BBI Contributing Editor
CHICAGO, Illinois – As was amply demonstrated on the exhibit floor of the annual meeting of the Radiological Society of North America (RSNA; Oak Brook, Illinois) medical imaging embraces five major modalities. Four of the five – the exception being X-ray – are customarily digital. Computed tomography (CT), magnetic resonance (MR), ultrasound (US), and nuclear medicine (NM) all lend themselves to networking in medical image management systems, more commonly known as Picture Archiving and Communication Systems, or PACS, which was the subject of our RSNA Part 1 report in the February issue of BBI.
Although PACS is the part of medical imaging that has been growing most rapidly, the modalities themselves are the heart of medical imaging, which has been assuming ever greater importance in medical practice. Contrary to expectations, the number of imaging procedures of all kinds in the United States has been growing in recent years by 8% or 9% per year. It had been thought that, because medical imaging procedures cost some $63 billion per year, managed care would take steps to curtail use of these fairly expensive examinations. But just the opposite has happened. The reason may be that, as managed care has put a premium on faster patient encounters and shortened hospital stays, clinicians find that imaging and laboratory procedures, because they are less time-consuming than histories and physical examinations, are the preferred ways to reach definitive diagnoses. In any case, the number of imaging examinations in the U. S. now stands at 389 million per year, as shown in Table 1.
As demonstrated by the more than 600 exhibitors at RSNA, the imaging modalities continue to improve, not through dramatic new breakthroughs but instead through steady extensions and refinements of techniques introduced in the past. Chief among these is the conversion to digital detection, control, and processing, which with the exception of X-ray is now virtually complete. Digital systems are not subject to accumulating errors in the way that analog systems are, and they provide a degree of flexibility that makes it possible to avoid errors or to compensate for them when they do occur. The fact that computational power continues to grow ever faster and less expensive fuels this conversion.
In the case of X-ray imaging, the conversion has been gradual. As noted in our earlier report, computed radiography (CR) is the usual means by which film-based imaging can be converted to the digital realm, but less than 10% of all X-ray procedures use CR. The reason is both economic and practical. While CR makes X-ray imaging digital and offers certain technical advantages over film, the conversion is not inexpensive, and the advantages not decisive. (Direct radiography (DR) is even more expensive and only marginally superior to CR.) As PACS grows, CR and DR will continue to grow, but the growth will be gradual.
In the meantime, there were developments at RSNA that support the continuing role of film. Computer-aided diagnosis (CAD) is an area that seems poised for growth, as two new competitors to R2 Technologies (Los Altos, California) – Scanis (Foster City, California) and CADx (Laval, Quebec, Canada) – showed up at this year's meeting. In a sense, CAD treads on delicate ground because, if it were good enough, it could displace highly paid radiologists. But the FDA has shown a marked reluctance to approve products that take human interpretations out of the loop, even in areas where humans are known to perform poorly, as in examining Pap smears used to screen for cervical cancer. Partly for this reason, the CAD devices are marketed not as substitutes for radiologists' interpretations, but as supplements. The FDA-approved R2 ImageChecker has a light box to display film as well as a digitizer that creates an image of each film for computerized analysis and display on a CRT with suspicious areas highlighted. In this way, the machine calls to the radiologist's attention areas on the film that deserve careful scrutiny.
Tuning such devices is a tricky business. If the threshold of suspicion is set too low, many areas are marked as suspicious, and radiologists find themselves either examining an excessive number of normal areas or ignoring the promptings of the computer. If the threshold of suspicion is set too high, the computer does not add much to the radiologist's performance. And suitable thresholds are different for radiologists at different levels of competence. So, despite evidence of their contribution to more accurate diagnoses, use of CAD devices is growing slowly.
The computer-controlled masking device developed by Smartlight (Hackensack, New Jersey) also contributes to accurate diagnosis. Everybody knows that bright ambient light diminishes the eye's sensitivity to subtle details on a backlighted film, but reading rooms still are often not properly darkened. In particular, the bright area not covered by films on light boxes is often unmasked. In theory, it would not be too difficult to add pieces of dark film or dark paper to these areas, thus improving viewing conditions, but as a practical matter radiologists in a hurry do not bother to do so. Smartlight has demonstrated convincingly that the ability to discern subtle details in X-ray images on film is greatly diminished by this practice. Their device detects the position of film on a light box and automatically darkens the surrounding areas. In spite of its undeniable contribution to accuracy, Smartlight's product has met only limited success (some 300 systems installed worldwide), perhaps because it is out of step with today's trend away from film and toward PACS. Nonetheless, it is remarkable that significant improvements in X-ray viewing can still be made after more than a century of practice.
A few years ago, it seemed that CT was a mature technology with little room to grow, but that perception has proved misleading. Spiral scanning and multislice detection have revolutionized CT. In conventional CT, the X-ray tube is tethered to the frame by its power cables and consequently can rotate only through an angle limited to about 540 degrees. This is sufficient to collect data for a single slice, but after the rotation, the tube must be brought to a halt, the table stepped forward, and the tube re-accelerated for collection of data for the next slice. The result is that the fastest conventional CTs take two or three seconds to collect data for a single slice.
