The trusted source for
healthcare information and
RSNA meeting provides insights on where technology is heading
By Philip G. Drew, PhD
Developments on display at the 85th scientific assembly and annual meeting of the radiological Society of North America (RSNA), held in Chicago, November 28 to December 3, 1999, promise that the technology and practice of medical imaging are in for big changes in the new millennium. As in every other technically based enterprise, the revolution being wrought by digital computation and the Internet extends into every aspect of modern medical imaging practice.
The obvious part of the revolution lies in computer applications for image management, a field that goes under the rubric of picture archiving and communication systems or PACS. These systems aim to replace film with their electronic alternatives—monitors on which to view images, networks on which to distribute images, and digital archives in which to store images. Beside avoiding the cost of film and film-handling, these systems introduce other benefits in streamlining radiology department operations and improving service to referring physicians. PACS that fully eliminate film are still more the exception than the rule, but eliminating film is the goal of many hospitals and imaging centers that have embarked on programs to install PACS in stages over a period of years.
Of the 600 or so companies exhibiting their wares at the meeting, fully half concerned themselves in one way or another with PACS. The field is broad, and the suppliers approach PACS from a variety of starting points. The imaging equipment manufacturers recognize in PACS an opportunity to expand their businesses in the radiology departments of the world, and after a slow start they have leapt enthusiastically into this business. The film companies—such as Agfa Medical (Ridgefield Park, NJ), Fuji America (Stamford, CT), Kodak (Rochester, NY), and Konica (Wayne, NJ)—see in PACS an eventual end to their film business, and they have embraced electronic systems, though with varying degrees of success. The hospital information systems companies see in PACS a new function for their networks and they have added PACS to their systems, though, it appears, with some reluctance because PACS introduces a new and difficult set of technical requirements.
PACS Vie for Attention
But the largest and most interesting group is made up of companies whose medical imaging business is confined to PACS. These companies are home to the novel concepts that vie for a place in the imaging departments of tomorrow. One new concept is web-based image distribution. It has become clear that systems to distribute images to the hundreds or thousands of referring physicians typically served by an imaging department cannot provide specialized terminals for each physician nor can they use a local area network to reach all of them. The apparent solution is a distribution system that uses personal computers already in place for terminals and an intranet or the Internet for distribution. Currently, transmissions of images on such networks take too much time, but data compression alleviates this problem. (As reported below, if prognostications from Lucent Technologies [the former Bell Labs] turn out to be correct, the problem may not persist.)
Several companies offer systems that address this aspect of PACS. Algotec (Raanana, Israel), Amicas (Watertown, MA), Applicare (Zeist, Netherlands), eMed (formerly Access Radiology, Lexington, MA), Insite One (Wallingford, CT), Line Imaging (Atlanta, GA), and Medweb (San Francisco, CA) all start with the concept of Internet distribution, whereas most PACS suppliers include the Internet as an afterthought.
Using a compression scheme developed at the University of Pittsburgh, Stentor (South San Francisco, CA) provides software that compresses images and then transmits only as much of the image detail as is necessary to match the display capabilities of the destination terminal. Using a compression scheme from Los Alamos, LizardTech (Seattle, WA) provides software with a similar function. Both systems claim to speed up network operation by diminishing network traffic, transmitting only what is necessary for the viewer to make a decision.
New Reasons to Use 3D
Another aspect of PACS largely pioneered by small companies is 3D imaging. In the early days of 3D some 15 years ago, small companies developed 3D systems, but manufacturers of CT and MR systems quickly incorporated the software in their systems. As a result, most 3D companies failed, though some, like ISG Technologies (now Cedara Software, Mississauga, Ontario) continue to supply the software that the larger companies use. Although most buyers of CT and MR systems buy the 3D software, most of them do not use it because it is too slow and cumbersome. Now, however, there are new reasons to use 3D software. Because CT and MR machines are much faster, it is feasible to conduct studies that may comprise hundreds of slices, and interpreting the study by examining each slice can be unduly time-consuming. By presenting all the data in a single image—or in a more manageable number of views from various angles—3D images can make image interpretation faster. And 3D has always had the virtue of being more intelligible to surgeons, patients, and other users of CT and MR images.
Who's Doing What
AccuImage Diagnostics (South San Francisco, CA) showed software that produces 3D images and detects and quantifies calcification of coronary arteries. Dicomit (Richmond Hill, Ontario) offers a 3D system for ultrasound in addition to its connectivity systems. Taking advantage of years of development, Vital Images (Minneapoplis, MN) showed its successful Vitrea 3D software, which is now available to run on systems using Windows NT as well as more expensive machines. Voxar (San Antonio, TX) introduced Plug'n View 3D software at an unusually low price—$5,000. Voxel (Orem, UT) produces 3D images using holography, an approach totally different from that of the other 3D companies. Having embarked on this path a decade ago, Voxel hit hard times when it encountered difficulties in designing and manufacturing the camera that makes the holograms, and sought protection in bankruptcy. The new owners believe the concept still has merit, though the holographic presentation has received only tepid interest from the radiology community.
