Feature Story

Digital imaging not picture perfect—yet

February 2000
Mark Uehling

You tell us. Are these dreams or nightmares? Fiction or fact?

Electron microscope images instantly available for review. Malpractice attorneys with Web sites dedicated to images from errant pathologists. Pathologists who preview images via the Internet using a system that costs nothing. Pathologists who want images in their reports and are told the cost will be as high as $550,000. Technologists in Beijing, grossly overpaid at $200 per annum, screening "digital" slides from managed care patients living in northern California.

Digital imaging for pathology is not a whirring gizmo in a Jules Verne novel. Its advance is being delayed only by the niche-stature of the information technology market for pathologists. But it was clear at the Advancing Pathology Informatics, Imaging and the Internet (APIII) conference, held in Pittsburgh last October, that digital imaging is now feasible. The caveat: It’s not as easy as setting up a new iMac.

"The major downside to this technology is the fiscal implications," says J. Mark Tuthill, MD, assistant professor of pathology at the University of Vermont. "This technology moves so fast that if you move too fast, you may buy something that’s already obsolete. This is where expertise comes in. Don’t go it alone."

Dr. Tuthill does not believe that pathologists need to become code-writing hackers or photographers. Instead, he says, physicians must only understand their own computational needs and know enough to talk to the information systems people who build digital-imaging solutions.

There’s more good news. Digital imaging is a fairly typical technological advance for which huge cost savings theoretically are possible. This savings comes not just from processing film, but from storage and education, and consumables and turnaround time. As with computers in general, the field is also benefiting from plummeting prices for key components—in particular, digital cameras, which once cost as much as $20,000—which intensifies interest in the topic.

The complicating factor is that, for now, incorporating digital images into an existing medical practice or laboratory information system appears to be about as easy as transforming a 1976 Ford Pinto into a 2000 Formula One racer. No doubt it can be done. But it’s still rare enough that one needs to pay attention to the early adopters—those happiest out in the high-tech garage. The recent APIII meeting had such individuals in spades.

Two such individuals are Alvin Marcelo, MD, and Paul Fontelo, MD, MPH, of the National Library of Medicine, who, fittingly, offered an electronically posted abstract.

Dr. Marcelo, a surgeon, and Dr. Fontelo, a pathologist, are part of the NLM’s Office of High Performance Computing and Communications. They decided to test whether compressed digital images were as good as noncompressed digital images. Why make a distinction? Because compressed images can be sent across the Internet—even between countries—far more quickly, especially if large numbers of images are involved.

Besides changing the way the image is created, the JPEG compression process shrinks the image files from a megabyte to one-tenth that size. For their paper, Drs. Marcelo and Fontelo asked colleagues in Chicago and Cincinnati to read representative shots of the same case at different magnifications. The investigators then compared how the readers fared with the compressed JPEGs versus the noncompressed BMP images. (JPEG files are inexpensive to create, manipulate, and archive, even on slow, archaic computer networks.)

The results were surprising. "There are no significant differences between noncompressed BMP and compressed JPEG images in areas of acceptability for diagnosis, accuracy of diagnosis, confidence level, and overall image quality," Dr. Marcelo and his coauthors wrote in their abstract.

That means low-paid physicians in other countries could acquire images from the United States quickly, diagnose them properly, and relay their findings back to the United States via e-mail. Once the legal issues were resolved-and this is nothing more than forward-looking speculation-there would be no scientific or clinical reason why a well-trained pathologist in Moscow could not stare deeply into a 14-inch computer screen and do just as good a job as a colleague in Peoria. The time difference between the two countries might even be an advantage, allowing 24-hour coverage.

Indeed, programmers in India already do large amounts of high-level work for U.S. corporations, and it may be wishful thinking to presume that cost-obsessed U.S. insurance companies would never hire similarly bargain-priced labor in other countries.

Certainly no one from the business sector who attended the APIII meeting expected pathology to be federally exempted from the digital winds blowing across change-resistant industries such as finance, travel, or retail shopping.

But Dr. Marcelo soberly refuses to be drawn into any such hypothetical scenarios. He sticks closely to the facts: "JPEG would be an acceptable format for exchanging pathology images. Before this paper, there had been no randomized material on the use of compressed images in pathology, [though] there have been a lot in radiology." This paper, he told conference attendees, is just one step in the process of rigorously ensuring telepathology won’t create more problems than it solves. Says Dr. Marcelo: "Telepathology is a developing science; its implementation, at present, is best left to local advocates."

