Oh, what a tangled Web we weave when first we practice to retrieve — images, that is, without a few basic elements: sufficient bandwidth and reliable performance, to name two.
The fact is that even though the Web is ubiquitous, its widespread use in radiology has yet to materialize. Interest is growing, but definitive numbers on how many radiologists use the Internet to store, retrieve and transmit images are mere guesses at this point. The best estimates are in single-digit percentages.
As the accessibility of bandwidth increases, as trust in encryption grows and as performance improves, the future of the Internet for storing, retrieving and transmitting images looks brighter with every technological improvement.
Web servers are delivering images in a timely and inexpensive way, and Web use is expected to grow. Industry observers predict that as radiology departments work to become more integrated and as healthcare drives greater integration between clinical departments, the Web will serve as a tool for both delivery and integration between departments — facilitating better care around the heathcare enterprise.
A computer screen near you
I think we’re pretty much at a juncture now that [Web use in radiology] is going to happen rapidly,” says Steve Munie, PACS administrator for Wade Radiology Associates (Raleigh, N.C.). “The Web server products from the vendors have matured within the past year to a point where they’re good, solid products.”
Wade Radiology Associates is comprised of seven centers, 40 radiologists and more than 1,000 referring physicians, not all of which are on Wade’s intranet. Munie says an Internet connection would be the easiest to implement for the group. He considers lower cost, both for initial investment and near-zero support and maintenance, as well as reduced overhead, important incentives for using the Web.
“The DICOM protocol introduces significant overhead in image transmission,” says Munie. “Using the Web will bypass the DICOM negotiations and allow for faster and simpler image access.”
Interestingly enough, when it comes to image and information systems, companies such as Mitra Imaging Inc. (Waterloo, Ontario, Canada), a service solutions provider of integrated image and information systems, and others do not see radiologists as their target customers. “Where we see [the market for the Web] is the rest of the clinicians in the hospital,” says Jeffrey Hendrikse, Mitra’s director of marketing. “We really look at the Web much more as a tool for clinicians to have the images, whether they’re in the operating room or the ICU. We look at this as a low-cost, high-quality mechanism for serving their needs.”
Mitra’s Web solutions are integrated into a PACS environment and into the workflow of a radiology department. “A whole lot of radiologists we deal with don’t mind using the Web, because it’s certainly better than driving into the hospital at 2:00 in the morning,” explains Hendrikse. “Still, their preferred reading environment is on a larger-resolution, high-brightness monitor that might be typically connected to the PACS.”
For the initial reads in the office, they prefer using the dedicated station with the larger, brighter monitor, rather than the Web. “The Web does not give them the tools or image clarity because of the inherent nature or limitations the Web can do through a browser,” says Hendrikse. “So, from that perspective, we don’t see PACS going away. We’re always going to need those core elements, not the least of which is years of core data [for longer-term storage].”
Serving it up
When American Radiology Services (Baltimore) was formed in 1997, four radiologist groups came together. All but one of the groups provided some kind of on-call coverage for hospitals they served. Some systems were old and required sending a technician to a radiologist’s home to install software, an upgrade or to fix the system. Elimination of technician costs was part of the motivation to look for a low-cost way to install a teleradiology system in radiologists’ homes and, at the same time, provide “high-quality service,” according to CIO George Bowers.
“We looked for a product that was Web-based,” he added, “with the idea that as long as the radiologist had a Web browser at home, such as Netscape or Internet Explorer, he or she would be able to download the necessary plug-in from the Web site and any updates could be delivered that way, as opposed to having techs running out there all the time,” says Bowers.
Mitra’s Web solutions are integrated into a PACS environment and into the workflow of a radiology department.
They also wanted to lay the foundation for eventually enabling referring physicians to use the Web to access images. Their radiologists currently are using the Web for on-call services, primarily in the evenings and weekends.
“We have found that it works well,” says Bowers. “There have been some problems with it, particularly in areas of the state where there is an inadequate wiring infrastructure. If radiologists cannot get a high-speed Internet connection to their home, they find that the Web-based product is a little slow and cumbersome for reading images.”
They have compensated for that, however. Where high-speed telephone line connections aren’t available, they have installed satellite dishes at a number of homes.
The radiologists use 17-inch monitors — although some use laptops with 14-inch screens — for Web reads. The final diagnostic reads usually are done the next day off film. The images transmitted on the Web server are compressed.
“Depending on the modality, they’re all within acceptable ranges of compression based upon the American College of Radiology (ACR) guidelines,” says Bowers. “At the same time, we have reminded the radiologists that there is a risk with compression and to look extra hard and if they feel uncomfortable, they should go in or have someone go in to read something.”
