An MD&DI October 1999 Column
Advances in computed tomography, magnetic resonance imaging, and ultrasound systems bring challenges as well as benefits.
Like voyagers looking at their world in pre-Columbian times, physicians have been trained to look at their patients as if they were flat. For more than a century, since the discovery of x-rays, flat images have characterized the practice of medicine. Film images viewed on a light box or held up to an outside window have provided the comfort of a "hands-on" experience.
This 3-D cardiac image, acquired using the Somatom Plus 4 CT scanner with Volume Zoom upgrade, clearly shows detailed coronary arteries. Photo courtesy of Siemens Medical Systems.
This dedication to a two-dimensional world, however, is coming to an end because medical imaging as we know it will soon cease to exist. Advanced imaging techniques in computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound are generating such large data sets that physicians can no longer effectively glean the necessary information to render a diagnosis. The next generation of CT scanners, high-performance MR scanners, and multiplanar ultrasound systems are leading diagnosticians into the future—a journey born of clinical and technological necessity.
Multislice CT scanners being sold by GE Medical Systems (Waukesha, WI), Siemens Medical Systems (Hoffman Estates, IL), and Picker International (Cleveland) process patient volumes so quickly that they threaten to overwhelm diagnosticians with the hundreds of slices created in little more than half a minute. Advanced MR scanners being offered by GE and Siemens capture whole volumes of the beating heart, as well as blood coursing through coronary and peripheral vessels. Ultrasound images generated by scanners made by Volumetrics Medical Imaging Inc. (Durham, NC) and ATL (Bothell, WA) create real-time 3-D images of a heart, blood vessels, or a fetus. Together they raise the prospect of more effective and accurate diagnoses, as well as less-invasive therapy for patients.
NEW PERSPECTIVES ON DIAGNOSTIC IMAGING
MR and CT scanners acquire and reconstruct blood vessels in three dimensions, demonstrating pinched stenoses and ballooning aneurysms. A 3-D presentation of a right renal artery can show a collapsed segment not visible in the projected plane of an x-ray arteriogram. Surgeons can then plan their approach to the pathology: selecting the route of least resistance and measuring the aneurysm to determine proper coil placement—exactly.
"A volume-rendered MR angiogram can be very intuitive, even to somebody who knows nothing about MR angiography." says John Paul Finn, MD, professor of radiology and director of magnetic resonance research at Northwestern University Medical School (Chicago). "This makes vascular surgeons much more confident that they're looking at what you tell them is there."
An emergency medicine specialist pages through a 3-D model of a trauma victim, axial slice after slice— from top to toe—through the head, the neck, the chest, the abdomen, the legs. This new genre of multislice CT scanners, which produce four slices with each turn of the detector around the gantry, render hundreds of extraordinarily thin slices. In little more than 30 seconds, they generate enough diagnostic data to replace the information yielded by hours or even days of conventional radiographic tests. Radiologists at the University of Michigan Medical Center (Ann Arbor, MI) have proven the potential of such scans, programming their multislice CT system with fields of view and algorithms optimized for different parts of the body.
"The result is a total spine, a total body, and a head evaluation," says Ella Kazerooni, UM associate professor of radiology. "This can only result in better patient decisions, and reduced morbidity and mortality."
REDUCING MOTION ARTIFACTS IN CARDIAC STUDIES
Beating 60 or more times per minute, the heart represents the greatest challenge for real-time imaging. The ventricles, walls, valves, and coronary arteries jumble about, their motion blurring conventional MR images. Advanced MR techniques, however, may prove to be safer than the gold standard, contrast-enhanced x-ray angiography; more sensitive than stress echocardiography; and faster and more accurate than nuclear cardiology.
"Tremendous strides have been taken in imaging the heart," Finn says. "Coronary arteries remain, I think, the most difficult task we have to tackle. But, even there, significant progress has been made."
Among the pioneers is Philips Medical Systems (Shelton, CT), whose Navigator technology, built into its Gyroscan MR scanner, keeps track of the patient's diaphragm, gating data acquisition to freeze out motion artifacts resulting from respiration. In the past, breath holding—sometimes 30 or 40 breath holds for 20 seconds each—was the only way to control such artifacts. The power of the scanner's gradients and special pulse sequences reduce blurring from the pumping motion of the heart.
