Digital detector systems for projection radiography are becoming commonplace in the clinical environment, in conjunction with the deployment and implementation of PACS in radiology and the medical enterprise. Currently available devices include image intensifier-TV (II-TV) camera systems, computed radiography (CR) using photostimulable phosphor detectors, charge-coupled-device (CCD) linear arrays and optically coupled 2D cameras, large area complementary metal-oxide semiconductor (CMOS) arrays coupled to a phosphor scintillator, and thin-film-transistor (TFT) two-dimensional arrays coupled to scintillator/photodiode detectors or direct detection semiconductor signal converters. Properly designed digital detectors provide high spatial resolution simultaneous to delivering high detective quantum efficiency. A comparison of detector systems reveals advantages and disadvantages based on applicability to a given imaging task, ease of use, integration and interfacing into a PACS, overall system cost, portability, image handling, preventive maintenance, recurring costs, level of quality control (QC) commitment, and practicality.
Typical questions one may ask include: What kind of digital detector(s) should be implemented and how many systems are needed? What are the equipment costs?? How are the systems to be deployed in and integrated to a PACS environment? How does one calibrate and verify optimal performance? Have downtime contingencies been addressed? What about quality control issues and verification of optimal image quality? What is the typical amortization schedule of digital imaging equipment, replacement costs, and turnover strategy? What are the hidden costs of digital radiography? What about niche imaging applications that are not directly amenable to the transition to digital acquisition, such as dental panorex, mammography, operating room, and multiple film studies (eg, scoliosis and long-bone)? All questions should be considered carefully, as they are extremely important for a successful switch to and continual operation of a fully digital department.
|Table 1. A comparison of positive and negative attributes of radiographic equipment.
SCREEN-FILM DETECTOR LIMITATIONS
Insofar as screen-film has served as the major radiography detector for the past 100 years, the technology has essentially reached its peak capability. Intrinsic limitations include narrow exposure latitude (good image contrast but susceptibility to under/over exposure), film grain noise (reduces signal to noise ratio), chemical processing (often the weak link for optimal conversion of the latent image and an environmental problem), inefficiencies in handling and storage (requiring a huge infrastructure of space and people), and lack of image postprocessing capabilities (image optimization is tailored to the detector, not the examination). With decreasing costs and improving functionality, digital detectors designed for projection radiography can overcome many (if not all) of these limitations.
DIGITAL DETECTOR CHARACTERISTICS
Design criteria for digital detectors must, at the minimum, emulate the capabilities of screen-film systems in terms of spatial resolution, contrast resolution, and field of view (FOV). A 400 speed screen-film detector delivers 5 to 7 line-pairs per mm (lp/mm) spatial resolution, corresponding to an equivalent discrete pixel size of 100 to 70 mm. For mammography, 15 to 20 lp/mm (33 to 25 mm pixel size) is achievable. The requisite spatial resolution of a digital detector that can provide the necessary resolution for a given imaging task has been shown to be somewhat less with edge enhancement processing. For most diagnostic imaging examinations (except possibly digital fluoroscopy), a minimum resolution of 2.5 lp/mm (200 mm) is required, and preferably 5.0 lp/mm (100 mm). Digital mammography requires higher resolution of 5.0 lp/mm (100 mm) to 10 lp/mm (50 mm) to detect and morphologically discriminate microcalcifications.
Contrast resolution is an attribute that is often misunderstood or overlooked. From the digital perspective, the image contrast can be manipulated and enhanced to an extent determined by the SNR (radiation dose), to a point when the noise becomes objectionable and further contrast enhancement is detrimental. In terms of data acquisition and digital conversion, spatial sampling and data quantization are imperfect and generate “electronic” noise in the output image. Data averaging of small details occurs over the detector element area (the pixel), generating a loss of contrast and injecting “aliasing” noise. Quantization, the process of converting the continuous analog signal amplitude into a discrete digital number, is determined by the number of “bits” of the analog-to-digital converter (ADC). With insufficient quantization steps, differences between the analog signal and the corresponding digital value cause “quantization” noise errors. Generally, most digital detectors for projection radiography require 10 to 16 bits of information (1,024 to 65,536 unique digital numbersgraylevels) to keep the errors to a low, practical value. Most digital systems use 12 bits (4,096 graylevels) in the final output image, and each image pixel requires 2 bytes of storage space.