Two strong local advocates were present in Pittsburgh, as it happened, with dramatically different stories about the nascence of digital imaging. At the risk of understatement, both have colleagues less comfortable with computers than themselves. Neither reported banishing film or printers from their practices. Nor were they willing to say digital imaging is as easy as logging onto America Online. But glimmers of the future may be buried in the remarks of both.

Eric Schubert, MD, staff pathologist at Memorial Hospital, Chattanooga, Tenn., gave presentations on two successive days-and by his own admission, at the time of his comments in Pittsburgh, his colleagues in Tennessee had not quite given him the go-ahead. The reason for that soon became clear: This is not easy.

Dr. Schubert is part of a group of five pathologists who handle 30,000 cytologies and 40,000 clinical specimens annually. "We’re overworked," he bluntly states. "That’s why I’m trying to put this system in place. I’m losing a lot of hair, gaining a lot of weight, and I can’t remember my kids’ names. That’s where I’m at." He is also chairman of his hospital’s information systems steering committee, however, and well-positioned to see what is happening as other specialties wrestle with the digital future.

Dr. Schubert thinks the exclusion of images from pathology reports is curious. "Ever see a pathology textbook without any images in it? I haven’t. This is a very visual field," he says. As he and others point out, many photographic costs, for the moment invisible because they are embedded in a variety of personal and group budgets, will inevitably be wrung out of the medical dollar once digital imaging is fully adopted in a decade or two. At one academic center, Dr. Schubert recalls, his colleagues estimated they were spending $30,000 to $60,000 annually on film processing for medical meetings alone.

Says Dr. Schubert: "The cost I quoted didn’t even include storage, personnel, or camera costs-just film and processing. So in the very beginning of our project, we got seed money from the hospital based just on that argument alone. And then there is the obvious argument that Kodachromes are hard to catalog, keep track of, are easily lost, and are expensive to duplicate."

But his first forays to save money by going digital were not promising. He slowly worked his way through the list of laboratory information systems vendors published annually in CAP TODAY and found, to his chagrin, that not all could incorporate images into reports. He discovered that some were not knocking themselves out to bring information systems for pathology into the 21st century. He recounts one call. "I say, ’Hello! I’ve got a check for $77,000. Can I put images into your system?’ And the answer is, ’What?’" His audience laughs.

Dr. Schubert is quick to note that he was not considering the cheapest possible package. That might cost as little as $6,000 and include items bought not from traditional hospital vendors but from mass-market retail outlets such as Best Buy or Amazon.com. A digital camera ($300), microscope attachment ($700 to $1,300), image databasing software ($50), color inkjet printer ($300), PC/video projector for conference room ($3,000), CD-ROM writer ($400), and Zip drive ($150) might all be on his suggested shopping list for cash-strapped departments. Buying these items is painless and relatively safe, with mistakes easily corrected as better, newer models emerge.

The real trick, Dr. Schubert adds, is linking these products with the rest of a practice or hospital. Some of the pressing needs? "Multiple capture stations, probably utilizing video input [which is much faster and simpler to use on a routine basis]; image capture software that is actually user-friendly; departmental support for color printing; hospital support for image-based reporting [customized hospital information system]; and an interface between image capture/database software and the laboratory information system or, better yet, a laboratory information system that actually directly supports images for capture, database, and reporting. This last item is the current Holy Grail in this field," says Dr. Schubert. "Lots of vendors say they can do this; lots of pathologists say they want this. Yet almost no one has actually implemented it in any real way."

Dr. Schubert continues: "I made a bunch of telephone calls and ended up talking my way up the ranks of this company and talking to someone who was head or VP of research and development. Basically, I was told that if we were radiologists they might be interested in looking at this. ’The people in pathology, they don’t even look at microscopic images.’"

Dr. Schubert finally was able to identify about a half-dozen LIS vendors who he thought could probably meet his needs. However, when he budgeted in the cost of implementing a new LIS with the cost of the new imaging hardware, the total ballooned to at least $300,000 to $500,000. His colleagues were understandably reluctant. "When I presented this to my group, they said, ’Eric, you’ve done a nice job here. Now go back to work.’ Clearly, there had to be another approach because this one wasn’t flying," Dr. Schubert says.