"Without the Navigator I think it is very hard to image the coronary arteries," says René M. Botnar, PhD, a member of the team working on cardiac MR at Beth Israel Deaconess Hospital (Boston) and a clinical scientist from Philips Medical Systems. "The breath-hold method works fine in volunteers—but as soon as you have patients, it gets really difficult."
The researchers documented that MR angiography can provide both anatomic and functional information about coronary circulation. Among the quantitative measures are coronary blood flow and flow reserve, ejection fraction, heart-wall motion, and perfusion.
Echocardiography, the application of ultrasound to cardiac assessment, is fighting to maintain its place as a premier imaging modality. Volumetrics is helping. Its Model 1 scanner, capable of real-time 3-D cardiac imaging, captures the entire heart volume 30 times per second. Enough data for a complete exam are captured in several seconds of scanning. Nowhere is that kind of speed more important than in the stress echocardiography lab, where cardiac patients are evaluated for heart function with ultrasound.
"Stress echo puts this technology in a truly clinical application, one that's widely used today and is growing," says Gary Abrahams, Volumetrics vice president of marketing and sales. "We've been able to demonstrate that we can do more accurate and precise diastolic and systolic calculations by using volumetric measurement."
There are other such opportunities in ultrasound. Among them are 3-D renderings of vasculature, such as the vascular bed that grows around tumors. Similarly, fetal exams may uncover 3-D features that indicate abnormalities such as a cleft palate or club foot. Fetal heart structures—including the four chambers, ventricular and atrial septa, foramen ovale, and cardiac valves—can be defined.
CONVINCING THE SKEPTICS
As stunning as these images may be to the untrained eye, many prospective users of the technology remain unimpressed. Radiologists do not need 3-D reconstruction, they say, because their mind's eye has been trained to visualize two-dimensional images in 3-D.
Research aimed specifically at finding clinical value in performing on-screen 3-D reformatting has uncovered only a few cases in which 3-D is actually useful in diagnostic interpretation. These involved complex cases, such as discovery of an anomaly in a shoulder or knee, or the subtle signs of malformation in the fetal head. None has been enough to convince many physicians to use 3-D imaging routinely.
The virtual cockpit, developed at Stanford's 3-D Medical Laboratory, shows the human colon as it would appear from the inside.
Practical issues now looming on the horizon, however, may force the medical community, including radiologists, to accept three-dimensional imaging. In the evaluation of trauma victims at the University of Michigan Medical Center, a single multislice CT exam typically generates 1000 or more individual images. Kazerooni notes that the physicians do not have enough time to look at them all before settling on a treatment. Nor do they have much opportunity to go back for what they might have missed. "We have to dump the raw data because we don't have enough archive space," she says. Three-dimensional reconstruction, on the other hand, promises to allow physicians to quickly assess a large volume of data.
Also being considered are computer-assisted diagnostic aids that might highlight certain findings indicative of disease. Such advanced interpretation methods, however, are barely more than concepts at present. Nevertheless, there is general consensus that advanced methods are needed to handle and benefit from the information overload.
Similarly, few would argue that 3-D ultrasound is anything but a research tool at this point. Advances in scanning equipment, however, could foment a revolution like the one now going on in CT, where huge volumes of data need to be evaluated efficiently. In September 1999, ATL unveiled a new type of technology that gathers data in real time from nine separate points. At present, these data are all in the same plane, forming a tomographic two-dimensional image. But company engineers are working on a version that will gather data in three dimensions. The advantage with this technology, says Jacques Souquet, ATL chief technology officer, "is in the area of processing the information in real time. It's speed."
Still more dimensions may be added in the form of functional data. For example, measurements of cardiac wall thickening and heart ejection fraction might be added to the ultrasound image. Biochemical data plumbed with spectroscopy from cellular activities might be integrated into the MR image. Additionally, novel ways of viewing image data are getting the attention of physicians. Virtual endoscopy is among the most intriguing.
VIRTUAL FLIGHTS OF FANCY
For the past several years, software developers have been inviting physicians to "fly through" their data. CT, MRI, and ultrasound have been used to generate the playing field for inside-out imaging where users venture into CT and MR recreations of the carotid, twisting and turning past plaque formations; delve into the bowels, pass through the bronchioles; or fly through fallopian tubes created from ultrasound data.
The renal arteries and kidneys are displayed in these images acquired by the Somatom Plus 4 CT scanner (Siemens Medical Systems).
