Piggyback PACS: Pursuing Enterprise Image Management
Radiologists have become PACS experts, sometimes by default. Because diagnostic images made the greatest demands on early information systems in health care, the most sophisticated systems were first developed to handle these images and associated data. These systems became PACS, which grew out of the homegrown image-management systems of academic radiology departments to become commercial products; eventually, they became essential to the practice of radiology in hospitals and radiology practices. Because PACS grew out of radiology departments, radiologists have never been simply consumers of a PACS product in the way that banks, for example, are purchasers of financial-transaction software. Radiologists were fully involved in specifying and troubleshooting commercial PACS, even after the systems had been widely adopted, and commercial PACS offerings could not have been developed without the input of the radiologists who used the first systems. This did not mean, though, that the needs of radiologists always drove the implementation of PACS. For some institutions, the cost savings associated with eliminating film (and its libraries and associated staff) became the primary driver; the increased procedural volumes that could be handled when the film-based radiology department went digital were also powerfully attractive. The increase in the quality of care that is likely to result from PACS use is another strong motivator, since the typical PACS installation can reduce the time that elapses between image acquisition and patient-care decisions, can reduce the need to repeat examinations due to the loss of films, and can provide better access to prior studies for comparison. In addition, PACS can keep referring physicians happy by providing rapid access to images and reports. Many of the reasons to adopt PACS, therefore, were as compelling to the institution as to the radiology department. This continues to be the case—and now, the same cost, quality, and competition concerns have extended the need for PACS beyond radiology to every specialty that uses and retains images in the course of diagnosis and treatment. Because radiologists are the physicians with PACS experience, they are being consulted in the process of expanding PACS use to the enterprise level. Sometimes, they are also being expected to allow other specialties to ride piggyback on their existing PACS. Because making PACS work outside radiology involves much more than simply enlarging an archive’s capacity or adding workstations in other departments, this piggyback PACS may be an idea whose time has not yet come. According to “Enterprise Imaging,” which was presented on November 29, 2010, at the annual RSNA meeting in Chicago, Illinois, it could be too soon to expect a PACS solution that is perfect—straight off the shelf—for the entire enterprise. Because the need for the advantages of PACS is obvious to institutions, to specialists who use images, and to many of the physicians who refer patients to them, the search for the fully integrated enterprise PACS continues. As information systems increase in processing power and archival capacity, that goal comes closer to reality, but as the three presenters report, each specialty has workflow and image-handling needs that must be taken into consideration as the enterprise PACS is built. Radiologists, to preserve the integrity and function of their own PACS, should be prepared to explain to cardiologists, pathologists, and other image users that they can’t just plug their departments into the existing PACS somehow. Planning, hardware and software outlays, and workflow adjustments are all to be expected first, and a thorough understanding of how each specialty uses images is required. PACS for Cardiology Benoit Desjardins, MD, PhD, is assistant professor of radiology at the Hospital of the University of Pennsylvania in Philadelphia. He is a noninvasive cardiovascular radiologist who works closely with the hospital’s cardiology service, he reports. “I have dual appointments in cardiology and radiology, and by knowing very well what IT is in radiology, I can appreciate that IT issues in cardiology are different,” he says. “The first major difference is cardiology data: It has another level of complexity, and there is a mix of both imaging and nonimaging data.” He continues, “Cardiology information is inherently 3D in nature, in a way that’s quite different from radiology information. In radiology, we use 3D all the time in reconstruction, but it doesn’t typically give us much more information. In cardiology, though, you’re dealing with branching structures, and when you look at them in 3D, it really adds a lot of information. Cardiology imaging also makes use of color, so PACS needs to be able to handle these images with color displays.” Cardiology, however, has much in common with radiology where image management is concerned, and many cardiology exams use the same modalities involved in radiology studies. The two specialties have CT, MRI, ultrasound, and nuclear-medicine modalities (including PET, SPECT, and scintigraphy) in common; in smaller facilities, they might be sharing some or all of that equipment. These similarities have had a twofold effect on the adoption of PACS in cardiology. The broad use of digital imaging has made cardiology the easiest first step in extending a radiology PACS to another department, especially in cases where the equipment is already delivering images to the existing PACS for radiology studies. Unfortunately, their