STATUS AND PROSPECTS OF DIGITAL DETECTOR TECHNOLOGY FOR CR AND DR Ulrich Neitzel Philips Medical Systems, Röntgenstrasse 24, D Hamburg, Germany

Transcription

1 Radiation Protection Dosimetry (2005), Vol. 114, Nos 1-3, pp doi: /rpd/nch532 INVITED PAPER STATUS AND PROSPECTS OF DIGITAL DETECTOR TECHNOLOGY FOR CR AND DR Ulrich Neitzel Philips Medical Systems, Röntgenstrasse 24, D Hamburg, Germany Projection radiography is in the middle of the transition from conventional screen-film imaging to digital image acquisition modalities, mainly based on imaging plates (computed radiography, CR) and flat-panel detectors direct radiography. Cassette-based CR has been available for the past 20 y, and constitutes the major part of direct radiography installations in hospitals today. direct radiography systems based on large-area amorphous silicon active matrix arrays are commercially available for the last 5 y and exist basically in two different types: with scintillators or photoconductors as X-ray conversion material ( indirect or direct type). direct radiography systems allow for improved image quality and/or dose reduction due to their high detective quantum efficiency and enable faster workflow because of instant image availability. However, new technology developments are improving the performance for CR systems as well, rendering it competitive to direct radiography in many practical aspects. Therefore, it is assumed that the CR and direct radiography systems will coexist for many years to come. This paper reviews the digital detector technologies and the possible future directions of development. INTRODUCTION In the past decade, digital radiography (direct radiography) systems have gained widespread acceptance for their clinical use. Many observers now expect the eventual replacement of all conventional screen-film X-ray imaging by digital electronic detectors, although the time frame for this transition may differ widely for different hospitals and countries. Whereas, for example, in Scandinavia the majority of hospitals is already fully film-less, in other parts of the world especially in the developing countries direct radiography is still uncommon. Direct radiography offers a number of clinical advantages compared with screen-film imaging. Some advantages are related to the availability of the images in electronic format, which allows easy storage, retrieval and transmission, even over large distances. The digital format also enables monitor (softcopy) reading of clinical images, which potentially reduces the running cost of an X-ray department compared with film-based reading. Other advantages are related to the separation of the image acquisition and image display stages in a direct radiography system. This feature allows to overcoming most of the limitations of film as an X-ray detector. Digital detectors now offer higher sensitivity, lower intrinsic noise and greater dynamic range than screen-film systems, which opens up new possibilities for dose reduction in clinical application. In this paper, a review is given on the status of the main digital detector technologies for (static) medical X-ray imaging together with some remarks on possible future developments. Corresponding author: ulrich.neitzel@philips.com DETECTORS FOR DIGITAL X-RAY IMAGING The X-ray detector is the key component of a direct radiography system. It has to fulfil several requirements: (1) the field size must be large enough for all radiographic applications; (2) the pixel size must be small enough to allow sufficient resolution; (3) the sensitivity must be high enough to allow low-dose operation; (4) the dynamic range must be large enough to cover a wide range of intensities; (5) internal noise sources must be small enough to preserve image quality; and (6) the readout time must be fast enough to allow efficient workflow. The specific requirements (i.e. what is enough?) will depend on the application area considered, for example, chest radiography or mammography. In this overview, we will limit ourselves to detectors in use for general applications, i.e. chest and skeletal radiography, which constitute the major part of the radiography workload in a hospital. Typical requirements for these application areas include a field size of up to cm, a pixel size <200 mm, a dynamic range exceeding 1:1000, a sensitivity that allows quantum noise-limited operation for detector doses in the order of 1 mgy, and a readout time of a few seconds. Detectors for direct radiography typically consist of three conceptual stages: a conversion or detection stage, whose function is to absorb the incoming X- ray photons and to convert them into a signal that is amenable to subsequent electronic handling, a ª The Author Published by Oxford University Press. All rights reserved. For Permissions, please journals.permissions@oupjournals.org

