| Technologies designing the “futurescape” of |
|
|
|
August 2007 William Check, PhD At the June Futurescape of Pathology conference, one in a series sponsored by the CAP Foundation, the notion of profound change seemed to pervade more than just the talks. It appeared to amplify the caffeine in the morning coffee, to flavor the pastries and yogurt served at the continental breakfast, and to spice the very air circulating throughout the meeting room. In her opening remarks, former CAP president Mary Kass, MD, who is now Foundation president, set the tone by posing dramatic questions. “In 20 years are we still going to be looking through the microscope to make a diagnosis?” she wonders. “Or will the practice of pathology change so significantly that microscopes will be outmoded?” She answered her own questions: “I have a feeling we are not going to be using glass slides, we are not going to be cutting paraffin blocks, and we are not going to be looking through microscopes.” New, molecular-based methods of diagnosis are being developed, Dr. Kass noted. “Will we take part in this revolution? Or will it pass us by?” On an encouraging note, she added, “We have a tremendous fund of knowledge relating what we see in the microscope to disease outcome. We can take advantage of it if we adopt these new technologies.” In his keynote address, Bruce A. Friedman, MD, active emeritus professor of pathology at the University of Michigan Medical School, Ann Arbor, presented an extensive list of changes occurring in pathology. He defines anatomic pathology as surgical pathology, cytopathology, and autopsy. Some of his ideas are also pertinent to hematopathology, which is a hybrid of a clinical pathology subspecialty and morphologic observations. Histopathology remains largely a subjective skill based on experience in pattern matching. Even training methods reinforce this notion of histopathologic diagnosis as a qualitative and subjective skill set: Residents and fellows sit next to an “artful practitioner” who declares a lesion to be closest to another “known” lesion. Yet for the most complex lesions there may be wide variation among national experts about which diagnosis to assign. Even with these drawbacks, it is “irrefutable,” he said, that histopathology remains the gold standard for rapid generation of tissue diagnoses and will continue in that role for some time. It provides relatively inexpensive and rapid turnaround time—sections can provide a diagnosis in minutes. Even new medical imaging modalities can rarely generate definitive diagnoses, Dr. Friedman said, while molecular imaging is in an early phase. Histopathology will serve as a validation tool for these new imaging modalities. At the same time, powerful forces are re-shaping the landscape in anatomic pathology. Dr. Friedman enumerated 10 such forces, though he conceded that the number is somewhat arbitrary. Many high-level forces for change are beyond the control of practitioners, Dr. Friedman said, while others are amenable to what he called “local action.”
What shape does he predict the model might take in practice? “Personal surveillance with panels of serum biomarkers, which will be more cost-effective than imaging,” he said. “Pathologists will be at the epicenter of this revolution” in diagnosis by biomarkers.
The key question, he says: How can anatomi pathology be converted from a morphology-driven discipline to a molecular-driven discipline?
Pressure is growing for more cost-effective health care. “Large panels of biomarkers are not cheap,” Dr. Friedman cautioned, and much of the cost of these new molecular screening techniques could end up being paid out-of-pocket by the patient. On the other hand, he speculates, multiplexed biomarker testing for diagnosis and monitoring could turn out to be less expensive than imaging. Over the long term, of course, wellness monitoring and healthy lifestyles may prove to be cost-effective by avoiding complications of chronic diseases and prolonged hospital stays. Testing costs may also decrease as walk-in clinics in retail drug and discount stores begin to offer CLIA-waived tests.
In this area, the research challenge will be to develop algorithms that more accurately interpret the significance of subtle changes in the titers of complex sets of biomarkers. “Some medical experts have raised doubts about the accuracy and cost of such biomarker surveillance of wellness,” Dr. Friedman noted, “but I believe that this approach will ultimately be beneficial for better health.”
