In anatomic pathology, there’s an elephant in the living room. Everyone knows it’s there, and everyone knows how big it is, but few want to be the first to talk about it.
Ready or not, it’s the replacement of glass slides with digital slides in day-to-day use. The date is far from fixed, of course, but virtual microscopy—already in widespread use for secondary consults, education, and research—could send glass slides directly to the archival warehouse as pathologists start using digital slides to make primary diagnoses.
“It’s potentially a billion dollar a year market, and it’s about to turn like the cusp on a hockey stick,” says Dirk Soenksen, founder and CEO of Aperio Technologies Inc., Vista, Calif., referring to a popular business projection curve where slow growth is followed by a sudden steep slope of climbing sales. “The field is far enough advanced that there are probably several scanning solutions where image quality is adequate, in principle, to make primary clinical diagnoses.”
But promotion is a different matter. “Without FDA approval, nobody is permitted to promote products for that use,” he says.
Digital imaging reached a milestone last October, when Trestle Holdings Inc., Irvine, Calif., filed the first pre-market notification or 510(k) application with the Food and Drug Administration to use its digital whole-slide imaging system for primary diagnosis. Aperio, which hired a regulatory expert only a few months ago, plans to follow suit.
“Some pathologists are comfortable with it already and ask, ‘If my microscope doesn’t have to be FDA-approved, why should I worry about using a monitor?’ It’s unclear if that will be a prevailing attitude,” Soenksen says. But others may be waiting for an FDA nod. “Securing 510(k) approval for primary diagnosis will probably be a ‘gating factor’ in driving adoption.”
The concept of a digital ‘pathologist cockpit,’ where all the information
needed to make clinical decisions is at one’s fingertips, is a tantalizing
one. Imaging robots are now available that can capture a high-resolution whole-slide
image in five minutes; a typical pathology case might contain 10 such slides,
with a total file size of 2 gigabyte (GB).
But can virtual microscopy really unseat the microscope?
For the time being, one could say conventional microscopy remains in a state
of unplanned obsolescence, because virtual microscopy is still most commonly
viewed as a supplement or an enhancement to pathology practice. And it places
significant demands on storage. The file size of a typical pathology slide
image is 200 times as big as that of a typical 1-megabyte (MB) photograph.
The technology is most helpful for reports (marketing), tumor board presentations, teaching, and collaborative consultation on difficult cases, says Tom Vuke, vice president of imaging for MIS Inc., Villa Park, Ill., which makes the PAXcam system for those purposes.
Using the PAXcam, a pathologist signing out a case has the glass slide, but may solicit a second opinion from a remote colleague by supplying a real-time digital view of the slide in a live consultation, to point out the features to be used in the decisionmaking.
“The recipient is not limited to select views, but has the whole slide at his or her disposal, and has the complete set of tools on the sender’s microscope available as well. Even if pathologists are uncomfortable with basing a signed diagnosis on what can be seen on a monitor, they are quite able to render an opinion as to the likelihood of another’s diagnosis being consistent with what they see.”
When it comes to final diagnosis and sign-out, however, Vuke believes most pathologists would prefer to work with the glass slide. “There is currently no substitute for having the entire specimen available for review, as opposed to a few static images that someone else has chosen as representative.”
Even in the case of the “virtual slide,” where more of the sample is available and the pathologist may be able to zoom up on particular areas, he adds, the ability to ‘focus through’ and to subject the sample to different lighting conditions is not a readily available option—and that leaves room for an incomplete assessment.
For some, at least, the promise of telepathology has brought mainly “anticipointment”—that
feeling one gets when an over-hyped product or event falls short of expectations.
Howard Robin, MD, medical director of laboratory services at Sharp Memorial
Hospital, San Diego, was an early adopter of the technology in 2000. Armed
with a foundation grant, Sharp Memorial purchased a remote telepathology system,
including microscopes, cameras, and computers, that connected its off-site
cytopathology laboratory with the hospital laboratory.
But the system quickly began to gather dust. “We used it twice,” Dr. Robin says. “Once when we put it up and once when we showed it to the foundation people to demonstrate how we used the money.” Another system’s remote viewing station, purchased in 2002 to allow transmission of images to an on-site pathologist, fared no better. “Once again the remote telepathology was not utilized,” he says. “At least in my experience, telepathology is a technology looking for a purpose, and I haven’t found it yet in my practice.”