2 coupling stage, and a pixel readout stage that serves to sample the signal in both the space and the intensity domain (Figure 1). A number of different detector technologies have been developed in the past 20 y (1,2). However, the vast majority of clinical installations of direct radiography systems to date are based on only two generic types of detectors: the storage phosphor-based imaging plate (computed radiography, CR) and the large-area amorphous silicon flat-panel detector, which exists in two slightly different types, the direct conversion type, employing selenium as an X-ray converter material, and the indirect type, which X-ray converter Coupling stage Pixel read-out Photoconductor Scintillator Storage phosphor (Imaging plate) Optical glue Direct deposition Lens/fibre optics Electrode array Photodiode array CCD, CMOS Optical scan Mechanical scan Figure 1. Generic model of the different digital detector principles. An X-ray conversion layer absorbs the incoming X-ray photons and produces electrons or visible light photons, which are coupled to a readout stage that produces the image (pixel) signal. DIGITAL DETECTOR TECHNOLOGY FOR CR AND DR uses a scintillator screen. A third principle, which has found limited clinical acceptance, is based on the use of charge-coupled device (CCD) technology as readout stage (3). Owing to the small size of available CCD imagers, systems employing this principle must use either a demagnifying optical coupling or a scanning slot approach. As a consequence, these systems are limited either in their dose efficiency or in their application flexibility. Both flat-panel detectors and CCD-based systems are usually referred to as direct radiography systems (Figure 2). The selection of a specific detector type for clinical use is influenced by a number of factors, including performance, cost and operational properties. For example, if applicational flexibility and cost issues are predominant, cassette-based CR is probably the best choice, whereas high-image quality with dose reduction potential and an efficient workflow can best be obtained with an integrated flat-panel detector. From a physical point of view, the performance of the X-ray detectors is commonly quantified by the modulation transfer function (MTF) and the detective quantum efficiency (DQE). CR IMAGING PLATES Storage phosphor imaging plates as used for CR are well known since the first commercial CR system was introduced more than 20 y ago (4). A comprehensive review of the physics of this imaging technology has been recently given by Rowlands (5). CR systems use photostimulable phosphor screens that upon irradiation store a latent image, which subsequently can be readout as luminescence light in a point-by-point fashion using a scanning laser beam. The phosphor components used in the imaging CR Storage phosphor imaging plates Fuji, Agfa, Kodak, Direct Selenium Hologic, Toshiba Flat-panel (a-si) Indirect Gadox Cesium Iodide Canon GE, Philips, Siemens CCD Optical lens Gadox Cesium Iodide Imix, Wuestek, Imaging Dynamics Swissray DR Slot scanner Cesium Iodide Delft Diagn. Imaging Figure 2. Main types of digital X-ray detectors commercially available to date. 33

3 plates of the various manufacturers are usually of the BaFX:Eu 2þ type, where X means any of the halegonids Cl, Br or I or a mixture of them. The structure and composition of the imaging plates is quite similar to typical conventional intensifying screens, and it is therefore not too surprising that they are also similar in their basic imaging performance (MTF, DQE). Different balances of MTF (sharpness) and DQE (dose efficiency) can be obtained by adapting the thickness of the phosphor screen (standard and high-resolution screens). Standard resolution screens of different manufacturers may show slight differences in DQE, but in general their performance is more similar than dissimilar (6). Dual-side reading The most straightforward way to improve the X-ray absorption efficiency of an imaging plate is to increase its thickness. This has, however, two limitations: (1) the lateral spreading of the light in the phosphor layer (both for the stimulating laser light and the emitted light) will increase proportional to the layer thickness, impairing the resolution; and (2) the signal intensity will increase only marginally when the layer exceeds a certain thickness, because most of the light stimulated at greater depth will not reach the surface and therefore cannot be detected. The idea to circumvent the latter limitation is to make the substrate of the imaging plate transparent and to detect the stimulated light exiting both at the front and at the back side of the plate (7,8). This requires two light-collection