In the face of such tumultuous change, he said, “we pathologists need to adapt or become irrelevant.” In his view, the first reform will be closer integration of anatomic pathology with clinical pathology, followed by integration of both with radiology. Ronald S. Weinstein, MD, professor and head of the Department of Pathology and director of the Arizona Telemedicine Program at the University of Arizona College of Medicine, Tucson, provided an illustrative case study in the application of some of the preceding transformative trends. The title of his talk: “The ‘S’ Curve Phenomenon as It Relates to New Technologies in Anatomic Pathology.” Dr. Weinstein is recognized as the first person to use the term “telepathology,” and he did so in 1987 at CAP Foundation Conference IV (Weinstein RS, Bloom KJ, Rozek LS. Arch Pathol Lab Med. 1987;111: 646– 652). At that time he defined telepathology simply as “the practice of pathology over a long distance.” Working with breast tissue to validate the concept, he and his colleagues showed that “the pathologist’s ability to discriminate benign from malignant breast tumors is similar using either a conventional light microscope or a video monitor with approximately 1000 lines of resolution.” The adoption of groundbreaking technologies such as telepathology follows an S-curve, Dr. Weinstein showed the meeting-goers in June. On the initial horizontal lower leg of the curve are represented many years of effort and investment of funds with little improvement in performance. At some point the technology matures, and a sudden dramatic increase in performance occurs with relatively little additional input. Finally, the technology reaches a performance plateau. “You can get blindsided by a new technology,” Dr. Weinstein concludes. Applying the S curve to digital pathology, Dr. Weinstein said the first phase took place from 1967 to 1987; telepathology is now on the upward accelerating phase of the curve, in his estimation. For one sub-component of digital pathology, virtual microscopy, the beginning occurred in about 1990. This technology has rapidly entered the growth phase, Dr. Weinstein said, with about 250 Aperio systems sold to date. Dr. Weinstein’s adventures in telepathology grew out of his experience at Rush-Presbyterian-St. Luke’s Medical Center in Chicago in the 1980s. At that time he was director of the Central Pathology Laboratory for the National Bladder Cancer Program. He found many minor and major discrepancies between local laboratory readings of bladder biopsies and those by the central laboratory. Variability rates varied among labs. One possible remedy, Dr. Weinstein realized, was telepathology, accomplished by whole-slide digital imaging using an image acquired through raster scanning and disseminated via Internet. Today, this can be done with Aperio’s virtual slide scanner. Dr. Weinstein, who had by then become head of pathology at the University of Arizona, worked with a team of scientists from the university’s College of Optical Sciences, particularly Michael R. Descour, PhD, to devise their own ultra-rapid virtual slide processing system. Their aim was to increase the field of view of the light microscope. For this purpose they designed a miniaturized microscope, a digital imaging engine consisting of small arrays instead of single lenses. They created an array of lenslets on chips; each lenslet array is the size of a quarter. The system is called DMetrix DX-40 digital slide scanner. Dr. Weinstein and his colleagues have been issued 12 U.S. patents from this work. They have published this description of the operation of the array microscope (Weinstein RS, et al. Hum Pathol. 2004; 35: 1303–1314): The array microscope optics consist of a stack of three 80-element 10 ¥ 8-lenslet arrays, constituting a “lenslet array ensemble,” he said. The ensemble is positioned over a glass slide. Uniquely shaped lenses in each of the lenslet arrays, arranged perpendicular to the glass slide, constitute a single “miniaturized microscope.” A high-pixel-density image sensor is attached to the top of the ensemble. In operation, the lenslet array ensemble is transported by a motorized mechanism relative to the long axis of a glass slide. Each of the 80 miniaturized microscopes has a lateral field of view of 250 microns. The microscopes of each row of the array are offset from the microscopes in other rows. Scanning a glass slide with the array microscope produces seamless, two-dimensional image data of the entire slide, that is, a virtual slide. The introduction of virtual slides into his department, facilitated by an infusion of $1.7 million by the university, has transformed their anatomic pathology practice and medical student and resident education. Use of virtual slides has enabled the university’s University Physicians Healthcare group to establish UltraClinics (www.uphkino.org/Hospitals/UPH-HospitalatKinoCampus/UltraClinics/tabid/239/Default.aspx), which provide same-day results from breast biopsies for women who have a positive mammogram. They accomplish this