The vast majority of images read at night by the group are CT scans, where the resolution is not really an issue, because it is well within the range of the number of pixels on the standard workstation. Flat films, which are more problematic, normally are sent to a central reading center at one of the hospitals where they are read on a
2K by 2K diagnostic workstation. American Radiology Services also plans to expand its Internet access with images and reports to referring physicians’ offices.
Compression challenges
Although challenges with image compression remain, others believe compression has gained acceptance and credibility. “People are understanding that you still can get the diagnostic yield out of an image using various compression techniques,” explains Fred Goeringer, CTO of Radiology.com (Chantilly, Va.). “So, that opens the door for Web applications. Coming in the other direction are technologies like DSL (digital subscriber line) and ever-increasing opportunities with browser-based and Web technology. And, if you’re prudent about your integration approach with the Web application, you can build a diagnostically effective system, not only for primary interpretation, but for image referral to those customers of radiology within the enterprise and outside the enterprise.”
Radiology.com offers Web-based services to radiologists, referring physicians and patients. Small to mid-sized radiology practices or hospitals that were left out of the PACS loop because of out-of-reach costs can take advantage of the application service provider (ASP) model, a pay-as-you-go subscription approach to accessing radiological images.
This allows a radiology practice to “go digital in a hurry and use and pay for just the amount of digital services they need,” explains Goeringer. “They don’t have to have the heavy investment in archiving if [we] can archive their images for them and provide the image distribution through a Web-based approach.”
The user eliminates the burden of networking services and pays on a fee-for-use basis, at approximately $4 to $5 per exam compared to $15 per exam to manage the cost of an X-ray study over its life.
Five years ago, radiologists questioned soft-copy reading, because they were fearful that the display did not have the same spatial resolution as a hard copy X-ray film and that they would miss things.
“That fear has now gone, and we find when you’re looking at an image and some of the recent studies that have come out are saying you’re talking about not so much [whether] can you duplicate the spatial frequency resolution, but what is the overall effectiveness of the imaging,” says Tom Johnson, M.D., chief medical officer of U.S. Radiology Partners (Dallas). “What we have found in the digital world is that what you give up in spatial resolution, you more than make up in contrast resolution. Therefore, the effectiveness of the image is better, even though the resolution is slightly less.”
Johnson says that recent studies are showing that even with 30-to-1 compression, if it is done properly, there is no significant, detectable difference in diagnostic capability. “In fact, the new DICOM standard that will be coming out this year or next will have a new FDA-approved compression algorithm that allows up to 30-to-1 jpeg compression that has been proven not to change your diagnostic capability,” says Johnson.
Paul Chang, M.D., director of the University of Pittsburgh’s School of Medicine’s Division of Radiology Informatics has a philosophical problem with compression. The UPMC Health System is comprised of 20 hospitals in western Pennsylvania. It includes a wide geographic area that looked to the Internet to bind its entities together. The system needed a way to be able to provide images, but couldn’t afford to distribute $100,000 workstations throughout and supply the bandwidth these workstations require.
“We asked our vendors to provide a Web PC-based solution to provide enterprise-wide image distribution of images without compromise with respect to either image fidelity and navigation” says Chang. “The vendors have tried to respond. Almost every vendor offers some sort of Web client. The problem is those Web clients are crippled in many ways. Most of these Web clients use some sort of lossy compression, which compromises some high-frequency information.
“I agree that perhaps 90 percent of the time, you can probably get away with that,” Chang continues, “but the problem is 10 percent of the time you can’t, because the high-frequency information is where the pathology sometimes lives. The problem is that you can’t tell a priori, if you’re seeing a patient whose image represents the 90 percent where there is no significant high-frequency information or the 10 percent where there is. It’s not that I don’t ‘believe in compression’; I just don’t have the ‘guts’ to use it and risk missing something.”
The other problem, adds Chang, is that Web solutions offer very limited navigation capabilities, such as the ability to quickly review multiple CT and/or MR sequences and compare images with other projections and prior studies. “This navigation capability is just as important as image fidelity when attempting to make the diagnosis,” says Chang.
The assumption, he adds, that only radiologists need full-fidelity “diagnostic quality” images is false. “There are a large number of physicians throughout the enterprise that require full-fidelity diagnostic-quality images with full navigation capabilities; the usual compromised vendor Web offerings fall significantly short for them,” says Chang. “We radiologists must remember that there are quite a number of sophisticated users of images out there outside the radiology department. In addition, it must be noted that since [for most of us] there is a significant delay in getting a radiologist’s report to the clinician, a large percentage of physicians are making patient management decisions based on their [not the radiologist’s] interpretation of images. It doesn’t matter if we or the vendors say that the Web clients are only for ‘secondary reads’; they are in fact being used for primary diagnosis and patient care. Given this reality, compression and compromised navigation is simply unacceptable.”