systems, but still needs only one stimulating laser beam. CR readers incorporating this principle have been first developed for an integrated chest imaging system and for highresolution mammography. An extension of this principle to general purpose CR readers seems to be straightforward. Owing to the greater thickness of the phosphor layer and the different origin of the light exiting at the front and at the back side, the image signals from both sides have different spatial frequency content: the back-side image is more blurred than the frontside image. In order to balance the improvement of DQE and the preservation of spatial resolution a spectrally weighted addition of the two images is performed, in which the contribution of the front image dominates at high-spatial frequencies, whereas both images contribute about equally for low-spatial frequencies. As a result, an DQE improvement of 30 50% compared to single-side reading can be achieved with only minor effect on the MTF (7,9). Parallel reading (line scan) All CR systems on the market today still use the same basic readout principle as the first reader unit introduced in the early 1980s: the flying spot scanner. U. NEITZEL 34 A finely focused laser beam is moved over the imaging plate in a raster-like fashion to address each pixel individually for the stimulation of luminescence. This principle has a fundamental limitation in readout time and thus in throughput. Owing to the decay time of the phosphor luminescence (0.7 ms for typical storage phosphors), the dwell time per pixel cannot be 4 ms; otherwise, the resolution in the fast scan direction would be negatively affected (5). Therefore, the readout of a high-resolution 4k 2 image needs ¼ 64 s. A new development that promises to improve this situation is known as parallel reading or line scanning (10). A full-line of pixels is addressed and readout at each point of time instead of a single pixel. This requires a linear laser light source (consisting of many laser diodes and beam-shaping optics) and a linear detector array (photodiodes or CCD) over the full-width of the imaging plate. Both the light source and the detector are combined into a compact unit ( scan head ), which is then moved in one direction over the imaging plate. As line scanning reads several thousand pixels in parallel, a corresponding increase in readout speed would seem possible; however, there are other limitations, such as the available power of laser stimulation light and mechanical restrictions. Still, it seems possible to reduce the scanning time for a large-area, high-resolution image to <10 s. Other advantages of the line scan approach are as follows: a more compact reader unit that can be integrated into radiographic stands or tables and a more efficient readout process since the detectors can be brought closer to the imaging plate. Structured phosphors The CR imaging plates available to date are very similar to intensifying screens in their structure. They consist of phosphor grains of 5 mm embedded in an organic binder and deposited on a substrate. The screen is turbid, that is, it scatters light (both excitation laser and stimulated light) strongly and isotropically. Thus, the light diffusion limits the useful thickness of the phosphor layer. The advantages of structured, needle-like phosphors have been known for a long time, first as the input screens for image intensifiers and more recently as the converter layers for indirect type flat-panel detectors. The phosphor needles act as light guides, channelling the luminescence light preferentially along their axes. Therefore, needle phosphor layers can be made thicker without loosing too much resolution. Another advantage is the high packing density. A structured phosphor has 100% effective density, whereas a powder screen consists of only 60% phosphor and 40% binder. This means that the efficiency of a structured phosphor can almost

4 be four times higher than that of a powder screen with similar resolution properties. The reason why structured phosphors have not been used for the CR systems so far is probably 2-fold: (1) it is not easy to find compounds that can be grown in needle shape and at the same time exhibit appropriate photostimulation properties; and (2) needle crystalline phosphors tend to be brittle and/or hygroscopic, which makes their use in a cassette-based system difficult. Recently, CsBr:Eu 2þ has been proposed as a storage phosphor for the CR systems (11). When used in a thickness of 500 mm in a laboratory setting this phosphor has been shown to reach DQE values that approach those of CsI-based flat-panel detectors. However, no commercial products based on this phosphor are available so far. DIGITAL DETECTOR TECHNOLOGY FOR CR AND DR (a) MTF (b) 0.5 CR imaging plate