reliance on the same technologies has also meant that, in many hospitals, cardiologists and radiologists have been competing for the same cases ever since the imaging equipment was installed. This has influenced the ability of the two specialties to cooperate in planning and implementing a PACS that could serve them both. In some hospitals, the result has been the creation of an isolated PACS for each specialty; in others, cardiology has made do with less-than-PACS information systems because no agreement to let cardiologists make any use of a radiology department’s PACS could be reached. Despite the similarities between cardiology and radiology in equipment use, the piggyback PACS is not an adequate solution, Desjardins says. The primary problems, he adds, are cardiology information’s lack of integration, lack of standardization, and lack of access/distribution. For technologies not shared with radiology, Desjardins says, there are several problems that are an additional challenge for enterprise information management. Many systems are proprietary; dedicated, single-purpose workstations are frequently in use; there is little or no standardization of output (with free text being common in reporting); and the equipment is often unable to communicate with PACS or with other information systems. Desjardins describes image-management systems in cardiology as being roughly as advanced as radiology PACS were in the 1990s, so there is substantial room for improvement. Even if cardiology is the specialty most like radiology in its image-management needs, its workflow is still quite different from that of radiology, and the most efficient PACS implementation for cardiology must accommodate these differences. “The workflow in cardiology is also very different from the workflow of radiologists. Typically, the cardiologist will have a mixed workflow involving some imaging,” Desjardins says, but also incorporating outpatient visits and inpatient rounds. “In radiology, most studies are elective and are, therefore, scheduled in advance and interpreted later. In cardiology, most studies are stat cases. They are performed with almost no data entered in the information system beforehand, and the cardiologist interprets the images as they are being acquired. In radiology, images are shared with the rest of the hospital; in cardiology, they are typically not shared, although there are exceptions,” he adds. “Another difference is that cardiology images can be separated by hours, or even days. Exams take much longer to acquire than a typical radiology exam,” Desjardins says. Some two-part tests might be conducted with 24 hours between segments, for instance, so that the cardiologist can observe changes in perfusion, or between stressed and resting states. In some cases, radiology-department PACS have been customized to accommodate the workflow of cardiac radiologists (which resembles that of cardiologists), so these systems will be less likely to require major modifications before they can be used by cardiologists. Elsewhere, a fair amount of tinkering might be required before the highest levels of efficiency can be achieved in a joint cardiology–radiology PACS. “One of the main differences in imaging is that cardiologists are trying to image the heart—a structure that moves. It moves in a cyclic fashion, though, so imaging only needs to record one cycle and then play it over and over again to get all the information that’s needed about the motion of the heart,” Desjardins says. “Other images in cardiology are vascular, so they do not deal with a structure that moves, but they are dynamic themselves—they could represent the progression of the injection of a contrast agent into a vessel, for example, to detect lesions or asymmetry in filling of the vessel.” Desjardins says that capacity is another issue that requires attention. “There’s been a huge volume surge in cardiology imaging data. The imaging technologies used in cardiology now generate much more information, and there has also been an increase in imaging activity in cardiology. There are more patients, but there’s also more of an imaging focus in cardiology, and there’s more of a need for digital management because of this explosion of information,” he says. “You need a lot of storage, and you need generous bandwidth to handle the data.” Cardiology uses a different set of views (based on echocardiography conventions) than those that radiology uses, Desjardins explains, so a PACS must be able to manage those differences, in addition to the more quantitative nature of cardiology data and the presence of nonimaging signals (such as ECG output) that must be coordinated with the images. Enterprise integration of cardiology can be done, Desjardins says, in steps. Initially, he recommends sharing storage with radiology for economies of scale. Further integration will proceed in stages, but the organization must first decide whether support for any merged systems will be provided by cardiology, radiology, IT, or some other department. “More advanced ways to display cardiology images are coming to market right now,” Desjardins reports. Some hurdles will be technological, but the main obstacles will be logistical and political. “After the problems of integrating cardiology have been solved, then the issues involved in integrating the enterprise can be addressed,” he says. PACS for Pathology Paul J. Chang, MD, is professor and vice chair of radiology informatics and medical director of pathology informatics at the University of Chicago School of Medicine. He notes