by making use of telepathology to send virtual slides of the biopsy from the clinic, which is situated in a place convenient to individuals, to a central reading site. UltraClinics even offer women the opportunity to “see” an oncologist the same day by video conferencing. “Having laboratories online changes the paradigm,” Dr. Weinstein said. It also offers easier access to second opinions. He foresees a great future for this service. A survey at a breast cancer clinic revealed that 30 percent of women would pay $200 or more out of pocket for a same-day second opinion. “And will there still be a place for pathologists in walk-in clinics?” “Of course there will,” he said. Commercial telepathology services are already available through US Labs. Dennis P. O’Malley, MD, medical director of US Labs, at the outset of his talk, titled “Practical Application of Telepathology Using Morphology-Based Anatomic Pathology,” said he would be talking about the present, not the future. Like Dr. Friedman, Dr. O’Malley put emphasis on the forces impinging on pathology. Molecular and genetic testing are becoming more solidly rooted in the pathology laboratory today, he noted. Molecular methods are allowing what he called “an explosion in testing choices and capabilities for diagnosis, prognosis, and therapy.” And pathology practice today is suffused with more connectivity. Dr. O’Malley acknowledged that there are barriers to adopting these new technologies, some of which Dr. Friedman already noted. But there are also strong motivating factors. “Turnaround time is everything,” Dr. O’Malley declared, and telepathology can ideally shorten TAT. Telepathology’s other advantages include shared expertise and a concern about medicolegal issues. “Sharing information doesn’t necessarily decrease risk,” he said, “but maybe it spreads the risk.” He provided a demonstration of US Labs’ service. A pathologist express mails a slide to US Labs, where it is stained. (Perhaps the client pathologist doesn’t have easy access to a special stain, such as a cytokeratin.) One of US Labs’ 16 consultant histopathologists reads the slide and posts an interpretation on the company’s Web site. Using an individual access code and password, the client pathologist obtains the consultant’s diagnosis and checks it against the pathologist’s own diagnosis. (The client doesn’t have to be a pathologist. A health care institution can designate a technologist to make the slide and to interact with US Labs’ system. In this case, no pathologist is involved.) “Even the most discriminating pathologist would agree—these slides are readable and diagnosable virtually,” Dr. O’Malley said. “Someday you may be able to sign cases from your condo in Hawaii.” The client can also use a template on US Labs’ Web site to create customized (paperless) reports for clinicians, importing slide images from the site into the report. Dr. O’Malley called the images on the site “portable and exportable.” They are compatible with the electronic medical record. “This is not jetpack technology,” Dr. O’Malley said. “This is today.” In closing, he quoted Abraham Lincoln: “The struggle of today is not altogether for today; it is for a vast future also.” Perhaps the future of digital pathology can be seen in the recent past and present of digital radiology. What happens in radiology comes to pathology with a lag time of about 10 years, as Dr. Weinstein noted. To provide this perspective, Douglas E. Johnson, executive director of sales for NightHawk Radiology Services, spoke on “NightHawk Teleradiology Services: A Template for Pathology?” He addressed these questions: How has radiology changed with technology? What are the barriers? the limitations? What value has “telemedicine” added to radiology? NightHawk was founded in the 1990s to help radiologists with night call, “a small unproductive segment of a radiologist’s job,” Johnson said. After looking for a place that would be attractive for U.S.-trained radiologists to live, NightHawk placed its initial office in Sydney, Australia, and started with two relocated American radiologists. The company has now expanded beyond night call and has an office in Zurich, Switzerland, and several in the United States. Expansion of this service has been aided by greater adoption of DICOM, a unified digital language that allows for multi-vendor platforms and provides for easy support, as well as by greater reliability and wider availability of the Internet. Initially, instrument vendors balked at adopting a universal language, but they soon saw the futility of retaining proprietary standards. Today, “technology is outpacing the availability and training of radiologists,” Johnson said. He estimates that 75 percent of U.S. hospitals now use digital plain films. Barriers to the adoption of teleradiology included state licensure, credentialing, malpractice insurers, and perceptions of quality. Malpractice insurers are “beginning to get it,” said Johnson, who noted that “today all radiology is teleradiology,” even if images are only sent across the street