An MD&DI January 1999 Column
Digital Imaging Heralds Waning of Film EraAdvances in digital detectors drive the evolution of radiography.
As digital technologies begin to emerge in radiography, a century of total dependence on x-ray film is drawing to a close. Several companies are now selling medical products that generate electronic radiographs. More are in the final stages of development.
But unlike other digital imaging technologies such as magnetic resonance imaging (MRI) and computed tomography (CT), there is as yet no clearly superior way to make digital radiographs. Engineers have available a bewildering array of solid-state x-ray sensors. Even those devices in the mainstream differ markedly. They may use different materials, such as selenium or silicon, which may be applied in various ways and shapes. Some sensors rely on charge-coupled devices (CCDs) similar to the ones built into digital cameras. These CCDs can be optically linked or stitched into a mosaic of chips. Or a CCD array might be transported mechanically across the patient.
Figure 1. Cutaway view of the DirectRay detector array packed into a flat box. An amorphous selenium coating is placed over the thin-film-transistor matrix and the associated readout electronics (Sterling Diagnostic Imaging; Greenville, SC).
The engineering challenges are similarly diverse. Silicon requires a scintillator to generate flashes picked up by photodiodes. Selenium records the impact of x-rays directly but must be electrically charged to record the x-ray strikes and recharged before the next exposure. Both silicon- and selenium-based systems are expensive to manufactureand vulnerable to defects. A single dust particle can short circuit a whole line of data.
CCDs are a mature technology and, as a result, are inexpensive. Their manufacture is relatively consistent and defects, consequently, are minimal. But their small size requires either tiling or optical coupling to a scintillator to record the flashes indicating that x-rays are passing through the body, and both of these solutions have drawbacks.
Meeting these challenges, while difficult, offers extraordinary opportunity. The demand for digital data is growing. Physicians are looking for increased efficiency through networking and the rapid transmission of data from point to point. Digital detectors promise improved diagnostic confidence, and even expanded clinical utility as pathologies not apparent on film appear in computer-enhanced images.
A DIVERSE TECHNOLOGY BASE
The promise of such enhanced imaging is already apparent in the three digital products now on the market. Not surprisingly, these digital systems differ substantially in the form or way they capture digital data. The first modern digital x-ray system, Thoravision from Philips Medical Systems (Eindhoven, Netherlands), records x-ray data on a thin film of selenium wrapped around a drum. The selenium directly records the impact of x-rays as electrical charges, which are read by a probe and transmitted to a computer for reconstruction into a two-dimensional image. This system, commercialized some four years ago, can handle the largest patient, providing 19 x 17-in. coverage.
"We have done millions of patients," notes Hans Kleine Schaars, marketing manager of the common imaging subsystems group at Philips Medical Systems. "The only issue is that it isn't flat."
Figure 2. Developed for fast diagnosis and therapeutic intervention, this system includes an integrated CT scanner, a solid-state x-ray detector (center, above the table), and an articulated arm. The silicon detector produces real-time images to guide physicians during interventional procedures (Picker International; Cleveland).
The drum-based approach also makes the system heavy and difficult to site. Ideally, digital systems should fit neatly into existing radiography suites. Such is the case with the radiography product from Sterling Diagnostic Imaging (Greenville, SC), called iiRAD. This product, like Thoravision, uses a selenium-based detector. But the sensor is fashioned into a thin-film-transistor (TFT) array and packed into a flat box that fits into standard radiographic tables or hangs vertically like a conventional chest-film Bucky. Charges deposited by x-rays in the selenium are recorded and read off the TFT for interpretation by a computer. Sterling began selling a general-purpose and a dedicated chest system earlier this year. "A lot of orders [for these systems] are being tied to PACS [picture archiving and communications systems]," says Jim Culley, a product manager for Sterling Diagnostic. "We are getting into major hospitals."
The digital detector, called DirectRay, will get an even wider dispersion as OEM deals with Sterling kick in (Figure 1). DirectRay is being provided to device companies, including Fischer Imaging (Denver), for integration into other x-ray machines under their own labels.
The AddOn-Multi-System from Swissray International (Hitzkirch, Switzerland) does not use selenium but rather four CCDs, which record flashes of light that occur when x-rays strike a scintillator. A fiber-optic system conveys flashes to the CCDs, consisting of photodiodes configured into vertical and horizontal rows. Each photodiode stores an electrical charge proportional to the intensity of the flash, then transfers this energy row by row to the edge of the CCD, where the signal is amplified and sent on to a computer.