Several companies seek to exploit the ubiquity and convenience of the Internet in ways analogous to Yahoo! but specialized for radiology. AuntMinnie.com and radiology.com both aim to be sites where a user can find all kinds of information about radiology—case files, locations of imaging facilities, equipment for sale, current and archived patient images, maintenance services, discussions of current issues, classified ads, etc. As every reader of the business pages knows, companies of this kind in other venues are showered with venture capital and IPO enthusiasm. Their viability is sustained by their potential as places to advertise, but their effectiveness in this role is not yet clear. What is clear is that companies must stake out their territory on the Internet early or it will be taken up by someone else. This fact explains the urgency with which such companies add features to their sites.
DR vs. CR
Still on the front burner this year were direct radiography (DR) systems, which provide a means to capture X-ray images in digital form. Doing so is a necessity for PACS that aim to do away with film, since 70% of all imaging procedures in the United States are plain film X-ray procedures. Computed radiography (CR)—the system originated by Fuji (Tokyo, Japan) and available from Agfa (Mortsel, Belgium) Konica (Tokyo, Japan), and Kodak (Rochester, NY)—substitutes a reusable phosphor plate for film. After exposure, the image on the plate must be read out in a special reading device analogous to a film processor. Lumisys (Sunnyvale, CA) offers a low-cost desk-top CR processor, which can accept plates from any of the three manufacturers noted above.
In DR, the receptor plate is permanently installed in the X-ray machine and produces the digital image data immediately after exposure. Avoiding the need to handle a cassette and producing images slightly superior technically to those produced by CR, DR seems as if it should quickly replace CR. However, that is not the case, because DR is much more expensive. Where a CR processor that can service three or four rooms simultaneously costs $200,000 or less, a DR system for a single room costs around $400,000. The main reason for this high cost is the difficulty of fabricating the large transistor arrays necessary to convert image data to electronic form. In the long run, there is little doubt that DR will supplant CR, but the conversion will take years, probably decades.
In the meantime, there is no dearth of companies interested in DR. One version of the process was pioneered by DuPont (Wilmington, DE), which eventually sold both the X-ray film business and the DR development to a start-up called Sterling. Most recently, Agfa bought the film business from Sterling, but left the DR system on the table, where it was eventually acquired by Hologic (Waltham, MA). Kodak has indicated that it will market a system designed by Analogic (Peabody, MA) using the Hologic plate. Another version has been under development in Europe by Trixell (Paris, France), which announced that it will begin delivering plates early this year. Trixell is a joint venture of Siemens, Philips, and Thomson, but the company will sell plates to anyone. Having created an alliance with defense contractor EG&G, General Electric Medical Systems (Waukesha, WI) added a second product to it DR offerings, a general-purpose X-ray machine supplementing a chest system.
There is another approach to DR, which is more simple and less expensive. Instead of using a large array of transistors, these systems use a phosphor screen to convert incident X-rays to light, which can then be focused optically on charge-coupled devices (CCDs), which convert the light to electronic signals. The difficulty faced by designers is that CCDs are small—no larger than about 2 × 2 inches—while the area of the phosphor screen is large—as large as 17 × 17 inches. The focusing optics suffer in two respects—they are inefficient because they collect only a small fraction of the light produced, and they require considerable distance—several inches, at least—of stand-off to accommodate the focal length of the optics. The latter condition means that such structures tend to be thick, so that they cannot be retrofitted to existing machines without modification of the machine.
Nevertheless, a number of companies have built systems based on this approach. Swissray (New York, NY) is one of these, and the company has had some success selling its product, mostly abroad. Caresbuilt (Keyport, NJ) uses a 20 × 20 array of small lenses and, thus, is able to squeeze the thickness of their plate down to 2.5 in, making it possible to retrofit into existing machines. They have chosen an image array of 7000 × 7000 pixels, making it the most highly resolved digital X-ray detector on the market. Theoretically, film has even higher resolution for high-contrast objects, but numerous studies have shown that with the possible exception of mammography lower resolution does not affect diagnostic accuracy. Furthermore, the 49 million elements of data that make up a Cares Built image tax the data-handling capabilities of computers and networks.