However, Dr. Robin does use the Ariol digital imaging system, made by Applied Imaging, San Jose, Calif., as a diagnostic tool, because of its ability to standardize and objectively evaluate breast prognostic and predictive marker reporting.
“You can use immunostains with specificity for hormone receptors, proliferative markers, and HER2 receptor expression in the breast cancer sections. The Ariol system scans the whole slide and digitizes the immunostain, and evaluates using a software program or an algorithm that calculates the percentage of the nuclei that stains for hormone receptor or calculates the number of cells that have Mid-1 expression, or completeness of HER2 receptor staining.
“If we did not have digital imaging, we would do it manually,” he says. “Just eyeball it and subjectively say 50 percent of cells are staining. The subjective impression of pathologists is difficult to standardize, while computers with digital imaging standardize the quantitative process.”
For HER2 quantitation, there is a great benefit to having an objective endpoint that’s standardized, he says. He’s happier with the quality of reports and notes that the standardized HER2 receptor by IHC correlates almost 100 percent with the FISH assay for HER2/neu gene amplification, “which is the gold standard.”
Optics and imaging hardware have improved so much, Dr. Robin notes, that the images now available have brilliant colors and stunningly high resolution. “We just had a renal biopsy conference this morning with our nephrologists observing the images on a plasma screen, and saying ‘Wow, this is incredible.’ The images are so impressive I had the nephrologists sit back because we don’t want them to fall down one of the tubules.”
But whole-slide imaging for diagnostic purposes, in his view, is not ready for prime time. “It will take more time to get pathologists away from glass slides than it did to get radiologists away from black and white films.”
Dr. Robin says he flirted with the idea of using printed images on pathology reports as a marketing tool. “My competitors use it, but we don’t. It would be time-consuming and costly, and at this point I don’t see any benefit to the physicians.”
However, James Piper, MD, chairman and medical director of laboratories at Pinnacle Health System, Harrisburg Pa., reports that such marketing-oriented use of digital images has given a giant shot in the arm to his hospital’s outreach program.
Since 2001, Pinnacle’s nine pathologists have each used a SPOT Insight digital camera (made by Diagnostic Instruments, Sterling Heights, Mich.) mounted on their microscopes, along with software to create, manipulate, and edit images. “Our IT department sets up files for storage of images, and from there they get pulled onto the surgical pathology report.”
“We did not set out to fully digitize all the microscopic slides we produce and store and transmit back and forth,” Dr. Piper says, noting that the images lack the detail to be useful for diagnosis. “Our goal was to enhance our surgical pathology reports for our client physicians and their patients, and we’ve accomplished that.”
Pinnacle’s pathologists could not be happier with the system they have, he says. In addition to routinely attaching images to surgical pathology reports, “we use imaging capabilities to prepare for conferences within the hospital, for weekly tumor board, and weekly surgical M and M.”
“It makes it a lot easier for us. We can sit at our desks, get the cases, review them, make some images to demonstrate the pathology, package them in a PowerPoint presentation, and put them on the laboratory’s Intranet site so residents and physicians can pull the images out and use them in conferences.”
It does take more time to sign out cases, because the pathologists have to compose the image, snap it, label and file it, and attach it to the report. “It probably adds close to a minute per case, so if you’re signing out 50 cases a day it could add nearly an additional hour of work per day. But we feel it’s been well worth it.” Client physicians have provided a lot of positive feedback, and the pathologists have been able to capture substantial additional outreach business for their department because of it.
The key to moving beyond glass slides is more than just a pathologist’s ability to render a diagnosis from an image, says Stephen M. Hewitt, MD, PhD, clinical investigator and chief of the Tissue Array Research Program of the National Cancer Institute. “If this were the threshold, we would already be there, as digital slide images have been on the pathology boards for a number of years now.”
He believes validation of digital imaging will require a multi-step process, demonstrating equivalency over many different types of tissue, diagnosis, and image types (H&E, special stains, and IHC).
Yet telepathology and imaging have made substantial advances just in the past two years. “In the early days, you had to wait for the pathologist to have time to make images. That was the hiccup, because within the pathology laboratory environment, we’re the most expensive person. So the development of technology where you could hand the images to the technologist to get them scanned was a significant change.”