These three devices may soon be joined by digital products using a markedly different type of technology. Unveiled this past November at the annual meeting of the Radiological Society of North America, these products will each incorporate a sensor made largely from amorphous silicon. This film of silicon is matched to a TFT array. A scintillator coating the silicon produces flashes with each x-ray strike; the TFT then turns these flashes into signals, which are conveyed to a computer.
GE Medical Systems (Milwaukee), Siemens Medical Systems (Erlangen, Germany), and Philips are each expected to unveil a dedicated chest or general-purpose radiography suite based on flat panels made from amorphous silicon. On October 27, GE announced receipt of FDA clearance for a digital radiography system optimized for chest exams. And radiography is likely to be only the first step.
Panels made from amorphous silicon have the potential to support applications in real-time x-ray imaging technologies, including general fluoroscopy, angiography, and cardiac catheterization. Picker International (Cleveland) may be the first to commercialize such a real-time imaging solution, as part of its advanced system that combines an 8 x 10-in. silicon-based detector, made for fluoroscopy, with a CT scanner (Figure 2). The flat-panel fluoroscopy is designed to allow real-time guidance for interventional procedures, while the CT provides three-dimensional or slice-based information. Clinical tests of this integrated system began in the fall.
"The oncology market is really embracing this technology for procedures such as brachytherapy [in which radioactive seeds are implanted near tumors]," says Richard Silver, sales and marketing manager for the Picker x-ray division. "We will continue to focus on the fluoroscopic applications, taking it into larger fields of view as the flat-panel technology evolves."
CHARGE-COUPLED DEVICE APPEAL
Detectors made of amorphous silicon are also being groomed to play key roles in digital mammography. But they will likely be preceded in the marketplace by other types of digital detectors. Trex Medical (Danbury, CT) has developed a full-breast digital detector comprised entirely of CCD chips. Charge-coupled devices have a long history in mammography, having been applied more than five years ago in stereotactic biopsy equipment. In this role, the CCD is used essentially for targeting suspicious lesions. Its major constraint is size: in order to cover an entire breast, Trex engineers have stitched the relatively small CCD chips together like a high-tech quilt. A coat of cesium iodide, acting as a scintillator, is deposited directly on a fiber-optic taper. This taper is bonded to the CCDs, which read the flashes of light resulting from x-ray impacts.
Fischer Imaging is clinically testing a mammography system that uses a CCD but in a much different design than that of Trex. Fischer uses a slot-scanning device, which emits a narrow beam of x-rays that fan out laterally, exposing breast tissue between the x-ray source and a slot-shaped CCD detector. The detector sweeps across the breast, creating in its wake a stream of digital data that are compiled into an electronic image.
The Trex and Fischer designs, while different, both address the same problem inherent in CCDsa lack of coverage. Coverage is an even greater challenge when doing radiography. Swissray engineers solved the problem by optically coupling the CCDs to a scintillator so that the total chest could be captured.
But there are drawbacks to these solutions. The abutments between CCDs in the Trex detector create a checkered pattern of thin lines where data cannot be recorded. Algorithms that interpolate data from one point to another fill in the missing pixels, smoothing over these lines to create a homogenous image. In reality, some data are lostbut the loss is so small that it is not likely to affect the diagnosis, according to company executives.
"Any adverse effect on imaging is kept well below any level of significance," says Richard Bird, director of clinical and product development at Trex.
Swissray boasts that no data are lost in its use of CCDs, thanks to the optical couplings, which actually provide overlapping data sets. A scintillator is divided into four regions, each covered by rapid scanning optical systems. Flashes in these regions are recorded by each of the CCDs, which create corresponding electrical signals for computer reconstruction.
The challenge is to process the data into a cohesive image, subtracting redundant data and then creating an accurate image optimized to depict bone and soft tissue clearly. All this processing is done in less than 20 seconds, according to Ueli Laupper, CEO of Swissray America Inc. "When the images appear, the appropriate algorithms are already applied," Laupper says. "The radiologist can immediately start his diagnostic study. No change of contrast or windowing or leveling is needed."
The accurate presentation of these data must take into account the differences in sensitivity that inevitably occur from using more than one CCD. Even though the manufacture of these devices is routine, performance parameters can vary slightly from one CCD to another. CCDs in the Trex mammography detector are also susceptible to this problem. But the digital nature of the technology provides the solution. Calibration software has been written to align the different CCDs while adjusting the data accordingly during image acquisition.
The Fischer Imaging device is spared these problems. Slot scanning creates neither voids nor overlaps in the data. Instead, the sensor creates myriad individual pictures of small areas of the breast that are then pieced together by the computer to form a mosaic of the breast.