In a digital system with sufficient bit depth and resolution, increased SNR can be obtained by simply increasing the radiation exposure to the detector, but the cost is increased exposure to the patient. When comparing digital detectors, another comparison benchmark is the detective quantum efficiency (DQE), a measure of the detector’s ability to capture the incident radiation and convert it to a useful signal. This measurement is determined by quantitative measurements of the noise-power transfer of the detector and its response to incident radiation. A perfect system would have a DQE of 100% at all spatial frequencies. “Real” systems lose efficiency over smaller area (at high spatial frequency) due to the inability of the detector to efficiently capture the information. In general, digital detector DQE ranges from 10% to as high as 80% depending on the detector design and x-ray converter characteristics for large area objects (low spatial frequency). For smaller object size (higher spatial frequencies), the DQE drops rapidly to the point where the system can no longer retain the identity of small anatomical detail.
DIGITAL DETECTOR TYPES
CR and DR are the commonly used terms for digital radiography detectors. CR uses a passive detector known as a photostimulable storage phosphor that appears in shape and function like the screen-film cassette. Direct radiography is a term that describes a digital x-ray detection system that produces an image without user intervention after the exposure is completed. There are several digital detectors that can be classified as DR systems, including automated CR systems that do not require technologist intervention to read/process the image.
Image Intensifier/TV systems. The earliest digital x-ray detectors (circa 1975) were based upon an II/TV system, which adapted the output video signal to an ADC. State-of-the-art image intensifiers have a high gain, low noise, and moderate spatial resolution capability, largely due to the structured cesium iodide (CsI) input phosphor material; when coupled with a light converter system (eg, analog TV or CCD camera), the system produces a low noise, high quality signal. Limitations arise from the geometric distortion of the II, large bulky size of the tube and housing, and limited dynamic range due to the saturation characteristics of the TV camera. A light-limiting aperture (iris) in the optical coupling provides the capability to adjust the incident exposure by adjusting the amount of light that is output by the II. Primary uses of these devices are for fluoroscopy, fluorography, and angiography dynamic sequences. Image matrix sizes contain as many as 2000 x 2000 pixels. The future portends increased use of flat-panel array detectors that have high gain and low noise, allowing the direct replacement of the 50-year-old intensifier tube technology. Reduced space requirements, better positioning flexibility, and combined digital fluoroscopy and high-resolution radiography imaging are all positive benefits.
Computed Radiography. CR was first introduced in the early 1980s, but high system costs, large size, and image quality issues blunted its widespread clinical appearance until the early 1990s with the introduction of more reliable systems with a smaller footprint. In function, CR emulates the screen-film paradigm very closely. The phosphor plate is housed in a cassette that resembles a screen-film cassette; in fact, it is used in a very similar fashion. Exposure to x-rays elevates electrons in the phosphor material to energy traps called F-centers, in numbers proportional to the incident x-ray intensity, forming a “latent image.” Subsequently, the exposed imaging plate is electronically “processed” with a mechanical-optical reader and scanning laser beam (a HeNe or diode laser of red wavelength). The trapped electrons absorb the laser energy and move out of the trap to a higher energy state. As the electrons fall back to the ground energy state, blue wavelength light photons are emitted. A light guide positioned close to the surface of the phosphor collects the photostimulated luminescence photons, converts the light energy to an electronic signal with the use of a sensitive, high gain photomultiplier tube, and produces a digital signal. By reflecting the laser beam off a rotating polygonal mirror, a line of image data can be obtained. Simultaneous mechanical translation of the phosphor plate in the optical stage allows the detector to be fully scanned, with a readout time of typically 45 to 135 seconds, depending on the size of the detector and the throughput of the reader. Image data is processed in three steps, including image scaling, contrast enhancement, and frequency enhancement. Image matrices typically comprise 100 to 200 mm detector elements, producing 8 to 40 MB of digital data, depending on field of view.
