle, in the United States, about 35 million mammograms a year.
The images range in size from 40MB to 200MB per patient. Early calculations regarding funding on the part of the National Institutes of Health were that if all mammography screenings were converted to digital images, youd have to figure out how to handle 20 petabytes a year—and thats just mammography images, which is a small subset of all medical images.
But it gets worse. Mammography is unique in requiring physicians to pull back a patients prior year or two of images for signs of change. That means a system has to manipulate, distribute and pull back approximately 40 petabytes of images per year.