CHICAGO - Few would dispute the Internet's potential for revolutionizing image management, and ultimately patient care. RSNA '98 saw the first crop of truly practical Web tools, and the show fairly buzzed with Web talk. The evolution continues this year with a new generation of products that enable better manipulation, archiving, and management of Web-based medical images.
Yet for all its potential, Web-based image transfer has yet to achieve widespread clinical usage. Early concerns about security, archiving, and administration have largely been solved, but bandwidth, or rather the lack of it, remains a big stumbling block.
Start-up solutions
Two Northern California companies, Stentor of South San Francisco and RealTimeImage of San Mateo, have developed Web-based imaging systems that use limited bandwidth more intelligently. The products are similar in several respects: Both rely on smart server software and raw computing power to manipulate image data. And both ration bandwidth by giving the user small amounts of image data only as it's needed.
Stentor's iSyntax technology is currently in beta testing, with commercial release scheduled for March 2000 pending approval of the 510(k) application the company submitted in August. iSyntax runs Active-X on a Windows NT server, and is designed for use on standard PCs connected to a hospital Ethernet. Future versions of the product will operate on slower LAN or dial-up Internet connections.
Based on technology developed at the University of Pittsburgh that was a highlight of the 1998 RSNA show, iSyntax is designed to balance a hospital's network load and eliminate the data bottlenecks that can occur when large image files are sent over the network.
The limits of perceptionStentor president Oran Muduroglu says that limitations in human perception make the technology possible.
"There's a very real limit to the rate at which a human being can perceive data," he said. "If you look at the PC, the room is not in focus. If you look at the room, the PC is not in focus. Human beings can capture data at (only) about 7-8 MB per second."
The goal, then, was to limit the display, and thus bandwidth use, to the maximum amount users can perceive and monitors can show. iSyntax uses wavelet compression to create several copies of an image, each drawing on data from the previous image as a starting point, then adding more detail. When a file is opened for viewing, the server sends the PC just enough data to maximize resolution based on image size, monitor size, and the monitor's resolution.
Data continues to load in the background while the original picture is on the screen. When the user zooms in for a closer look, the server processes new coefficients to deliver a more detailed view of just the image region selected. iSyntax eliminates the need for the network to deliver the mountain of data required to show the whole image in full detail (which couldn't be seen on the monitor anyway), and eliminates the need for the receiving computer to process it.
Muduroglu said the system can provide lossless real-time imaging at network speeds of 1 MB or greater, but said that versions being developed for lower bandwidths will necessarily be lossy.
At the RSNA show the company is demonstrating software for a companion product, iSite, that can function at data speeds as low as 400 Kbps, a speed that can be achieved with a fast DSL line. iSite will incorporate lossy wavelet compression technology, Muduroglu said. Since the technology is an internal server process, it operates separately from, and is compliant with, DICOM, TCP/IP, and the new JPEG 2000 standards being developed, he said.
A different approachRealTimeImage takes a different tack to bandwidth economy with its Pixels-on-Demand technology, originally developed to facilitate the transfer of large image files in the graphic arts industry.
Like Stentor's products, Pixels-on-Demand uses progressive wavelet algorithms to create a compact mathematical representation of image data. It then calculates and sends the PC just enough image pixels to maximize image resolution based on three factors: image size, monitor size, and monitor resolution.
Unlike iSyntax, however, Pixels-on-Demand does not create multiple archives of varying resolution for each image, and the current version does not load image data in the background before additional image data is requested, said Gene Rubel, the company's vice president of marketing. When the user clicks on an image, the server delivers the highest resolution image that will fit on the monitor.
"It could be one or ten clicks," Rubel said. "It's totally dynamic. Each time you click we fill up a screen of data. If your screen is bigger you get more data; if your screen has more resolution you get more data."
RTI's RSNA exhibitAt the RSNA show this week, the company is introducing a new server for the system. Called iPACS, the server runs Windows NT on 64-bit Intel hardware the company says is much faster than previous 32-bit processors. The server will be used in clinical trials of the product set to begin next month in Israel.
Rubel is enthusiastic about the market for Web-based imaging products, and will aim the company's marketing efforts on that segment rather than hospital-based networks, where he says it would have to compete with other methods of image distribution.
"Our claim to fame is our technology, our development," says Rubel. "I think the real breakthrough here is not inside the hospital but for the doctor at his home or office being able to access these images."
By Eric BarnesAuntMinnie.com staff writer
November 30, 1999