Over the past 2 years, CR technology has significantly improved in image acquisition speed and detector efficiency. Parallel line-scan systems can read a full 35 x 43 storage phosphor detector in as little as 10 seconds. Compared to conventional CR, dual side readout methods that consist of a transparent base imaging plate and two light collection guides can read a greater fraction of the photostimulated light to improve the DQE by as much as 50%. Adaptation to digital mammography has been achieved with a 50 mm laser spot size and 50 mm image pixels for both 18 x 24 and 24 x 30 cm imaging plates for high spatial resolution. Structured photostimulable phosphors are soon to be introduced, which promise enhanced detection efficiency and high spatial resolution.
The major attributes of CR are portability, selectable field of view, and integration with existing x-ray equipment. A single reader can service multiple rooms and therefore can be very cost-effective. Extra handling and time, however, are required to process the images, which reduces patient throughput, particularly if films are still being printed. Imaging plates and cassettes eventually wear out or are damaged due to mishandling or machine malfunctions, and must be replaced over time. Adding to operational costs are preventive maintenance of the CR reader equipment and periodic quality control. CR detectors, although not physically handled in normal use, attract dirt and dust from the environment and can be bent/scratched while being processed in the reader system. Image artifacts, including dust specks, streaks, and vertical lines (the latter resulting from dirt deposits on the light guide) are common. The requisite cleaning of screens as frequently as every 2 weeks has been the experience of many sites, including our own. Under normal use, the CR phosphor can become discolored or cracked, and the cassette housing can be damaged, requiring replacement. Typical detector/cassette life cycles depend on use; in an outpatient clinic, the lifetimes of cassettes and imaging plates are much longer than in a portable imaging environment. A minimum of 5,000 cycles is typical. Machine jams, removal from the cassette holder for specific procedures (eg, scoliosis imaging), and improper cleaning can subject the imaging plates to damage that requires replacement before the “guaranteed” number of exposures. Review of system operation, including field uniformity, spatial and contrast resolution, and distance measurement accuracy, should be performed at least annually or as needed after maintenance. FTE support for performing these duties, in addition to reviewing exposure trends and retake rates, must be considered for the overall implementation and operational costs.
CCD Cameras. CCD camera-based detectors represent a very cost-effective alternative to large field detector systems; however, the optical coupling and relatively small size of the active CCD camera elements require a significant demagnification of the light image arising from the scintillator/phosphor. A secondary quantum sink arises when the statistical integrity of the image data is dominated by the loss of light photons due to inefficient lens coupling, and not the number of x-rays absorbed in the scintillator. Thus, these systems are often less radiation dose-efficient but can be a very cost-effective alternative to CR and flat-panel DR systems. Advances in scientific grade CCD detector technology include larger area sensors and excellent low electronic noise properties to minimize the quantum sink problem. The sensor is radiosensitive, and typically requires a mirror to reflect the light from the scintillator to the lens/camera component. As a result, the housing depth can be considerable, on the order of 30 to 50 cm and greater, a consideration for system positioning in tight spaces. Other digital detectors using CCD sensors include fiberoptic or lens coupled mammography biopsy systems, and a linear array CCD with slot-scan geometry for a full-field-of-view digital mammography unit. The latter system does not require an antiscatter grid and is very dose efficient, but requires mechanical scanning and a long exposure time. Geometric aberrations and intensity variations introduced by optical lenses require flat-field processing on a periodic ingle, and multiple lens-coupled CCD camera systems are available. All electronic acquisition is a major benefit of the CCD camera system.
