Making room for a digital image view
At Dianon Systems, an anatomic pathology reference lab, digital imaging of slides is not on the cutting edge. It’s old hat. For years all of the pathologists at the Stratford, Conn., company have had digital cameras.
Dianon’s pathologists routinely zap images into their reports. Some 30 percent to 40 percent of Dianon reports contain digital images, says chief information officer Steven Clayton. Those reports, in turn, are a key marketing feature when Dianon attempts to win accounts from labs with less jazzy reports. His pathologists can annotate images to help draw attention to a specific area of interest. "They can put in text or an arrow," Clayton says. "They can say, ‘Look at this spot.’"
Dianon’s proprietary technology is about as available to competitors as the formula for Coca-Cola. Other pathologists will have to find their own imaging solutions. In the coming years, laboratorians may even have to learn digital retouching tricks on the computer, massaging and storing images just as they work with text today.
In the meantime, digital imaging for labs is inching forward on two fronts. Some tech-savvy pathologists are building their own systems by hand. Also under development: commercial hardware and software packages designed to be purchased, unwrapped, and plugged in with no headaches.
While hardly coordinated, the two efforts complement each other. The companies are eager to delve into the murky twists and turns of the pathologist’s workflow. And the imaging pioneers, in turn, know not everyone will be able to build a digital imaging system from scratch. They know the railroads often followed the routes of those brazen enough to cross the continent in covered wagons.
A case in point is Eric Schubert, MD. The intrepid Dr. Schubert is shepherding his five-pathologist group at Memorial Hospital in Chattanooga, Tenn., into the digital age. When he last recounted his pioneer odyssey in CAP TODAY, early last year, the challenge of finding vendors was paramount.
Now he has spent $100,000 on cameras and other equipment, all of it for computer pictures. He’s also budgeting $150,000 for software, some out of the box and some custom-coded.
Why so much? Because Dr. Schubert is trying to get the equipment to work within his hospital’s Meditech information system-and to create reports with images outside Meditech. Says the indefatigable Dr. Schubert: "We’re dealing with trying to get data into and out of Meditech, through an interface and into the imaging database. That’s proving difficult, but we’re making progress." In addition, they are developing automated report generation and electronic signout capabilities, as well as improved transcription and report delivery systems. "This is functionality that Meditech simply does not support," he says.
The wrinkles he’s ironing out may benefit others. The Shams Group, his consultants, will be offering the same software to other Meditech users. And Dr. Schubert’s hope is that in the next few months he will be training his colleagues to use the new system. But even he acknowledges managing electronic pictures is tricky. For one thing, he’s not just taking digital snapshots of the occasional slide.
"There are a fair number of pathologists who are doing that," Dr. Schubert says. "They are using imaging in a manual way. They bought an off-the-shelf digital camera, they take pictures, and they put them on their hard drive. The more adventurous ones are printing them out and stapling them to their reports. And that’s fine."
But Dr. Schubert did not think that approach would work with his colleagues. "That’s actually a lot of work. It gets you through your conference support. But it’s a very manual, labor-intensive process that only the truly driven and computer-literate will do. Not all the pathologists I work with are like that."
To make the process more automatic, Dr. Schubert made at least one unconventional choice for his microscopes. Rather than buy high-resolution digital still cameras (which take snapshots), Dr. Schubert chose lower-resolution digital video cameras (which take movies of family vacations). Why? Because still cameras require several cumbersome steps to drop a selected image into a report.
With the system he is designing, Dr. Schubert explains, "You have a live image on your monitor that you don’t have with digital. You’re at your scope, you’re focusing, you’re changing your lighting. And then the act of capturing it grabs it and puts it into the database. It’s a one-step deal."
He has done his homework. "We found essentially an off-the-shelf imaging system made by TriPath Imaging," says Dr. Schubert. "They have a piece of software called the Autocyte Image Management System, or AIMS. AIMS is a system that Dynamic Healthcare Technologies is incorporating to be the imaging module of CoPath. What we’re doing is essentially the same thing, using the same imaging platform, but we’re doing it for Meditech."
