image: Grid server catalogs anonymized DICOM files to enable secure access by geographically remote users. view more
Credit: Childrens Hospital Los Angeles
Using a new grid computing system, radiologists, physicians and pediatric oncologists at 40 hospitals all over North America are now quickly and securely exchanging high-resolution medical images.
One hoped-for result will be that the doctors of young cancer patients will know more quickly whose treatment is not working and be able to change course. Others include making second opinions from specialists anywhere easily available; and quicker, closer monitoring of ongoing clinical research and diagnostic practice.
"We have broken the medical image communication barrier," says Stephan Erberich, a computer scientist who is the Director of Functional Imaging and Biomedical Informatics at Childrens Hospital Los Angeles and a faculty member of both the USC Keck School of Medicine and the USC Viterbi School of Engineering.
He will demonstrate the Globus MEDICUS system at the upcoming annual meeting of the Radiological Society of North America (RSNA) in Chicago every day from Sunday Nov. 26 through Thursday Nov. 30.
The Globus MEDICUS project makes pediatric cancer researchers and the medical imaging profession at large the latest in the rapidly growing number of scientific and professional communities using Globus open-source grid collaboration software developed at the USC Viterbi School of Engineering's Information Sciences Institute (ISI) and Argonne National Laboratories (ANL).
Carl Kesselman and Ann Chevernak of ISI, who worked with Erberich in creating MEDICUS, built the system basing themselves directly upon earlier work by the Digital Imaging and Communication In Media (DICOM) standards committee.
DICOM created a uniform electronic format for medical images, one that allow the whole range of commercial imaging devices -- X-ray, MRI, and CT -- to display and manage images from any other.
But DICOM's potential for transparent exchange between collaborating researchers, and physicians has so far not been realized, because of technological, administrative, and security challenges of confidential patient data, according to Erberich.
As a result, access to the interchangeable data was limited to the hospital where the images are acquired -- not even available to a patient's point-of-care facility, if different, unless physically carried there.
"Today if you leave the hospital, you either leave your digitized images behind or you have to carry them on a CDROM," said Erberich. "This is not the 21st century healthcare we need in a networked society. All kinds of other fields, from banking to air travel now rely on instant information exchange and decision making online. We should be able expect the same level of sophistication in healthcare."
That day has now arrived, says the scientist. Using the DICOM Grid Interface Service (DGIS) DICOM records at medical facility anywhere are now easily accessible and exchangeable over Grid-secured Internet connections.
The MEDICUS project began when Erberich approached ISI grid experts Kesselman and Chervenak asking them "to translate DICOM into Grid," as Erberich described it.
Kesselman had, as part of the Globus project previously helped more than a dozen scientific communities ranging from high-energy physicists to earthquake simulating engineers and geologists share instruments and data, securely and easily.
He immediately saw that the need was a perfect fit for Globus open-source Grid solution. "There had to be new code developed to handle the medical-specific things like DICOM translation and patient confidentiality assurance," Kesselman said, "but the cool thing is this leverages all of the existing underlying Globus technology that we use in so many other projects."
In creating key grid components for MEDICUS, ISI research scientist Chervenak and Kesselman, who is director of the center for grid technologies at ISI and a research associate professor of computer science in the USC Viterbi School of Engineering worked with Manasee Bhandekar, a computer engineer at the USC Alfred E. Mann Institute. ISI researchers Robert Schuler, Shishir Bharathi and Gaurang Mehta also made significant contributions.
Erberich developed the DICOM to Grid interface and led the inter-disciplinary collaboration between the engineering and clinical teams, working with Childrens Hospital radiologist-in-chief and Chairman Marvin D. Nelson.
The system has been in place since September, and as Nelson describes it, "it's totally transparent. Each facility is now connected to the Grid, using its own interface -- you only have to one interface at the hospital, and that serves the whole hospital, reusing the hospital's capital investment in DICOM visualization devices."
The cost of installing a DGIS node is "trivial," said Erberich: on the order of $1000 for a Grid gateway, attached to a high-bandwidth net connection. The gateway provides two-way access to the Grid, allowing upload of local images (after de-identification) and also continuing access to a catalog of archived DICOM records. "The nice thing, " said Nelson, "if a researcher has authorization for a specific record in the catalog, it can be downloaded for use on her own image display."
One dramatic change in practice will be the ease of review. Researchers can look at observations made anywhere on the grid without leaving their offices.
"We store the images here in the Data Center, " said Erberich, "but the people who have been assigned to review images, can review them from virtually anywhere."
"Before" he continued "when we were documenting a research study, it meant that radiologists would have to physically come to a single facility and look through a file cabinet full of physical images. Now, radiologists all over the planet can look at the images at their leisure in their own offices, on their own favorite commercial medical imaging system."
One critical advantage of this is elimination of backlogs reviewing images, with potentially life-saving results for patients in studies. "We'll probably have a more timely review of scans," said Robert C. Seeger, M.D., of the Saban Research Institute of Childrens Hospital Los Angeles, a specialist in neuroblastoma who is part of one of the research groups now using the system.
Besides the 13-institution New Approaches in Neuroblastoma Therapy group (NANT.org) that Seeger is part of, the 27-member Children's Oncology Group (CURESEARCH.org) is now active.
Both the doctors and the computer scientists involved expect this number to skyrocket in coming years, because the entry cost is so low and the possibilities are only beginning to be tapped. Other advantages include:
Greatly increased ease of radiological consultation and study. Any radiologist practicing on rare or unusual conditions can now see only see the small fraction of the total cases that present in one place. Now, "he could sit in Boston and potentially review every single case, from anywhere in the country," says Seeger.
Imaging research. Scientists studying new techniques will be able to exchange samples instantly. And "we can develop expertise not just for reading, but also processing images," said Erberich.
Drug development. New techniques depend on imaging experimental animals, typically mice, using bioluminescent markers. Analysis of large bodies of such images requires great computing power. Grid techniques can both share images and the computing power necessary to extract their meaning.
The Globus Alliance is a community of organizations and individuals developing fundamental technologies behind the "Grid," which lets people share computing power, databases, instruments, and other on-line tools securely across corporate, institutional, and geographic boundaries without sacrificing local autonomy.
Grid computing work has been named one of "Ten Technologies that Will Change the World" by M.I.T. Technology Review, and has received a "Top 100" award as well as a "Most Promising New Technology" honor from R&D Magazine.
The Globus MEDICUS project was supported by the Children's Oncology Group Phase-I Consortium, NIH (grant UO1-BA97452), and the NANT Cancer Foundation.