As one of the early adopters of prototype imaging systems, NCI’s Tissue Array Research Program began using digital imaging for tissue microarrays in 2000 and now has an image archive of about 5,000, and growing.
“With the auto-loader on our instrument, you can just pop the slides in, push the green button, and walk away,” Dr. Hewitt says. “It’s vastly different than before, when the pathologist would be wandering around the screen looking for which parts to image.” In addition, instruments that used to take an hour to two hours to take an image have been displaced by machines like Aperio’s ScanScope, which takes 20 minutes for the largest scans and averages three or four minutes per image.
The memory needed to store the images is cheaper now too. “I know groups in the clinical environment that are talking about scanning every slide out of histochemistry. We’ve shifted the burden; now we can afford to store the images, but it’s a management issue of figuring out how to do it responsibly and efficiently.”
Still, he admits, “there’s a learning curve for those of us using computer screens as a means of observing our images. I do it every day and I’m good at it, but I still love a microscope. The Aperio instruments allow us now to create little 3D zones and focus up and down—but there’s nothing like taking an entire slide and making the lymphocytes twinkle.”
Katherine M. Scott, MD, assistant professor of clinical medicine and pathology at the University of Arizona, Tucson, was underwhelmed by her first experience with digital imaging when she was a resident in the mid-1990s. “I remember looking at it and saying it was labor-intensive, not intuitive, and not worth the amount of work going into it for the result.”
But the technology progressed, and by 2004, “we decided to put one of the DMetrix digital slide-scanning instruments into one of our laboratories.” The DMetrix system was co-invented by the university’s director of pathology, Ronald S. Weinstein, MD.
The UA laboratory hoped to do outreach work for small community hospitals in Arizona that lack their own pathology staff. “That six miles really does you in several times a day,” says Dr. Scott, who serves as medical director for one of these laboratories. “So we set up a quality assurance program for review of cases, with diagnostics using virtual imaging.”
The system, which has been running successfully since Jan. 1, allows her to hold ultra-rapid prostate and breast clinics from once to three times a week depending on the biopsy schedule, and to issue final reports within four hours of the biopsy. Consultations with pathologists in solo practice are also much easier. “It’s such a boon to get an answer in a matter of hours, versus packaging up the slides and waiting for a report.”
“The advantages with this system are that the software is very easy to use, and the residents have all taken to it very rapidly. It’s not as perfect as having your own microscope and being able to switch from objective to objective. But it’s user-friendly; you can look at the entire slide, then focus on smaller areas, moving around and mimicking what a human pathologist does with a microscope.”
The DMetrix imaging system now offers scanning up to 20¥ magnification, but this fall 40¥ will be available. “I cannot do cytology with the current model, so I’m looking forward to the 40¥ scan and seeing how that will do,” Dr. Scott says. “Once it’s working properly, “then I think with our tissue diagnoses we should be able to switch over, and I’d anticipate being able to use it for primary diagnoses by the end of the year.”
New research at the University of Pittsburgh, a clinical validation study for Trestle Holdings, supports the claim that digital slides live up to glass slides when it comes to primary diagnosis. Three pathologists in the study each reviewed 202 cases including about 1,800 slides, both glass and digital, to render 606 diagnoses.
According to Trestle, “no meaningful deviation or bias was introduced by utilizing digital whole-slide images in performing primary diagnoses with a 4.3 percent discrepancy rate. This compares with a 1.4 percent to 5.7 percent discrepancy rate generally associated with rendering second opinions.”
Trestle announced last October that it had used the study in its 510(k) application for FDA clearance of its whole-slide imaging system as a class II device. Along with an image capture hardware device, the system includes software applications to conduct region-of-interest detection, focal mapping, and digital image archiving and clinical workflow management. The company is hoping to qualify its technology as “substantially equivalent” to other devices used for primary clinical diagnoses.
The difficult part of virtual microscopy is not making sure the quality of the image is good enough, says Yukako Yagi, PhD, director of telepathology/director technical management at the University of Pittsburgh Medical Center. The missing link now is standardization. For example, laboratory information systems in general are poorly equipped to manage image-level information.