Ultimately, the project will give Dr. Schubert and his colleagues a searchable repository of images. In technical lingo, it’s a database. It will contain not only digital images but also the practice’s written reports. So if Dr. Schubert needs to look at every DCIS he has ever photographed digitally, he will be able to. In seconds. Searching electronic case records-visual or textual-should be easy even for computer-phobic colleagues.
The tradeoff? Images that are unsuitable for submitting to JAMA. "These are certainly not publishing quality," Dr. Schubert says. "As far as diagnostic quality, sometimes yes, sometimes no." But the images are still of sufficiently high resolution to be able to pass along to any surgeon or oncologist.
Knowing the needs of those physicians was critical. "We don’t need a high-resolution image to enhance our reporting," says Dr. Schubert. "I need something that is fast, that is economical, that interferes minimally with our workflow but that achieves the goals of enhanced reporting."
Part of the reason the images are necessary in reports, Dr. Schubert allows, is that his partners want them for defensive marketing. He uses an agricultural analogy: "In the old days, there were a lot of family farms. Now you have factory farms that are half the size of Oklahoma. The family farmers are not, in fact, doing as well as they used to be.
"There is a similar progression in pathology," Dr. Schubert continues. "The bigger private labs out there try to get into local markets by having enhanced reporting. For us, this is a way of saying, ’You don’t have to look beyond us. We can deliver this level of service as well.’"
But Dr. Schubert, like many other pathologists, is also looking toward digital imaging as a tool to bolster a patient’s faith in the quality of a diagnosis, much as x-rays or ultrasound can help a patient understand his or her condition. "It wouldn’t surprise me for some doctors to use these images as teaching aids to explain to their patients what’s happening to them," says Dr. Schubert. "To say, ’Here’s a picture of the colon that we removed, and you can see there is a mass here.’"
Digital imaging is not just about the whiz-bang technology, Dr. Schubert suggests. "A friend of mine had a tumor diagnosed about 10 years ago," he relates. "It was hard to come to grips with this tumor because he could never see it. It frustrated him. It’s important for the clinicians we work with to have confidence in what we’re saying. And it’s important for the patients to have confidence in the clinicians. By providing this extra level of information, it helps everyone come to grips with these life-altering diagnoses."
A fellow pioneer is Steve McClain, MD, director of dermatopathology at Albert Einstein College of Medicine at Montefiore Medical Center in the Bronx. Dr. McClain, to be clear, does not present himself as another Edison: "Increasingly, pathologists are putting images in reports. There are three derm-path labs that I know of that are doing it. And there are a couple labs focused on prostate that are doing it."
Nor is Dr. McClain hesitant about confessing a source of inspiration: the supermarket. He recognized three years ago that part of the hassle of digital imaging in a laboratory was pairing up a particular case number or medical record with a particular slide. The solution, he thought, was to add bar-code labels to his slides.
He mounted a bar-code scanner and a zoom-equipped Polaroid camera to his microscope. Each of his slides has a label with a bar-code printed on it. The red crosshairs of the scanner’s laser automatically read the bar-code label as he puts the slide on the stage. The scanner even emits a tell-tale "beep," signaling to him that the slide’s data have been read successfully and triggering the camera and database to begin capturing an image. He takes an image of every case, because "it’s so easy to do and can be done in 10 to 20 seconds," he says. "I look at the slide, capture an image, and go on to the next case."
Do bar-code scanners make errors? Yes, perhaps at a rate of one in a billion. "[In a pilot project] I scanned 200,000 slides consecutively. The only errors we saw were where the scanner could not read the code," says Dr. McClain."I saw 60,000 slides last year. So one might expect an error in the next 1,600 years."
In all, Dr. McClain spent $10,000 for new hardware, cables, and computers, and $20,000 for software, generic or custom-made. Like Dr. Schubert in Tennessee, he did not find all the components he needed off the shelf. A few pieces of software-to link the scanner and his reports-were missing.
Not deterred, he simply hired a programmer-Jan Tepper of Seattle-to create a three-way marriage of his existing reports, his medical images, and his bar-code numbers. The effort was customized to how he likes to work-not to satisfy a consultant or administrator.