In spiral CT, the tube gets its power from slip rings and, as a result, can rotate continuously while the table translates continuously, reducing scan time to a second or so for each slice. With multislice detection, the arc of detectors is doubled or quadrupled, thus speeding up data collection and reducing scan time by the same factor. Elscint (Haifa, Israel) pioneered multislice detection, but that company's CT division was subsequently sold to GE Medical Systems (Waukesha, Wisconsin). Now all the major CT manufacturers – including GE; Hitachi (Tokyo), which does not market CT in the United States; Marconi Medical (formerly Picker; Cleveland, Ohio); Philips (Eindhoven, the Netherlands); Shimadzu (Osaka, Japan); Siemens (Erlangen, Germany); and Toshiba (Tokyo) – offer spiral scanning and multislice detection.
The value of these features lies less in reduced examination or faster throughput than in the ability to perform examinations where involuntary patient motion had been a problem. One such application is lung scanning, where these features make it possible to collect data for a complete examination during a single breath-hold. The RSNA meeting included a report that for diagnosis of pulmonary embolism contrast-enhanced spiral, CT is comparable to the more-complicated angiography procedure. Another new application is assessment of calcification in the coronary arteries. Heretofore, such scans could be performed only by Imatron's (South San Francisco) electron-beam CTs, of which there are only a few dozen installed.
The increased speed and dense data made possible by spiral scanning and multislice detection also can be exploited to improve spatial resolution, making possible, for the first time, submillimeter slice thickness. One application opened by these innovations is virtual colonoscopy or virtual cystoscopy, wherein a 3-D reconstruction of CT data yields images similar to those seen by endoscopy, without the invasiveness of endoscopy.
MR likewise has been advancing in new directions opened by reduced scan times. The reductions have been achieved through innovations in switching gradients. MR images depend on encoding the signals emitted by excited nuclei so that the spatial origin of the signals can be inferred. The means for encoding is to apply gradients to the main magnet's field by sending programmed currents through three sets of supplementary coils called gradient coils. Characteristically, MR signals fade away because the nuclei emitting them get out of step, owing to local field inhomogeneities. The nuclei can be brought back into step by reversing the evolution of their motion, either by applying a phase-reversing pulse or by reversing the gradient. It has turned out that reversing gradients, particularly if it can be done rapidly, enables collection of enough data for an image of a complete slice after a single excitation. This development is now a standard part of all modern MR machines as offered by GE, Hitachi, Marconi Medical, Philips, Shimadzu, Siemens, and Toshiba.
As with CT, this increased speed is valuable not so much for increased throughput as for new applications that would have been impossible with conventional MR. One such application is demonstration of soft plaque in coronary arteries. Even though they do not cause symptoms like angina, such plaques may rupture and break away and are now thought to be a major cause of heart attacks. Thus, their detection is a necessary first step in therapy, which may include lifestyle changes, lipid-lowering drugs, angioplasty, and bypass surgery.
Unlike CT, MR, and X-ray, which are dominated by the large imaging companies noted above, ultrasound includes other companies, among them Acuson (Mountain View, California), Aloka (Wallingford, Connecticut), Biomedica Esaote (Indianapolis, Indiana), Agilent/Hewlett-Packard (Andover, Massachusetts), and Medison (Seoul, Korea). The latter showed a fully featured ultrasound device listed for less than $30,000. In this respect, the ultrasound world seems to be re-entering a state last encountered in the early 1980s. At that time, ultrasound technology had reached a plateau, and the companies in the field were competing chiefly by lowering prices, which had fallen from $100,000 to around $60,000 for high-end machines. Suddenly, Acuson burst on the scene with a markedly superior scanner for abdominal examinations at double the going price. Overnight, the company reversed the trend toward lower prices and opened a new high-end market. Whether any such revolutionary developments are in store for ultrasound remains to be seen. They were not in evidence at RSNA.
One interesting development on the exhibit floor was the re-introduction of hand-held ultrasound machines. Such a product was developed in the 1970s, shaped like a dumb-bell, with the transducer on one end and the tiny CRT display on the other. At the time, it seemed too much like a toy without an accepted use, but in the intervening years the roles for ultrasound have much expanded, and there may now be a part that these machines can play. SonoSite (Bothell, Washington) and Teratech (Burlington, Massachusetts) displayed such products. SonoSite's product is self-contained and weighs five pounds. The company believes that the machines will find a role for examinations in cardiologists' offices and emergency rooms. The Teratech device connects to ordinary PCs and weighs only eight ounces. It would most likely lend itself to use in physicians' offices.
The technology of nuclear medicine evolves, but is not marked by real breakthroughs. Machines with multiple heads offer improved images by capturing more of the emitted radiation, and their advent has raised the average price of NM machines. Most suppliers now offer coincidence detection for multi-head machines, which allows them to carry out examinations using positron-emitting isotopes, mainly fluorine in fluorodeoxyglucose (FDG). These developments are not new, and in this respect there was not much to report from this year's meeting. ADAC Laboratories (Milpitas, California) now offers a gantryless gamma camera, where the detectors are located in the walls of the room, though the advantages of doing so do not seem decisive. Nevertheless, the future of nuclear medicine is bright because of advances in molecular understanding of normal and pathological metabolic processes. The easiest way to detect abnormalities in these processes is with radiopharmaceuticals, of which there are dozens in the pipeline.