RQA 5 Direct flat-panel Indirect flat-panel Spatial frequency [mm -1 ] RQA 5 FLAT-PANEL DETECTORS Flat-panel detectors based on amorphous silicon, active matrix readout arrays were introduced for radiographic imaging in 1997 and have experienced rapid clinical acceptance, despite their quite high costs (12,13). An excellent review of the physics and technology of these detectors has been given by Rowlands and Yorkston (14). Flat-panel detectors exist basically in two different variants, one using a semiconductor (selenium) and the other using a scintillator (caesium iodide or gadolinium oxysulphide) as conversion layer. The first type is also called direct because of the direct conversion of X-ray photons into electrons, whereas the second is called indirect since it employs an intermediate step of luminescent light generation. However, these names are just descriptive and do not per se imply advantages or disadvantages in imaging efficiency. In fact, the performance characteristics in terms of MTF and DQE show characteristic differences for commercially available direct (selenium) and indirect (caesium iodide) detectors (15 17). While the direct detector shows a markedly higher pre-sampling MTF than both the indirect detector and the storage phosphor imaging plate, its DQE is only second to that of the indirect detector, although still better than that of the imaging plate (Figure 3). The reason for this behaviour lies in the properties of the conversion process. The electric field in the selenium layer, necessary to separate the charges that are generated by the absorbed X-ray photons, inhibits the lateral diffusion of the charge cloud, thus preserving a high resolution or MTF. On the other hand, the high MTF causes noise aliasing, which together with the only moderate X-ray absorption efficiency of selenium results in lower DQE. There are also two different types of indirect flatpanel detectors: the most common employs needleshaped crystallites of CsI as phosphor material that DQE CR imaging plate Indirect flat-panel Direct flat-panel Spatial frequency [mm -1 ] Figure 3. (a) MTF and (b) DQE as a function of spatial frequency for the direct and indirect flat-panel detectors and for the CR imaging plates, measured according to the standard IEC at radiation quality RQA5. The data are taken from the publication by Illers et al. (16,17). allows to use a rather thick layer (500 mm) but has still good MTF. The curve shown in Figure 3 refers to this type. Another type uses a screen made of rare-earth phosphor particles (usually gadolinium oxysulphide, Gd 2 O 2 S) suspended in a binder, very similar to or even identical with conventional intensifying screens. For this screen type, lateral light diffusion is much more severe and the screen thickness must be kept much lower if a similar MTF to be obtained. Consequently, the X-ray absorption is lower and the DQE reaches values comparable with CR imaging plates or conventional screen-film systems only (18). Direct radiography systems based on the flat-panel detectors have a number of advantages compared with the CR systems. Since the detector unit is rather compact and does not need any moving parts for readout it can easily be integrated into radiographic stands and tables, making the examination unit selfcontained without the need to use cassettes and reading units. The systems can be interfaced with the hospital network and deliver their image data 35

5 directly to the reading workstations or the digital archive. Image readout is much faster for flat-panel detectors than for the standard CR systems, hence the image is available within a few seconds after the exposure and the detector is ready for the next image acquisition. One major clinical advantage of the flat-panel direct radiography systems, in particular for the indirect CsI type, is their high DQE compared with the CR and screen-film systems, which allows to operating the detector at a lower dose with maintained image quality. A number of technical and clinical studies have shown that up to 50% dose reduction is possible without loss of diagnostic image quality (13). New direct conversion materials Owing to its low Z-value, selenium (Z ¼ 34) is not ideal as an X-ray absorption material. The K-absorption edge lies at 13 kev and in the diagnostic energy range the mass energy absorption coefficient m en /r reduces rapidly with energy (Figure 4). Compared with the indirect conversion phosphors CsI and Gd 2 O 2 S, rather thick layers would therefore be required to absorb the X-ray quanta with equal efficiency at higher radiation energies. Other photoconductors are extensively researched for their use as direct X-ray converters, such as lead oxide (PbO), lead iodide (PbI 2 ), mercury iodide (HgI 2 ) and cadmium zinc telluride (CdZnTe) (19). All these have a much better photon absorption efficiency than selenium (and even CsI) for a given layer of thickness (Figure 5), and also show a higher charge yield, i.e. they produce a higher signal per absorbed photon, resulting in a better signal-to-noise Energy absorption coefficient [cm 2 /g] CsI Se Ga 2 O 2 S Photon energy [kev] Figure 4. Mass energy absorption coefficient m en /r as a function of photon energy for typical direct and indirect conversion materials. U. NEITZEL 36 ratio. However, further technology improvements are necessary to produce stable, highly homogeneous and defect-free layers of these materials with low dark current and low memory effect. Structured phosphors The phosphor screen necessary for the indirect-type flat-panel detector has to be balanced for absorption and resolution properties. Generally, the absorption increases with thickness but the resolution decreases at the same time due to lateral light spreading. For a given resolution, needle-shaped phosphors, such as CsI, allow thicker layers to be used than the conventional powder screens, resulting in higher absorption and better DQE. In the relevant energy range, the specific absorption properties of powder phosphors, such as Gd 2 O 2 S, are at least as good as those of CsI (Figure 4). Efforts have been made to suppress the isotropic light diffusion in Gadox screens by, for example, producing microchannel plates with reflecting walls in which the phosphor is loaded. Reduced pixel sizes In contrast to a CR imaging plate, a direct radiography detector has a fixed pixel size which is defined by the design of the readout array. Typical pixel dimensions for state-of-the-art active matrix arrays are in the range of mm. According to the Nyquist theorem this translates into limiting spatial resolutions of lp mm 1. Although this seems to be sufficient at least for general radiography, there is a trend to smaller pixel sizes particularly for mammographic applications. However, smaller pixels Photon absorption efficiency PbO PbI 2 HgI 2 CdZnTe CsI a-se Thickness (µm) Figure 5. Photon absorption efficiency as a function of layer thickness for several direct conversion materials. The curve for the scintillator CsI is given for comparison purposes.

6 reduce the fill-factor of the sensor array, that is, the proportion of the sensitive pixel area to the area of the pixel electronics forming switching elements and readout lines. The fill-factor reduction may lead to DQE losses, which are more pronounced for the direct detectors than for the indirect detectors (20). Owing to this reason, for amorphous silicon active matrix arrays there seem to be a practical lower limit of pixel size not much <100 mm. New types of readout arrays Virtually all large-area flat-panel detectors available to date use active matrix devices made of hydrogenated amorphous silicon (a-si:h þ ) as readout arrays. This technology has been developed for display devices (flat-panel displays) and makes it possible to produce large-area X-ray detectors with high radiation tolerance. However, a-si is less than ideal with respect to electronic properties and the minimum size of structures that can be obtained in this material. Another type of readout array that is investigated for X-ray detector applications is based on CMOS technology (21). CMOS imagers are active matrix devices as well, but produced on crystalline silicon wafers instead of amorphous silicon layers. They have the advantage of using the highly developed manufacturing processes of the semiconductor industry. CMOS sensors offer high fill-factor and resolution, good sensitivity and very fast readout. A special feature is the possibility to integrate on-chip electronics for each pixel, which can vastly improve the performance and may enable additional functions, such as photon counting or energy discrimination. CMOS arrays are limited in size to the available Si wafer size; typical wafer diameters to date are 6 or 8 in. For a future 12 in. diameter wafer, the maximum possible array size would be cm. Larger detector areas require tiling of several CMOS imagers. CMOS arrays may be used for both direct and indirect conversion detectors. Other directions of research for readout arrays aim at reducing the production costs, for example, by using a printing process instead of optical lithography or by employing organic semiconductors on flexible substrates (22). Considerable development work is still to be done before these approaches could reach the desired level of performance necessary for clinical application. CONCLUSIONS To date, various digital X-ray detectors are available for radiographic systems that fulfil all requirements for typical clinical applications. Differences between the detector types mainly affect