that the benefits of infrastructure integration are considerable for both pathology and radiology departments. “Their deliverable, like ours, is a report based on the interpretation of images, even if their images are slides. As we have embraced structured reporting, they also have a similar initiative (called synoptic reporting),” he says. There are many similarities in the image-management needs of the two specialties, but there are additional requirements, specific to pathology, that are unlikely to be met by existing radiology PACS without modification. “Anatomic pathologists are really trying to model what radiology has done, in the past few years, in their workflow. Some of that makes sense, but some of it does not, because we are very different—but similar enough to make some of the same mistakes,” Chang says. The nature of the anatomic-pathology workflow is a particular challenge, since it is quite different from the typical workflow pattern seen in radiology departments. Radiology workflow has been shaped by RIS availability, but anatomic pathology has not been affected by the use of similar systems because the laboratory information systems used by some clinical pathologists have not often been used to handle anatomic pathology. As a result, Chang says, many (if not most) anatomic pathologists have no advance notice that a case is on the way to them. Instead, it is added to the worklist when a new specimen bucket is delivered to the rack that often serves as a pathologist’s inbox. “If the first time we knew that a CT scan was to be interpreted was when a patient just hopped onto the CT scanner—without any scheduling—how efficient would our department be? That’s how pathology runs in the vast majority of places,” Chang says. In addition, pathology poses greater challenges in image management than cardiology does because so much of the work is analog in nature. Instead of beginning a case with the acquisition of digital images, as radiologists and cardiologists who use PACS nearly always do today, the anatomic pathologist begins with a piece of the patient. While the pathologist’s report can be structured so that it becomes digital more easily, and while pathology images can be digitized—or, in some cases, captured using cameras and microscopes capable of generating digital output—there is no digital replacement for the thoroughly analog anatomic specimen. It must be manipulated in several ways before a slide can even be ready for the microscope. In many anatomic-pathology laboratories, Chang says, the only digital element of the workflow is that the images are digitized at the end of the process (if at all). “The fundamental thing that we exploit, in radiology informatics, is the fact that at the time of acquisition, everything is digital. Anatomic pathology is still not digital until the end,” he explains. The typical case-organizing mechanism in anatomic pathology is also an analog item: specimens, slides, notes, and reports are held together using the thick rubber bands more often found in produce departments, holding together bundles of broccoli. It’s very difficult to justify complete digital pathology archives, from a business perspective, Chang says, because an anatomic-pathology department (when its images are digitized) can generate 10 terabytes of data in a day—more than a typical radiology PACS handles in a month. A single slide can require 15 gigabytes of storage for just one of multiple focal planes. Chang says, “Many of the researchers and vendors in pathology informatics actually have it wrong, though. They think that they are too much like us, when in reality, their use of digital information is very different.” Pathologists don’t require comprehensive prior studies for comparison; they usually need only a single slide, and it will have been designated as the one important image for future retrieval by being marked with a blue dot during the initial interpretation, Chang says. “This is a fundamental difference that vastly simplifies the storage requirements,” he adds. PACS for Other Specialties Steven C. Horii, MD, is professor of radiology at the Hospital of the University of Pennsylvania in Philadelphia. As he explains, the desire for enterprise PACS goes beyond the scope of adding cardiology and pathology to radiology’s image-management systems. Other specialties that use and store images, such as surgery, want access to advanced image-management capabilities. When they know what radiology can do with PACS, they can become increasingly insistent that they should be able to leverage the radiology system for their own purposes. For this reason, Horii says, radiologists should be prepared to answer the questions that other specialists have about PACS—and about what those specialists are likely to require before their departments can become functional parts of an enterprise PACS. “Things we don’t think of as images can wind up on our plate, if we’re not careful,” he says. “Virtually every specialty generates images. Medicine is largely a visual profession, but not all images are kept—and those that are kept are stored in a huge variety of formats. Surgeons who capture images in the operating room now want to be able to review them later in their offices,” Horii adds. “Obstetricians use video, gastroenterologists print gastroscopy photos to add to charts, and various departments print on thermal paper, inkjet and laser paper, polymer, and film. Dermatologists use film or digital cameras to take full-body photos to track the skin changes that can indicate melanoma.” He