Much of his and Tepper’s time, in fact, was spent on making the image-capturing process easy for the pathologist. "I knew that adding work and complexity without redesigning the system might cause the error rate to go up," he says. The aim for anyone should be to "remove labor steps, to streamline things, to let the new technology do some of the things for you."
His efforts to simplify workflow and use technology to avoid errors paid off with an astonishing improvement in quality. "The corrected report rate went from one in 500 to one in 2,500," Dr. McClain reports. "Digital imaging wasn’t just marketing fluff anymore."
"Where the images go directly into the report and the text gets entered in separately, then the images can be used to proofread the report, in much the same way that a pathologist might look at a slide while signing out a report," he says. "One need not rely on memory to proofread. It is usually readily obvious to spot the infrequent error-for example, where we made a diagnosis of carcinoma but the picture is clearly that of a seborrheic keratosis."
Fewer reporting errors is not the only benefit to clinicians who receive image reports. The image reinforces the text message. "The images allow clinicians a chance at clinical-pathological correlation," Dr. McClain notes. "With physicians working harder, cranking through incredible amounts of work, some of the opportunities for C-P correlation are being lost. Imaging allows another way to communicate with physicians, in my practice, dermatologists. They can see, for example, a shave biopsy where carcinoma extends to the base." Another benefit: Resident teaching is made easier with digital images captured for reports.
Of course, not every pathologist has the interest or the patience to follow in the footsteps of Drs. Schubert and McClain. At least three companies are trying to fill the gap.
One contender is Tamtron Corp., based in Silicon Valley. The company’s PowerPath 2000 report-writing software for anatomic pathology has been installed in more than 300 facilities, says company president and chief executive Steve Tablak. The program already permits users who possess images to incorporate them into reports.
But Tablak concedes few users have the equipment or confidence to obtain the images in the first place. Thus Tamtron is planning a solution for people who don’t know a TIFF from a JPEG.
Citing the product’s ongoing development, Tablak cannot say when the company’s product might be ready, how much it might cost, or what hardware or software components it might include. Company policy dictates that the information be released first to Tamtron’s customers. But he suggests Tamtron might recommend cameras or storage devices that would work fluidly together-and with larger information systems.
More specifically, he imagines the image-capture process being hassle-free. Says Tablak, citing a hypothetical pathologist: "They need to be sitting at their microscope, and in one or two clicks, say, ’capture that’ and have whatever is in the field of vision of the microscope automatically become part of that case."
It sounds great. But Tablak rapidly frames a list of questions that may prove harder to answer. Is it pointless to fax reports with digital images? After all, in a fax, such images will be unreadable dark rectangles. Should the size or quality of images be constrained? If not, printed or Internet delivery of color images to several hundred colleagues could overwhelm the printers or computer networks of many labs.
But one of the most surprising revelations in Tablak’s discussions with customers is that few seem to be enamored with Kodachrome. "There is a practical understanding of the advantage of not having to move the film around an institution, tracking and holding and managing it," he says. Teaching institutions, he predicts, will be the first to demote film and bet on digital pictures.
"Over the long run," Tablak says, "digital imaging is going to emerge as a natural part of the information system." One of the few hints he does drop is that his ideal solution might have the ability to create images of gross specimens.
"The gross images are the most practical images for single field capture," Tablak suggests. "After you embed the specimen into blocks, and cut it onto the slide, you no longer actually have the gross specimen in its original form. It could be helpful to the pathologist to see what the complex specimens look like."
Another company, a few steps further along, is InterScope Technologies. This Pittsburgh startup, like so many young ventures that have secured an initial round of venture capital financing, is now raising additional funds to commercialize its automated, high-throughput digital system.
As executive vice president and co-founder David Sholehvar, MD, recounts the story, he and two professors of pathology (Michael Becich, MD, PhD, and John Gilbertson, MD, of the University of Pittsburgh Medical Center) started the company. They all felt pathology and the biomedical marketplace needed something fast, something simple, something digital. As they saw it, the solution would be integrated into the workflow and serve images to the pathologist. Explains Dr. Sholehvar: "We decided to build an entirely new device. We really tried to get away from the idea of automated microscopes, or cameras on top of microscopes."