their dose efficiency and workflow integration. The CR systems based on DIGITAL DETECTOR TECHNOLOGY FOR CR AND DR 37 storage phosphor imaging plates are comparable to screen-film systems in their dose requirements and similar to the cassette handling, but offer already the full advantages of digital image processing and data handling. Flat-panel detectors, in particular those using CsI as the input phosphor, exhibit substantially improved dose efficiency that enables a dose reduction by at least a factor of 2 when compared with conventional or CR imaging. Also, due to the fast readout possibility and the integration into the examination system, the direct radiography systems allow a faster workflow that helps to compensating for their still higher cost. Further technology developments can be expected that will improve the performance and usability of both the CR and direct radiography detectors. While many research activities are focused on the improvement of the physical performance parameters, the practical aspects of using the detectors in a clinical environment may change. To date, most CR systems are cassette-based while flat-panel detectors usually are integrated. However, portable, cassette-like flatpanel detectors are already available, as well as CR readers integrated into the examination systems. Therefore, it can be expected that both technologies will coexist for the foreseeable future. REFERENCES 1. Yaffe, M. J. and Rowlands, J. A. X-ray detectors for digital radiography. Phys. Med. Biol. 42, 1 39 (1997). 2. Orava, R. New detectors for radiology. Phys. Med. 15, (1999). 3. Båth, M., Sund, P. and Månsson, L. G. Evaluation of the imaging properties of two generations of a CCD-based system for digital chest radiography. Med. Phys. 29, (2002). 4. Sonoda, M., Takano, M., Miyahara, J. and Kato, H. Computed radiography utilizing scanning laser stimulated luminescence. Radiology 148, (1983). 5. Rowlands, J. A. The physics of computed radiography. Phys. Med. Biol. 47, R123 R166 (2002). 6. Samei, E. and Flynn, M. J. An experimental comparison of detector performance for computed radiography systems. Med. Phys. 29, (2002). 7. Arakawa, S., Ito, W., Kohda, K. and Suzuki, T. Novel computed radiography system with improved image quality by detection of emissions from both sides of an imaging plate. Proc. SPIE 3659, (1999). 8. Arakawa, S., Yasuda, H., Kohda, K. and Suzuki, T. Improvement of image quality in CR mammography by detection of emissions from dual sides of an imaging plate. Proc. SPIE 3977, (2000). 9. Fetterly, K. A. and Schueler, B. A. Performance evaluation of a dual-side read dedicated mammography computed radiography system. Med. Phys. 30, (2003). 10. Schaetzing, R., Fasbender, R. and Kersten, P. New high-speed scanning technique for computed radiography. Proc. SPIE 4682, (2002).

7 11. Leblans, P. J. R., Struye, L. and Willems, P. New needle-crystalline CR detector. Proc. SPIE 4320, (2001). 12. Kotter, E. and Langer, M. Digital radiography with large-area flat-panel detectors. Eur. Radiol. 12, (2002). 13. Völk, M., Hamer, O. W., Feuerbach, S. and Strotzer, M. Dose reduction in skeletal and chest radiography using large-area flat-panel detector based on amorphous silicon and thallium-doped cesium iodide: technical background, basic image quality parameters, and review of the literature. Eur. Radiol. 14, (2004). 14. Rowlands, J. A. and Yorkston, J. Flat panel detectors for digital radiography. In: Handbook of Medical Imaging, Vol. 1. Physics and Psychophysics. Beutel, J., Kundel, H. L. and Van Metter, R. L., Eds. (Bellingham: SPIE) pp (2000), ISBN Samei, E. and Flynn, M. J. An experimental comparison of detector performance for direct and indirect radiography systems. Med. Phys. 30, (2003). U. NEITZEL 16. Illers, H., Buhr, E., Bergmann, D. and Hoeschen, C. Measurement of the detective quantum efficiency (DQE) of digital x-ray imaging devices according to the standard IEC Proc. SPIE 5368 (2004). 17. Illers, H., Buhr, E. and Hoeschen, C. Measurement of the detective quantum efficiency (DQE) of digital X-ray detectors according to the novel standard IEC Radiat. Prot. Dosim. 114(1-3), (2005). 18. Samei, E. Image quality in phosphor-based flat panel digital radiographic detectors. Med. Phys. 30, (2003). 19. Kasap, S. O. and Rowlands, J. A. Direct-conversion flat-panel X-ray image detectors. IEE Proc. Circuits Devices Syst. 149, (2002). 20. Cunningham, I. A. Degradation of the detective quantum efficiency due to a non-unity detector fill factor. Proc. SPIE 3032, (1997). 21. Graeve, T. and Weckler, G. P. High-resolution CMOS imaging detector. Proc. SPIE 4320, (2001). 22. Street, R. A., Lu, J.-P. and Ready, S. R. New materials and processes for flat-panel X-ray detectors. IEE Proc. Circuits Devices Syst. 150, (2003). 38