An easily discouraged person might have given up at this point. But Dr. Schubert persevered, electing to try to assemble his own system. Basically, his approach was to use a stand-alone imaging system and mesh it into the hospital’s existing LIS. This required building interfaces to move information between the two systems and customizing software to integrate the images into the pathology reports. The total cost will exceed $200,000, but customizations have been added to further enhance laboratory workflow, such as changing the transcription platform to Microsoft Word, instituting pathologist-driven electronic signout, and setting up an automated report printing and faxing module.

The good news? "I expect to be more or less done [in terms of routinely dropping images into our reports and storing these image-enabled reports in the hospital electronic medical record] within maybe six to nine months," Dr. Schubert told CAPTODAY in December. That could be truly innovative.

Part of his victory, he notes, lies in finding a piece of software that could handle all of the imaging functions and then locating the software consultants who could connect it to his LIS. That job could be custom-programmed by the Shams Group, database consultants in Carrollton, Tex. The work of the Shams Group, in practical terms, will allow the images of Dr. Schubert and his colleagues to drop directly into the report that is sent to a referring dermatologist or internist.

Asked how fellow pathologists could locate a company such as the Shams Group, Dr. Schubert concedes he got lucky. "My introduction to the Shams Group was serendipitous," he explains. "I met them through the hospital CIO. Clearly, it helps to have good relations with the hospital information services people if you are going to do any developmental work. I would also expect that getting involved with user support groups for the major LIS vendors can lead to good contacts."

Consultants for any ancient or cutting-edge LIS system may be hard to find, he observes, but they do exist on the periphery of the ecosystem surrounding any vendor. You’re not alone.

But has Dr. Schubert’s path been so pocked with potholes and strewn with scary prices that it will not be followed by less adventuresome souls? He admits one politely insistent physician needs to take the lead, and he concedes that individual could be just about anyone.

"Progress on any front is always driven by, and largely dependent on, the perseverance of one doc who leads the charge," he says. "If that doc is in radiology, then the hospital will implement PACS in radiology first. If that doc is a cardiologist, then cardiology imaging will lead. It is a lot easier to generate support for a radiology system than a pathology system," he adds, "simply because most clinicians are interested in seeing the radiograph. Microscopic images tend to be seen as somewhat less relevant unless you’re dealing with oncologists. In my environment, the hospital-based radiology group is behind us, but the outpatient radiology group is ahead."

Another expert on hand in Pittsburgh was Dr. Tuthill of the University of Vermont. At 37, Dr. Tuthill finished a fellowship in pathology informatics last July, but he became comfortable with programming in his college days, when he took a minor in computer science. As a medical student and resident, he understood that computers could ease the life of a pathologist a bit. He regards his programming as not so much a hobby but as an outgrowth of his medical training. "It comes from my realization as a medical student that all of this information you need is on a piece of paper, and it’s locked up in the basement, and we need it up on the fifth floor now," he says. "It’s all tied together. Images are just one part of it."

If Dr. Tuthill had not chosen medicine, he probably could be making a comparable living writing Perl scripts, the lingua franca of the Web masters who create and maintain Internet sites. Does every practicing pathologist with an interest in this area need to learn to write Perl scripts? Nope. "I don’t know that that is possible or desirable," he comments. "It’s not that it’s particularly hard, but it does get into the question of how much time do you want to devote to this?"

More important, he believes, is being able to serve a mediating role. "You have to be able to sit with computer scientists and technicians and tell them enough of the specifications of the system you’re trying to design, understanding enough of their lingo so they understand," he explains. "There is a role for a pathologist who has an interest in this sort of thing to stand in between pathologists and computer programmers."

Dr. Tuthill’s programming ability, however, is not incidental to what he’s done for his department. It’s central. The big picture is that he’s allowed his colleagues to see electron microscopy images immediately, as soon as they’re "scoped." The consulting pathologist gets an e-mail saying case #12345 is ready for review; the message includes a hyperlink to a secure Internet page that displays the images.