DICOM3 WG26 for Pathology was established in Budapest in September 2005 and officially approved in January 2006. Dr. Yagi is a member. “We’re working with many other organizations such as the CAP and international organizations. We can expect this group to make progress on pathology imaging standardization,” she says.
Pathology is not like radiology, she notes. “We cannot have exactly the same quality of images by whole-slide imaging. Even the same systems from the same vendor give different image quality by slide, by stain, or by time of day.”
Dr. Yagi thinks a new quality assurance system is needed for whole-slide imaging. One way, she suggests, is to collect images periodically from the organizations using the technology so that a pathology organization like the CAP could evaluate the image quality.
Steve Jones, manager of the Anatomic Pathology Laboratory at US Labs, Irvine, Calif., says the LabCorp subsidiary has been using virtual microscopy for several years for its immunohistochemistry tests, mostly in cancer diagnostics. “We had a digital imaging product that wasn’t particularly flexible, so about two years ago we looked at an approach to run different test algorithms, and we came up with a modularized system—a scanning system from one vendor, algorithms developed by a separate software vendor—and we provide clients with the ability to access their results.”
The system, which now runs about 250 slides a day using Aperio’s ScanScopes, has succeeded very well. “The optical system is very good, and some pathologists say they prefer to look on a screen rather than in the microscope. We can use the images as straightforward pictures as in a telepathology function. At the other end of the scale is full image analysis; we’ll image the slide, then perform our test algorithms on that slide to give it a numerical score.”
Breast prognostic imaging is the laboratory’s bread and butter, but the laboratory is branching out to colon, prostate, and other types of tumors that can be helped by having prognostic markers analyzed. “It’s very analogous to looking at any type of glass slide. Typically they are antibody-antigen reactions to identify fairly specific markers within a particular cancer, such as a receptor to prove that a cancer is hormonally sensitive and will respond to tamoxifen treatment.”
Not only have imaging hardware speed and quality increased, but the algorithms to interrogate the data are also now much more effective, he says. It was difficult, in the early days, to write algorithms to identify malignant cells. In the old-fashioned software systems, “they used to use comparisons of brown and blue areas. It was a very crude method of discriminating. But the software is now much more intelligent and intuitive, using algorithms that look at the shape and size of cells. There are a lot of criteria being applied in a much more constructive way.”
There is a certain reluctance among ‘old school’ pathologists to trust digital technology, Jones says. After scanning, it’s not necessary to physically handle the slide, and this should theoretically speed up results reporting, but many pathologists have spent their careers looking at glass slides and want the slide with the results. “So a lot of the time savings we hoped to provide the community at large haven’t been realized.”
But he thinks their attitude might fade with time. “Once people can understand there’s no alteration on color or morphology or anything, they become more accepting. And certainly the newer breed of pathologists are a lot more used to sharing images with colleagues and conducting peer reviews over video screens, so their acceptance is a lot quicker.”
Still, Jones doubts that microscopes will disappear. “Most smaller laboratories will always want to produce glass slides for their pathologists. You have to have something you can go back to at some point. And smaller laboratories wouldn’t want to spend time imaging when they can walk next door and give a slide to the pathologists. For larger facilities, we can produce an image for a client fairly quickly—but they will probably still want the slide back.”
For teaching, digital images are making pathologists’ lives much simpler, says Tim McCalmont, MD, professor of clinical pathology and dermatology and medical director of the UCSF Dermatopathology Service in San Francisco.
“In academic pathology in general, but particularly in dermatopathology, we historically used Kodachromes to teach residents or community physicians. I’ve got mountains of Kodachromes. You were endlessly shooting the images—and that was tedious in the first place—and putting them away, and then getting them back out. I used to tell people I’m not a pathologist, I’m a file clerk.”
That gave him a strong incentive to have images on a computer. “There have been digital cameras to stick on top of the microscope for eight or 10 years and I had one of the early ones, but the quality of the images was not very good, so you’d go into a meeting and they looked grainy or dim.” He continued taking standard film images because of that. “But that’s now completely gone as an issue. You can actually take better images digitally.”
His laboratory receives cases on a different basis since it is university-based and performs a lot of second opinion work. “We are going to have images in our reports soon, but only on an occasional basis, in instances in which the image shows something interesting or unusual.”