The several seconds needed to traverse the breast, however, makes the system susceptible to motion artifacts created by movement of the patient or gremlins in the mechanism that transport the detector and x-ray source. Fischer engineers reduceif not eliminatethe risk of diagnostic error resulting from motion artifacts by compartmentalizing the individual images, addressing them as a series of subsecond, high-resolution snapshots. Additionally, the Fischer engineers designed the breast tray to ensure that the source and detector would remain at the same distance throughout the sweep. In the process, they designed a machine that causes less discomfort to the patient.
"They've designed the breast support tray to be slightly curved and it is that curvature that keeps the array and tube at exactly the same distance throughout the sweep," says Cynthia Malin, a product marketing manager at Fischer Imaging. "As a side point, when they did that they made a more comfortable mammography machine, because the compression is more even across the breast, with less force being applied."
ADVANTAGES OF PLATES
Amorphous silicon plates would seem to have the edge over competing technologies. Flat-panel sensors come off the production line at EG&G Amorphous Silicon (Santa Clara, CA) as 41 x 41-cm sheets that require no stitching, optical coupling, or mechanical transport across the body. When plugged into their surrounding electronics, these panels can capture an entire chest image in a single shot of x-rays (Figure 3). Alternatively, the company can precut these sheets into smaller sizes for detectors to be used in mammography, angiography, or even bone-densitometry products.
"It's the same concept as a semiconductor manufacturer that makes a bunch of devices on a single wafer and then cuts up [the wafer]," says Andres Buser, the general manager of EG&G Amorphous Silicon.
Getting to this point, however, has not been easyor cheap. The basic research to develop this detector technology cost GE Medical Systems more than $100 million during the past decade. The cost of developing and building the manufacturing line at EG&G Amorphous Silicon is not publicly known but is likely to be in the tens of millions of dollars. And production still has a long way to go. At full capacity, the EG&G plant can produce only a few thousand detectors per year.
"We run them in lots of eight, but there are some steps in the process that take a whole day," Buser explains.
If digital x-ray systems are to be widely adopted, the cost per unit must come down. The end-user products now in the marketanalog devices that use filmare among the least expensive of all diagnostic imaging systems. By comparison, a full-breast diagnostic mammography system from Trex, when approved by FDA, will cost more than $350,000.
The need to generate digital images and the promise of clinical efficiencies possible with digital imaging are expected to help justify the cost of digital x-ray equipment. But such justifications can go only so far.
Figure 3. GE Medical Systems (Milwaukee) and EG&G Amorphous Silicon (Santa Clara, CA) have teamed up to manufacture a digital x-ray detector that produces x-rays without film. Cynthia Landberg, PhD, of the GE Research and Development Center mounts a prototype detector in an apparatus designed to test functionality.
Vendors must control the cost of the sensor if they are to be successful. That harsh economic fact may be the greatest challenge facing this technology. It has already claimed the financial health of one supplier of sensor components, OIS Optical Imaging Systems (Northville, MI).
The company shut down its manufacturing plant in early September, citing unacceptable losses from the production of flat-panel sensors and displays. The plant reopened late that month by order of the U.S. Department of Commerce, which forced company owners to initiate production to meet government contracts to make displays for military weapons systems. The federal order was a reprieve for Sterling Diagnostic Imaging, which only weeks earlier had begun selling its iiRAD system, whose digital detector requires the selenium plate made by OIS.
"This is the best news," said Sterling spokesperson Jayne L. Seebach in response to the government's order. "But you better believe we will be simultaneously finding another supplier."
It will be a tough search. Only two companies other than EG&G Amorphous Silicon and OIS have the in-house expertise and equipment necessary to make solid-state imaging detectorsdpiX and Trixell SAS. One, dpiX, is the manufacturing arm of Xerox PARC (Palo Alto, CA), which can draw from deep financial pockets at Xerox, if necessary, until the market is fully developed. Trixell is a consortium formed in January 1997 by Siemens Medical Engineering Group, Philips Medical Systems, and Thomson Electroniques. Based in Moirans, France, Trixell has a built-in market for its solid-state detectorsPhilips and Siemensand hopes to sell them to other OEMs as well.
Other companiesthe makers of LCD flat panels, for examplehave the fundamental resources to make the components, but they will have to be convinced to do so. Simply, the global market for digital detectors is defined in the tens of thousands per year, as opposed to the millions of units per year of demand for LCD panels that comes from the computer industry alone.
At the very least, the vendors of imaging equipment are committed to digital x-ray. They have little choice. The practice of medicine is growing increasingly dependent on computers and networking. To be cost-effective and efficient, these networks must include x-ray images, which account for about 70% of all the imaging studies done in the United States. Additionally, as doctors grow increasingly accustomed to the benefits of digital image display in other modalities such as MRI, CT, and ultrasound, they will demand the same from radiography and the various applications of x-ray fluoroscopy.