For the first stage of the project, digital and traditional 8 by 10 glossies were prepared. The future may be simpler, with only a subset of the whole batch going to photoprocessing. Instead of 20 8 by 10 glossies per case, at a huge annual cost, the pathologists will be able to print only those four or five essential to the diagnosis. Notes Dr. Tuthill: "You can cut 35,000 to 40,000 bucks off your photographic budget relatively easily. That’s just [electron microscopy]. If you look at photography in general..."

For now—this was a pilot project, with 100 electron microscopy cases logged into the system—the pathologist reviewing a batch of such images needs to consult a different database to learn more about the patient. But Dr. Tuthill and his colleagues are creating several computer programs that would allow images, text, and other pertinent clinical information to be automatically inserted into the same Web page at the same time. One immediate hurdle on the way to that destination? Security.

Dr. Tuthill is already using Microsoft e-mail and database software, which he considers hard to break into; he wants to be extremely cautious about maintaining the privacy of patients and the physicians’ reports. "We are very concerned about security issues," Dr. Tuthill notes. "We tolerate almost no clinical information outside of the actual clinical systems themselves."

The foundation of Dr. Tuthill’s project—and the reason he could finish it while handling his other responsibilities as a fellow—is that he did not need to invent the Internet. The World Wide Web not only does most of the work of storing and transferring information, but also radically simplifies the process of creating computer programs. We’re not talking about Excel or Quicken. We are talking about smaller, nimbler Internet-based programs that can be built and tested quickly on microbudgets. We’re talking, in some cases, about Perl scripts.

As Dr. Tuthill explains: "This is the major incentive behind the Internet: The cost of developing an Internet application is so much lower than developing a stand-alone application in the prior generation. If you add to that the fact that you’ve got connectivity built in, you’ve got another win, because inherently you have applications that can talk to each other. You have an inherent savings there: You do not have to program for interoperability"—that is, ensuring that Windows users in private doctors’ offices and UNIX users on campuses can take advantage of the same program.

The cost of Dr. Tuthill’s project? Nothing, because the department already owned a Dell Power Edge 2200 server to hold the images. Dr. Tuthill jokes that the University of Vermont may have turned a profit on the project because graduate students in computer science paid tuition to work on it. But he also makes a serious point. For those in academic medical centers, pathology-related Internet programming projects can be highly desirable among computer researchers because of their real-world practicality. "Within academic communities, these problems are of such interest that they are the subject matter for master’s theses, for PhD theses," he explains. "These are things that can be a part of relationships with other departments in their undergraduate universities."

Unlike Dr. Marcelo, Dr. Tuthill has no reservations about making a few predictions. "Ten years hence, getting a pathology report with images from a consulting pathologist, a consulting physician, or in the patient’s actual report will be ubiquitous. We won’t consider the fact that images weren’t always there. We’ll tell the anecdotes about how images didn’t use to be in the reports."

If that sounds rosy, it is-and it isn’t. There is also an imperative not to be left behind that Dr. Tuthill admits he is a bit hesitant to discuss. "We get these requests from our physicians and from other physicians," Dr. Tuthill continues. "They say, ’Jeez, you don’t have images in your reports? When are they going to be there?’ This is not casual. This is, in many ways, the lifeblood to the laboratory. Once images can be readily integrated into the report, there will be no going back. The appeal of images in the reports for other pathologists, for other physicians, and for the patients themselves, is of such import that once the cat gets out of the bag, there will be no going back."

For his part, Dr. Tuthill is tearing holes in the bag to help the proverbial cat get out. In a year or two, he says, the team at the University of Vermont may have ways for pathologists elsewhere to use the same system. The Vermont computer programs may be licensed; they may form the basis of a startup company. Every traditional option for campus-incubated startups is under consideration.

While tight-lipped about the details, Dr. Tuthill does say: "We are developing a couple of applications that may have market impact. We are looking at the requirements of integrating images and the textual information that would be in the pathology report. If anything, we are increasing our embrace of Internet-based technology. This is decreasing our time for development quite significantly."

One clue Dr. Tuthill does drop is that pathology may not be the only target market. "We are looking at tools that could make a difference not just in the lives of other pathologists, but other physicians, other patients, other health care systems. It’s about getting this information out to the patients’ doctors and to the patients themselves so that they can be actively involved in their disease process," he explains. "By providing good communication for doctors, we provide good information for patients. Good lab data means good patient care. That’s what our motivation is."

Mark Uehling is a freelance writer in Chicago.