Flat-panel Detectors. Two types of flat-panel detectors are available. One type is based on a CMOS tiled array coupled to an x-ray scintillator. CMOS is essentially computer random access memory with a photodiode attached to each pixel, typically of 25 to 50 mm. To achieve a large field of view detector, individual CMOS modules are joined together. The other, more ubiquitous flat-panel detector is based on TFT arrays that were initially developed for flat-panel displays commonly found in laptop computers. TFT devices are hewn from silicon semiconductor sheets and consist of discrete small detector elements (typically 75 to 150 mm on a side), each of which consists of a charge collection storage capacitor and a switching transistor (the TFT). “Gate” and “Drain” lines are interconnected to each transistor and storage capacitor, enabling the electronic readout of the charges. An x-ray converter material is layered on the detector matrix. There are two general types of TFT x-ray detectors: indirect and direct, based on the conversion of absorbed x-ray energy in the converter into a corresponding electronic charge. In the indirect device, x-rays absorbed by a scintillator (eg, structured CsI phosphor) produce a proportional intensity of light photons that are captured by a “photodiode” sensor on each TFT detector element. Electrons produced in the diode are transferred to the local capacitor. In the direct acquisition detector, x-rays are absorbed by a thick photoconductor such as amorphous selenium (a-Se), which directly releases a corresponding number of electrons (and holes) that rapidly migrate under a large voltage placed across the selenium to the local storage capacitor. For both detector types, electronic image “readout” of the stored charge is accomplished by turning on the switches (the transistor gates) one row at a time, allowing transfer from the local capacitors along all data lines (columns) to charge amplifiers and subsequent digitization of the row data. This is repeated for each row until the pixel matrix is fully scanned and the “raw” digital image produced. In some systems, multiple groups of charge amplifiers allow readout in parallel from segmented parts of the array for faster acquisition. (Note that the CMOS detector functions in a similar fashion.)
A question often arises as to which method of detection (indirect vs direct) is superior. There is no clear-cut answer as both methods are effective in producing images of extremely high quality with high detection efficiency. In general, the indirect detector (eg, CsITFT array) has a faster readout time, and is more readily adaptable to dynamic fluoroscopic image sequences, while the a-Se photoconductor has a much higher intrinsic spatial resolution. The indirect detector can suffer from inefficient “fill-factor” effects, whereby the electronic components (eg, the TFT, storage capacitor, and readout lines) represent dead-space that reduces the overall x-ray detection efficiency. This penalty is more severe for smaller pixel areas, and therefore sets a limit on the “reasonable” resolution that can be achieved for reasonable detection efficiency, which is likely not to be much less than 70 to 80 mm. Direct acquisition devices are more susceptible to frequency aliasing, a situation resulting from the extremely high intrinsic spatial resolution of the photoconductor. Higher spatial frequency signals beyond the “Nyquist frequency” fold back into the lower frequency spectrum, and cause this “aliased noise signal” to be superimposed in the image. Charge trapping and recombination also reduce the optimal detector response. A significant research effort is under way to improve these devices and to implement others, with the overall goal to improve detection efficiency with high spatial resolution and excellent image quality at a cost that can be justified in the clinical setting. One example is the recent introduction of a portable DR flat panel detector that could work in the portable imaging domain now dominated by CR devices.
Flat panel detectors also require periodic evaluation of image quality to verify optimal performance. Direct digital detectors (CCD, CMOS, and flat-panel) have many flaws caused by dead pixel elements, defective column or row functionality, and subpanel variations in background levels, all of which conspire to increase visible flaws and introduce noise into the final image. Preprocessing correction techniques are successfully employed by all digital detector manufacturers to ensure the highest image quality; however, detector drift and aging of the detector electronics/components can result in the reduced performance of these systems over time.
Editor’s Note: Part II: A Comparison of Digital Detectors will appear in the July issue.
J. Anthony Seibert, PhD, is professor of radiology, University of California, Davis Medical Center, Sacramento.