The first such trips were in "visible humans"—volumetric data sets of a man and woman whose corpses had been electronically interred in the National Library of Medicine in 1994. These data, composed of CT, MR, and cryosection images, were ideal for testing early virtual imaging software.
Much like these early forays, current virtual endoscopy products are primarily engineering novelties. No one disputes, however, that they could someday replace conventional exams, which—like colonoscopy—are time-consuming, expensive, uncomfortable, and more prone to error.
The power of this software-based technology comes from the power of new imaging technologies. A multislice CT exam can cover the entire abdomen and pelvis in less than 25 seconds with a slice width of 2.5 mm, narrow enough to capture the microfolds in the colon that could be missed with traditional endoscopy. The vastly increased speed of multislice scanners is expected to force this transition initially by increasing throughput while making virtual CT endoscopy more cost-effective.
Some 3-D imaging exams, such as sagittal and coronal reformatting of CT brain images, have their own CPT codes and are reimbursable. To be economically practical, however, these reconstructions must be done with blazing speed, possible only with massive computing power that is just becoming available. Achieving advanced processing in real time requires having the bandwidth to handle massive quantities of data and the circuit boards to process those data instantaneously. A supplier to several major imaging vendors, Mercury Computer (Chelmsford, MA) specializes in components that meet these requirements.
"You have to be able to bring data in, to process it, and display it all in a real-time framework. To be able to do that, you need a balance between bandwidth in and bandwidth out, and the processing in between," says Don Barry, vice president of the Mercury medical business group. "The Mercury architecture is designed for that."
Tera Recon, a Japanese company with U.S. headquarters in San Mateo, CA, is embarking on an aggressive strategy by selling advanced imaging components to OEMs, as well as marketing turnkey workstations and advanced networking systems for conveying images in the context of other patient data. "We really want to achieve a kind of synergy from the integration of technologies," says Motoaki Saito, MD, Tera Recon president and CEO. "In doing that, we believe we can create a totally new modality unto itself."
Soon to be released is the IiVS (integrated image viewing station) 320. The workstation, which will be list-priced at $50,000 or less, is based on an SGI 320 computer, and can be upgraded to an SGI 540. The system will allow the operator to control the way images and data can be displayed. For example, several 2-D slices may be displayed along two sides of a 3-D rendering. The system is designed to support high-frame-rate cine images, even virtual endoscopy, while showing 2-D or 3-D views to provide positioning data that help orient the operator.
This 3-D reconstruction, acquired using the Somatom Plus 4 CT scanner, shows identifiable details of the coronary arteries (Siemens Medical Systems).
From the Stanford University 3-D Medical Imaging Laboratory (Stanford, CA), Geoffrey D. Rubin, MD, assistant professor of radiology and codirector of the lab, sees the future from a different perspective. Surrounded by supercomputers from SGI and Sun Microsystems, Rubin and his staff focus on the development of alternative means for volumetric analysis, including visualization and quantitation, of which 3-D and 4-D are subsets. The goal is to find more efficient and effective methods for evaluating imaging data—particularly CT and MRI data.
One approach is the "virtual cockpit" for flying through body structures. In this system, virtual cameras are arranged to capture images with a 180-degree field of view. The method has allowed display of the human colon, airways in the lung, and the aorta—each free of the geometric artifacts that otherwise commonly occur in wide-angle virtual endoscopy renderings.
NEW SOLUTIONS TO NEW CHALLENGES
By default, the future of medical visualization is sure to bring the unexpected, as contemporary solutions simply fall short of managing coming tasks. New ways of looking at data are inevitable. One may be flattening out the data, as in slicing the virtual colon down the middle, laying it flat and then flying over, rather than through, it. Another may be getting away from the hallmark of the modern computing age—the mouse interface. Rubin advocates an intuitive interface, such as the one built into the "pixel wall" in the Stanford 3-D lab. This six- by two-foot electronic light box, which he describes as "the alternator of the future," can be programmed to change digital images at the wave of the hand, a spoken word, or the flash from a laser light pen. "We have to be able to navigate and interpret efficiently," he said. "To do this we must have good control of the visualization."
In the end, practical—not technological—concerns are almost certain to win out. More than a decade of 3-D wizardry bears testament to that fact. Only now, when there are definable reasons for applying advanced methods, are practitioners interested in their adoption.