A highlight of this year's RSNA was a new program called the Integrated Health Enterprise (IHE), advertised as a meeting within a meeting. It is a response to what has become a perennial vexation for designers and users of PACS. Hospital information systems started out as accounting and billing systems and only gradually added the clinical functions that constitute most of the real work in a hospital; only recently have they addressed the issues of electronic archiving and retrieving of the full gamut of patient information that includes images, traces, and data. Partly for this reason, hospital information systems are often fragmented with different parts addressed by different suppliers. To facilitate interchange of information among systems provided by different manufacturers, the industry has adopted a standard called HL7 (for Health Level 7). This standard concerns alphanumeric data and is silent on the matters of concern when the data represent images.
Designers of PACS found that they needed a standard of their own to address the problems of interchanging images among systems from different manufacturers, and the PACS industry adopted the so-called DICOM (for Digital Image Communication in Medicine) standard. HL7 and DICOM are independent of one another, but system designers have need of both. Image management systems need data, such as patient demographics and insurance data, that come from information systems; when they address the electronic patient record, hospital information systems need data, such as images and traces, that come from PACS.
New Use of Existing Technology
This is the goal of the IHE program—not to develop new standards but to facilitate information exchange among medical information systems, whether they adhere to the HL7 standard, the DICOM standard, or any of the other standards now in development. To do this, IHE undertook a connectivity demonstration at RSNA involving 24 self-selected participants, who include some of the major HIS and PACs suppliers as well as a number of smaller companies. Terminals in their booths were connected to a common network over which they could exchange image and alphanumeric data whether the data originated in accordance with the HL7 standard or the DICOM standard. The demonstrations showed that each participant could support all the data exchange necessary for a hypothetical medical scenario, such as a patient coming to the emergency room with a heart attack.
The key to this demonstration was a test node previously developed at which companies could exercise their systems to assure that their code worked properly when connected to the network. This approach worked well in the past, when the organization RSNA set up an equivalent network for companies to demonstrate their ability to work in accordance with the DICOM standard. Though only 24 companies participated, and two important companies—McKesson HBOC and SMS, were absent—this was just the first year of IHE, and many more companies can be expected to participate in future demonstrations. The program will be repeated at this year's Healthcare Information Management Systems Society meeting in Dallas next April.
Echoing a Theme
This theme—integration of PACS and RIS—was echoed by a number of companies at RSNA. Picker, reincarnated as Marconi (Cleveland, OH), chose the new name to emphasize its commitment to interoperative systems. Agfa announced an "Embedded-IS," expected to be available this year, which integrates an RIS into their highly successful PACS. Starting from the RIS side, Dynamic Healthcare Technologies (Lake Mary, FL), IDX, SMS, and others asserted their compatibility with PACS. Aiming at integrated products, GE has an alliance with Cerner, and Philips with Sectra Imtec (Linkoping, Sweden).
In addition, to the demonstrations, IHE at RSNA included 24 sessions on subjects of interest to all concerned with interoperability of information systems and PACS. There were two talks at plenary sessions, one delivered by Harry Bosco, the chief operating officer of optic networking at Lucent Technologies, the other by Selby Wellman, a senior vice presidentof Cisco Systems. Both were distinctly upbeat, and both laid great stress on the role of the Internet and broadband communications. Wellman talked about the semi-magical possibilities that the Internet may provide, like being able to look at the interior of your refrigerator while you are at the supermarket, or having your house brought to just the right temperature by a signal sent from your homeward bound car.
Emerging Role of Fiberoptics
Bosco spoke about matters more germane to the medical world. In his view, fiberoptic cable will make so much bandwidth available to everybody at low cost that virtually any system one could conceive will be technically possible. Exactly how this will come about is one of the great questions facing the communications industry. He placed his bet on fiberoptic cable on every telephone pole with wireless light communication for the short links from pole to house, where, he said, lie 80% of the costs of realizing a system as fully connected as the telephone system. He added that stringing the cable is already proceeding at a furious pace, undertaken by dozens of companies in major cities. If he is right about available bandwidth, the issues of response time for image transfer and compression will vanish.
All in all, it is clear that PACS is the fastest growing part of medical imaging. Already at $550 million per year, the market for PACS products in the United States exceeds that for nuclear medicine equipment and is approaching the markets for CT and MR. It is no wonder that so many companies are active. Whether they all will survive is another matter, but the situation is not so dire as it was with the multiplicity of companies that entered CT or MR in their early days. In both of the latter cases, there were 25 or 30 companies producing products, while there are now only a dozen or so, most of them the larger imaging equipment companies. On the other hand, while there are only a handful of large HIS companies, there are still dozens of smaller companies with niches in particular segments. PACS seems more like HIS than like CT or MR. It seems likely that many of the smaller companies listed above will make their marks in this field—some surviving on their own, and others transferring their technology and their customers to larger companies.