to the local radiologist’s office. Acknowledgement of the equivalent quality of digital images is proceeding. NightHawk uses only American board-certified radiologists. Getting them credentialed for states other than the one in which they hold their initial license and accredited for at least some of the hospitals that subscribe to NightHawk’s services is a major task. More than 150 of NightHawk’s 400 employees are totally dedicated to licensing and credentialing radiologists. Reciprocal credentialing under JCAHO regulations has been helpful in some institutions. “It costs about $60,000 to get a new radiologist productive to read,” Johnson said. Image size is a major limitation to the service. “NightHawk is the biggest user of bandwidth in the country of Australia,” Johnson said. Data transmission can especially become a problem with the new 64-slice scanners. NightHawk uses minimal compression—only 2:1. Complicating the issue is that satellite doesn’t work well with DICOM, so much of the transmission of images occurs via DSL lines. Average TAT for emergency department patients is 17 minutes, not including the transmission of images. Radiologist training is a major activity at NightHawk; they have trained all their radiologists in cardiac imaging. In addition to technical barriers, the company must also factor in personal interactions, whether with attending physicians, patients, or family. Johnson listed several added values of the service: a better quality of life for radiologists, improved efficiency, higher revenues from improved efficiency, being able to serve underserved areas, lower overhead, and practice growth. New markets come from Americans living abroad who would prefer to be “seen” by American physicians. Even with digital transmission of virtual slides, standard morphologic pathology has serious limitations. Richard M. Levenson, MD, research director of biomedical systems at CRI (Cambridge Research and Instrumentation Inc.), discussed potential solutions in his talk, “Beyond Morphology: Spectral Imaging Applications.” In a way, anatomic pathology has progressed very little since the original microscope of van Leeuwenhoek in the early 18th century, Dr. Levenson said. “With the addition of H&E stain and just that microscope you could do 90 percent of what [surgical] pathologists do now,” he said. Given the relatively static nature of anatomic pathology, Dr. Levenson raised the possibility that, in the molecular era, the biopsy could suffer the same fate as the autopsy. A survey published in 2000 reported that a major reason why clinicians are reluctant to request an autopsy (second only to “difficulty obtaining consent from relatives”) is that advances in modern diagnostic techniques have reduced the need for it (Loughrey MB, McCluggage WG, Toner PG. Ulster Med J. 2000;69:83–89). “The same could be said of the biopsy,” Dr. Levenson said. He listed several problems with the biopsy, chief of which are that only one or two molecular markers can be evaluated per slide and that visual-based quantitation is “iffy” at best. Digital imaging can help, but it requires sophisticated software to be reliable. Dr. Levenson put forth two solutions, both of which CRI is developing: novel optical methods reveal structural proteins without special stains and multispectral imaging allows for many more labels per slide. Standard pathology slides/stains use color to see molecules—for example, collagen on H&E. But color is a subjective human perception not necessarily related to the underlying physical properties of light. “Color” and “spectral position” are not identical. For instance, the human eye will call “yellow” both the emission from a pure yellow dye and that from red plus green dyes present together. Once one looks at spectral imaging rather than simple color, the possibilities are multiplied. Spectral analysis can be done automatically, such as with CRI’s Nuance spectral imaging system, an electronically tunable spectral filter consisting of a scientific-grade cooled CCD camera with simple acquisition and analysis software. Not only can spectral imaging resolve multiple spectra from two or more dyes present simultaneously, it can resolve several shades of the same “color”—for example, three “red” dyes of different shades. For this reason, spectral analysis can co-localize multiple stains and resolve overlapping absorbers, such as ER and PR or cell cycle markers like Ki67, p21, and p27. Some techniques, such as flow cytometry and microarrays, give multiplexing information. Other methods, notably regular stained slides, give morphology. Spectral analysis gives both morphology and multiplexing from the same analysis. However, the resolving power of spectral analysis does have limits. Only a few chromophores that emit in the visible range can be used simultaneously on one tissue. Thus, fluorescence is “a necessary evil,” Dr. Levenson said. “Pathologists hate fluorescence,” he acknowledged, for several reasons: autofluorescence in fixed tissue, photobleaching, expense, need for complicated