Placing images in reports has a substantial cost in professional time, because the image acquisition is done by a doctor and not done automatically. “You have to make a decision to take a picture, set it up correctly, and snap the button, and that might add an hour to your day if you’re adding pictures to many reports.”
The UCSF laboratory bought its first SPOT camera about four years ago, and that was when Dr. McCalmont was able to start making the conversion to digital images exclusively. “Now every pathologist in our group has a digital camera on his or her own scope. With our first camera it required a great deal of adjustment, by the pathologist photographer, just to get the images the way you wanted them. But the current SPOT camera pretty much is self-adjusting, so the pathologist quickly takes the picture and it’s very close to being right. A reasonably skilled computer user should be able to take the images easily, and with the SCSI connection, it just plunks them right down into the computer.”
In a few months the UCSF laboratory plans to acquire a BioImagene system that will provide images to use for reporting, educational, and Web site purposes, in automated fashion. BioImagene, Cupertino, Calif., uses a high-resolution scanner to scan an entire slide. “So rather than look at still images, you can look at the image just as you would a slide. You can start at a low magnification, then move to a higher magnification or prowl around using the guide tools on the computer to move the image.”
“All the imaging technologies we’ve used up to now require a lot of pathologist time, but this is an advance that lets the machine do the work. You can acquire images without having to push the shutter.”
Dr. McCalmont hopes to decide after a pilot program with the BioImagene system whether this type of digital imaging will be suitable for diagnosis. “It’s possible we’ll decide that diagnosing images makes it harder, and then we won’t use the technique for routine diagnostic work. But if I felt I could give better patient care more quickly and effectively using a computer with digitized images, I’d give up my microscope in a heartbeat.”
The UCSF laboratory may also offer Web-based imaging. Rather than doctors getting a pathology report with an image on it, their staff could log onto the laboratory Web site and possibly link to the full digital slides. “For a lot of doctors with no training in pathology, that would just be fluff, but some of the dermatologists we serve have a great interest in details of the cases,” he says.
Is it still true that the additional labor required to perform virtual microscopy hobbles pathologists’ workflow? Because the technology is in evolution, it may appear to be a burden on pathology, Aperio’s Soenksen admits. “This technology started in a box that people bought and purchased and put in the laboratory, and if you wanted to use it, it meant extra work.”
But once virtual microscopy is integrated into the workflow of the laboratory, it will make the process of getting a slide diagnosed and reported out faster on a digital system, he says. “Eventually it will be more efficient to use than a microscope.”
The real payoff of digital will be in managing the information it provides, Soenksen says. “If you are a pathologist, and get 10 glass slides and a written report, your only tool is to look at one small area of one slide with one magnification at a time. You put one slide under the microscope, look around for the region of interest, remember it in your brain, and mentally compare it with the next slide.”
“It’s a massively inefficient process,” compared with virtual microscopy,
which allows an information management system to show slides side by side,
integrate them with gross images, supply different magnifications, make available
any radiology that’s been done, or call up information on a patient case from
six months ago.
The NCI’s Dr. Hewitt cautions that pathology needs to confront the functional challenges posed by new digital slide technology. For example, one problem in the field today is there are no real image quality standards for compression. “Digital imaging is great, but let’s make sure the instruments can do what we expect them to do. Otherwise we’re going to run into real headaches in the future.”
“If we don’t start addressing the technical issues of digital imaging, then patient care has the risk of slipping. I’d hate to see some guy set up shop and do all his pathology by looking at a screen, not knowing the screen is not as good as a microscope, and all of a sudden he misses a positive margin.”
Pathology has to go through some of the same adoption cycles and clear some of the same hurdles that the radiology community had to go through and clear before radiologists could make diagnoses on monitors, Soenksen says.
“But pathologists are surprisingly excited about this technology, because they see all the wonderful things it can do for them, and many of them would love to use it. They have questions about how it will integrate into the laboratory, where they will store images, how big the files will be—and they need answers. But I believe any pathologist you ask will tell you the future is digital.”
And it may be more fun too. Says the University of Arizona’s Dr. Scott, “Digital imaging is a way of making our lives easier—and a way to bring back some of the fun of doing pathology.”
Anne Paxton is a writer in Seattle.