"We also wanted to get away from the notion that pathologists needed to capture images, that pathologists need to be involved," Dr. Sholehvar says. "Our objective was to develop an ultrafast slide digitizer modeled after radiology systems." The company calls its revolutionary black box Abintra.
The device is based on a proprietary hardware platform and a montage image-capture process that identifies where tissue is on a slide and directs the capture to only those areas. Because of this, and other patented image-capture processes, whole slide images are captured in minutes rather than hours. The system then creates a multi-resolution image, stores it, and presents it in a comprehensive software wrapper via the Web. "This software includes functionality found in radiology PACS systems to improve the pathologist’s signout and allow integration of images with text," Dr. Sholehvar says. "And the multi-resolution imaging process results in more magnification options than traditional imaging solutions," giving pathologists more flexibility in how they view and navigate the images.
Dr. Sholehvar says more than 20 laboratories have expressed interest, and a commitment from the first customer is in.
But like the archetypal entrepreneurs in a computer startup, the Pittsburgh doctors are already striving for performance enhancements on their system. As Dr. Sholehvar explains, "Our prototype device, depending on the size of the tissue, takes less than 10 minutes to image an entire slide. We plan to make it better-to get to real-time imaging of pathology slides. We plan to get to sub-minute imaging times per slide."
The truly far-reaching part, for InterScope, is how such a device could affect the day-to-day practice of pathology. There’s ample evidence of unexpected benefits having hit radiology as digital imaging swept through that specialty, Dr. Sholehvar notes. It’s hard to know exactly how pathology will change. But he and his partners already spy glimmers of the future when they demonstrate their prototype.
"There are a number of things you can do now that quite frankly radiologists take for granted in their digital environment that pathologists never knew were really possible until we showed them the demo," says Dr. Sholehvar.
He continues, citing the questions posed by those who have seen Abintra. "’You mean I could get access to all my historical cases without calling anybody or getting the slides myself? You mean I can develop a conference in minutes, without having to have somebody go pull 10 cases and 30 slides and clinical reports? I can compare two cases at one time? I can put a case on a colleague’s or consultant’s work list in a nonintrusive way, instead of bundling it in a courier slip or walking it down the hall?’"
Dr. Sholehvar says the company is tentatively planning to sell Abintra for $100,000, plus software licenses. Based on its early analysis showing reduced hard costs and increased productivity, the company predicts Abintra will pay for itself. That could prove particularly true if buyers elect to purchase an adjunct product, the InterScope robot, to batch-load hundreds of glass slides into Abintra.
At Bacus Laboratories, in Lombard, Ill., founder and president James W. Bacus, PhD-like all the other executives interviewed for this article-tips his inventor’s cap to the work of the do-it-yourself pioneers like Drs. Schubert and McClain. Unlike them or his corporate competitors, however, Dr. Bacus has already sold 20 of his BLISS systems. The technology costs $120,000 for a microscope/scanner and $5,000 for a server to deliver images. Dr. Bacus clearly occupies the high end of the digital imaging market.
In the conceptual stages, Dr. Bacus had PACS, digital imaging for radiology, in the back of his mind. But he points out the irony that 11- by 17-inch radiographic images contain much less information than tiny slides. Pathologists’ penchant for zooming in and out, examining cells under different powers and levels of illumination, makes creating a practical digital version of a slide a technical hurdle of the highest order.
"The image content on a microscope slide is just immense," notes Dr. Bacus. "It’s an amount of data that we just haven’t had to acquire-and not just one slide, but in multiple slides. We’ve needed advances in storage capability, but they’re here," he says, citing the hard disks capable of swallowing gigabyte after gigabyte without complaint.
As politely and diplomatically as possible, Dr. Bacus says the time has come to move beyond the slide. "It’s archaic," he complains. "The technology has to move into the 21st century. What we’re really talking about is replacing that glass slide with a virtual slide that is for all intents and purposes equivalent to looking at the glass slide."
The centerpiece of his solution is a motorized, computer-controlled microscope stage. With thousands of tiny movements, the stage can divide an entire slide or an indicated portion of a slide into thousands of tiny quadrants, or image tiles. Each area of interest is scanned and separately stored on a special computer. When that process is finished, and the pathologist directs the BLISS system to show a certain field of view at 40¥, the computer quickly hunts through its stack of tiles and presents what the pathologist requested.