microscopes, and non-archival samples. Using a spectral analysis system such as Nuance, these problems can all be overcome, Dr. Levenson said. For optimal results, spectral imaging can be combined with machine-learning segmentation, which is an essential component to quantitative molecular imaging. The learning machine MIDAS is a “learn-by-example” tool developed at CRI that can be “trained” to recognize different tissues. For instance, in a breast biopsy, MIDAS can differentiate normal breast from ductal carcinoma from stroma containing inflammation and fat at 10¥. Dr. Levenson said these two instruments together can power a “molecular revolution” in which molecular data can be automatically collected from the right area(s) of tissue. Spectral imaging has the potential to put together a cell-by-cell snapshot of signaling and cell regulatory status, he added. It gives quantitative spatial understanding. “We do not need to abandon the structural architectural and cytological underpinnings of pathology,” he reassured attendees. In conclusion, Dr. Levenson said spectral imaging offers “a recipe for making pathology hot again.” At the University of Pennsylvania, Michael D. Feldman, MD, PhD, associate professor of pathology, is also working on spectral imaging and machine learning. His talk was titled “Beyond Morphology: Whole-Slide Imaging, Computer-Aided Detection and Other Techniques.” Dr. Feldman quipped that his approach could be called “slow cytometry.” He reassured attendees that digital imaging is not a pathologist replacement but an assistive technology to enhance pathologist functionality. The worry that pathology and morphology will go away is incorrect, he said. Assessing tumor cells for proteins is done now with immunohistochemistry. But this method is pathologist-dependent, subjective, and only semiquantitative, and it has low interobserver reproducibility. And it can assess proteins only one at a time. For clinical trials work, which Dr. Feldman does, it would be desirable to have a reliably quantitative approach for multiple analytes, such as the phosphoproteins of cell signaling pathways (Murphy DA, et al. Am J Pathol. 2006;169:1875–1885). Accomplishing this set of criteria on tissue slides requires multiparameter immunostaining, multispectral image capture, image processing to resolve individual stains based on spectra, tumor segmentation, computational identification of nuclei and cells in images, and assignment of immunostains to each nucleus or cell. Multiparameter immunostaining is limited now to no more than three dyes in the same spatial compartment. “We haven’t been able to co-localize three stains in one compartment, such as the nucleus,” Dr. Feldman says. In the future, higher-order multiplexing with fluorescence will expand this capability and “allow you to interrogate biological pathways.” He and his colleagues have been developing their own version of a learning/vision system that differs from MIDAS. It applies a Bayesian classifier model to virtual scanned slides and shows a pathologist the result; the pathologist must then interpret the image. Dr. Feldman and his coworkers have investigated whether the computer can be taught to identify prostate cancer in whole digitized slides. They use scale-based representation to increase the efficiency of detection: Each pass of the image rejects pixels and only the positive pixels are analyzed at higher scales. With this approach, increased accuracy does not lengthen execution time. They have demonstrated their system’s ability to distinguish stroma from benign glands from Gleason grade three and grade four carcinoma in prostate biopsies. It should be possible to obtain information about the presence of cell signaling markers such as p-ERK and Ki67 in nuclei pretherapy in carcinoma and then ask whether these biomarkers decrease post-therapy. “This is the future of this field,” Dr. Feldman said. “Quantum dots are not the total answer. You need to use imaging and cytometric analysis to fully understand biologic pathways at the cellular level to fully comprehend the efficacy of pathway inhibitors and how these pathways relate to tumor cell biology.” William Check is a medical writer in Wilmette, Ill. The co-sponsors of the CAP Foundation Conference were PathSync, BioImagene, Aperio, and Dako. |
|
Interest is growing in defining and executing the Early Health Model. This model, popularized by GE Healthcare and other for-profit companies, is characterized by pre-clinical/pre-symptomatic diagnosis. New methods derived from molecular biology have fueled interest in this approach. Established powers will oppose it or, perhaps better stated be reluctant to endorse it, he predicts—insurance executives, clinicians and health care professionals, and pharmaceutical companies. One reason for Big Pharma to be opposed to pre-symptomatic diagnosis, he said, is that all of the clinical trials for drugs are currently based on subjects with symptomatic disease. Redesigning trials for a new model will require major efforts and a new way of thinking.