Out of Chang’s frustration with “compromised Web solutions” and expensive PACS, he invented Dynamic Transfer Syntax (DTS), a just-in-time model for image data delivery and distribution. DTS, as its name implies, addresses the problem of delivering large image datasets by using a “just-in-time” or dynamic streaming model.
“For example, if you want to see a 500-slice CT exam,” says Chang, “we don’t send all 500 slices to you, since no human can look at more than one CT image at a time. Instead of globally compressing the whole dataset (which still results in a nontrivial block of data to transmit) and sending the whole dataset to the client, DTS sends you just the image (or portion of the image) that you need. Similarly, even though a CR image may be 4K x 4K (32 MB) in size, there is no such thing as a 4K monitor. So, instead of sending (compressed or noncompressed) the whole CR dataset (only to have your local client “downsample” the dataset to “fit” in on your monitor), DTS sends just what you can display on your PC monitor. When you zoom or pan, DTS sends you the incremental information necessary to display that new region in full fidelity. As long as DTS can anticipate and deliver just what you need when you need it, it seems to the user that the entire study is locally available — all without the need for lossy compression. DTS can live on an existing network and uses a regular PC as the workstation. I can leverage my existing network, and it outperforms a traditional PACS, and because it’s Web-based, I can use it within the intranet or the Internet.”
Mitra, which makes software for Agfa Corp. (Ridgefield Park, N.J.), Philips Medical Systems North America Inc. (Shelton, Conn.) and Cemax-Icon Inc. (Fremont, Calif.), recognizes that some radiologists’ views of compression vary, so the company tries to provide some flexibility.
“The way our product deals with this is allowing any sort of compression and in several different places,” says Mitra’s Hendrikse. “On the server side, the administrator can set up a compression level so all images coming in are compressed automatically. We also can do it on a per-user basis, so if one physician would like to have a higher level of compression than another physician, we are able to have those individual preferences.” They also can specify different rates of compression by modality.
The bang of bandwidth
The speed of transmitting from one place to another is a critical factor in ensuring acceptable performance of Web-based applications. “If you’re a radiologist who is accustomed to receiving your information in subsecond time, [bandwidth] is an issue,” says Milton Silva-Craig, general manager of clinical e-services at GE Medical Systems (GEMS of Waukesha, Wis.). “If you’re going over the public Internet exchange, forget it. You’re never going to get the response time. Even in a private intranet scenario, for you to get the response time, you’re going to pay quite a bit of money on the telecommunications.”
Silva-Craig sees the bandwidth issue subsiding. The bandwidth trends are positive, bigger pipes for less cost, which, he says, complement the evolvement of the Web in radiology.
Mitra’s Hendrikse sites some facilities’ lack of bandwidth as an impediment to Web use in radiology. “It is surprising how many hospitals are running 10 megabit networks,” says Hendrikse. “When you start to talk about multiple megabit networks of image data going everywhere for every physician’s viewing, that really becomes difficult.”
For the radiologist at home, trying to access images without a high-speed network, a cable modem or DSL connection could be an exercise in frustration.
It is one thing to say you’re Web-based, but another to make it a practical utility. “Doctors are basically impatient people,” says Charles Sykes, M.D., a radiologist and partner at North Fulton Regional Hospital (Roswell, Ga.). “I’m not going to sit in front of a computer screen for more than a second or two. If it delays me from doing what I normally do, then it’s of no use to me.”
Sykes is working with Medizeus (Atlanta) on developing its Web-based, computer-assisted diagnosis software to check mammography results. A radiologist digitally transmits a mammogram to Medizeus to match up results with computer-diagnosis software.
“By the time a mammogram is placed in front on me to read it, I have the benefit of the computer-assisted diagnosis read,” says Sykes. “I read the mammogram as I normally would and then refer to the computer-assisted diagnosis interpretation, which may point out something and assigns a percentage next to it [that estimates the likelihood of a malignancy]. It would compel me to do an additional work-up on that lesion, perhaps even biopsying it, where I might not have in the past.”
The results are sent via e-mail to the referring doctor who can analyze that information almost immediately. The patient could have results within hours.
Editor’s note: In the August issue, we will examine Web-related PACS, providing details on system functionality and which vendors are participating in the marketplace.