continues, “There are some specialties, such as ophthalmology, that use DICOM formats and digital images. Many other standards also are used, and then there are all kinds of proprietary formats that make our lives more difficult—and there are ways of storing data with no format at all.” As with cardiology and pathology, other specialties will probably find that their primary obstacle to smooth integration with existing information systems is that their workflow differs substantially from the pattern seen in radiology, for which PACS is designed. For example, dermatologists might benefit from access to prior images for comparison during diagnosis or to assess the patient’s response to treatment. Pathologists, however, use their images in the present, so there is no need for rapid retrieval of prior studies; images are stored primarily for use in education and in defense against possible legal challenges. Using the same storage rules that radiology employs to determine whether a case of a particular age needs immediate retrieval, slightly less rapid accessibility, or long-term (slow-retrieval) archival storage would be wasteful in specialties that have no need for rapid access. In addition, it could slow retrieval—depending on the system in use—for radiologists who are sharing PACS with other specialties, compromising speed of access for the specialty that needs it most. In a medical center or multispecialty practice, Horii says, the digital images generated can come from neurologists, ophthalmologists, otorhinolaryngologists, dentists and orthodontists, maxillofacial surgeons, pulmonologists, gynecologists, orthopedists, and infectious-disease specialists, in addition to more obvious sources. Horii calls on radiology and IT departments to conduct a careful evaluation of the radiology PACS to determine whether it is capable of handling any of these images from outside radiology—and if so, how much such activity it can manage without overloading the system and reducing its usefulness for radiology. What Horii defines as unexpected images that can become candidates for enterprise image management include ECGs, EEGs, and the output of fetal monitors and intraoperative patient monitors, since tracings and graphs are actually images. He asks, “Should these be stored as images or as the actual numeric data that they represent (such as voltage versus time) and then reproduced from those data as needed? How do I get back to the original numeric data if all I’ve done is capture it as an image?” Images alone, in such cases, have the same shortcomings as scanned handwritten documents: They can’t be indexed and they can’t be used quantitatively, so the opportunity to do anything with them, beyond looking at them again, is lost. For medicolegal reasons, however, it can be important to store the original images, since the numeric data are easily manipulated. It could, therefore, be necessary to store both the tracings and graphs (images) as originally captured and the numeric data first used to generate them. Horii considers enterprise PACS a challenge, but not an insurmountable one. He notes that one promising approach is the use of a federated PACS, in which a network of domain-specific PACS is created. Each domain has its own database, which is accessible throughout the enterprise; radiology becomes a component of the federated PACS, but is not its host. Another possibility is the use of an enterprise multispecialty PACS with a single database, but Horii says that this “is not realizable at present, with some exceptions.” In settings where radiology’s PACS will act as the host, a combination of federated PACS and custom integration—with some workflow modifications—will probably be required, he says. Piece by Piece Despite the conclusion of the three presenters that an adequate enterprise PACS should not be built simply by cobbling together extensions of the radiology PACS, they agree that the pursuit of the enterprise-imaging goal should not be discouraged. Information systems undergo continuing revision in response to innovations in both hardware and software; professional bodies are working to refine the standards that permit systems to exchange data without losing any of its value. Giving up on enterprise PACS would be advocating a return to the stand-alone information systems that have handicapped so many hospital departments because they didn’t allow information to be exchanged with other systems. Where these separate systems are already in place, they will become far more valuable if they are reconfigured to work with an enterprise image-management system. Where there is no legacy departmental system, the opportunity to create PACS components that will extend the benefits that radiology already enjoys to cardiology, pathology, and other specialties should not be overlooked. While the presenters emphasize that building an enterprise PACS is not likely to be a plug-and-play operation, that does not make the pursuit of integrated image management for the whole institution less important—any more than needing custom-made shoes would be an excuse to stay barefooted. There are gains to be had at each stage of PACS implementation, long before the complete enterprise solution takes shape at a given organization. For the sake of cost control, workflow management, staff recruitment and retention, referrers’ satisfaction, and better patient care, work toward enterprise PACS will continue at the vendor, professional-society, facility, and specialty levels. Kris Kyes is technical editor of Radiology Business Journal.