The BLISS system was designed to work over the Internet from the start, not as an add-on. Telepathology would be hard to simplify further. To cite one clever feature, the small size of the BLISS image tiles keeps the load on computer networks as low as possible.
Two medical schools (Michigan State and Loyola University in Chicago) are already using the BLISS technology to teach medical students. One selling point: Every student can see the exact same image. Says Dr. Bacus: "The world now is coping with small biopsies. You can’t get 100 slices of these prostate biopsies."
Not all the training is for youngsters. Dr. Bacus adds that the CAP will be selling CDs with his tiled images for continuing medical education. Other early BLISS customers include Mayo Clinic, Stanford, Johns Hopkins, the University of Michigan, and the Armed Forces Institute of Pathology. It’s an impressive list.
Dr. Bacus is softspoken and a bit apologetic about charging $120,000 for the scanning microscope. (Some of his sophisticated viewing software, he notes, can be downloaded for free on the Internet.) But Dr. Bacus expects that advances in the rest of the computing industry are unlikely to bring the cost of BLISS down as readily as they have the prices in the consumer computer trade, because the main costs are in the optical and mechanical components.
The ability to divide a slide into so many quadrants, he states, requires extremely precise instruments. "You have to be at a resolution that is, oh, less than micrometers," says Dr. Bacus. "You have to be better than the optical resolution of light."
Beyond the price of his system, one drawback may be the time required to scan a slide. To digitize an entire slide for teaching purposes at very high magnification may take more than an hour, Dr. Bacus reports. However, to scan a selected region of special interest at 40¥, perhaps for a second opinion, might take only 15 or 20 minutes. The time depends on how much you need to see, how many tiles you want to divide the slide into, and the magnification or optical resolution required.
In his defense, Dr. Bacus is quick to explain the computer can handle the scanning chore by itself. The pathologist can attend to other matters. Once the input is complete, assuming a relatively fast computer network, looking at one of his digital WebSlides is no more cumbersome than looking at a traditional slide.
And getting a WebSlide in an hour or so via the BLISS system is considerably cheaper and quicker than waiting for FedEx. Dr. Bacus speaks of one physician reviewing the slides. "Those slides are scanned in the U.S.," says Dr. Bacus. "He’s in Italy. He doesn’t care about the time needed for scanning. He does care about very high optical resolution image quality, however."
For all the fancy filmless cameras, for all the scanners and servers, the debate about digital imaging is not really about resolutions or pixels. It is about the future of the slide. Should it be relegated to one of many intermediate steps in the workflow, as staining is today? Or will it eventually be junked altogether, like the leech, the buggy whip, and the typewriter?
Younger pathologists may shape the answer. Some young pathologists are already at ease with digital images. Jimmie Stewart III, MD, a cytopathology fellow at the University of Wisconsin-Madison, has watched the prices of professional-quality digital cameras plummet from $13,000 and higher to $1,000 to $2,000. Yet that is not the most spell-binding tale he has to tell.
His most shocking discovery is that pathologists of all ages can in fact be trained to transform what amounts to raw data into dazzling diagnostic images. Even mastering one of computerdom’s trickiest pieces of software-a program called Photoshop-is within the grasp of the pathologist in middle age.
"Granted," Dr. Stewart says, "as far as older faculty members, when you talk about Photoshop and start mentioning PICT files and JPEGs, people’s eyes tend to cross. [Retouching] the first image is the hardest. The first one is the one where everyone moans and groans."
Dr. Stewart continues, describing his own workshops in teaching photo-editing at scientific conferences and more informal tutoring sessions. "I taught three or four of these teaching sessions, and afterward the faculty, older and younger, seemed to grasp it," he says.
"There were some people who had problems. But if you sit down and work with an image, and use terms that most pathologists understand, it’s not as hard as it seems. It becomes an easy and intuitive process."
Mark Uehling is a freelance writer in Chicago. To reach Dr. McClain, send e-mail to firstname.lastname@example.org. Dr. Schubert is at email@example.com.