DGZfP Proceedings BB 67-CD Paper 22 Computerized Tomography for Industrial Applications and Image Processing in Radiology March 15-17, 1999, Berlin, Germany Amorphous Selenium Direct Radiography for Industrial Imaging Peter K. Soltani, Daniel Wysnewski, Liberty Technologies, Inc., Conshohocken, PA, USA; Ken Swartz, Direct Radiography Corporation, Newark, DE, USA I. Introduction Filmless radiation imaging tools have been available for many years, primarily through the use of real time fluoroscopic (radioscopy) systems with simple analog video outputs. These systems are usually employed in high volume inspection applications where inspection speed is critical or when inspection automation is required. Fluoroscopic systems overcame their inherently poor contrast and spatial resolution through long integration times and the use of microfocus x-ray tubes to achieve projection magnification. The evolution of such systems led to the development of intensified CCD cameras, where the analog video outputs could be digitized, allowing for further analysis and enhancement. However, the intrinsic image quality of radioscopy systems has never matched that of conventional film, making them a poor choice for film replacement in most radiography applications. Over the past decade, several new digital radiography methods have emerged which can achieve sufficient levels of image quality to allow for film replacement in some radiographic imaging applications. These methods include what can be referred to as indirect digital radiography technologies, such as photostimulable storage phosphors employed in computed radiography (CR), phosphor and scintillator coupled amorphous silicon (a-si) arrays. These systems are characterized by the use of turbid media that convert radiation into visible light, which is, in turn, electronically detected and digitized to form an image. Because of optical scattering within the media, spatial blurring and increased noise can be encountered which degrade image quality as compared to film. However, these systems offer superior performance relative to conventional radioscopy systems, while exhibiting faster overall image display times as compared to film. More recently, our group has investigated and introduced a direct method for acquiring digital radiographs which eliminates optical blurring and reduces noise in the x-ray conversion layer. This approach employs an amorphous selenium (a-se) photoconductor deposited on top of an a-si thin film transistor (TFT) array. It has been demonstrated that this method produces image quality which exceeds indirect digital methods and is comparable to fine grain radiographic film. This paper will describe the particular structure and operating principle of this device, along with performance characteristics relevant to industrial imaging applications. II. Comparison of different radiography methods Before discussing the particulars of the a-se detector array, it is worth reviewing other digital filmless radiography methods, namely those that employ an indirect x-ray to digital conversion process. The different methods are summarized in Figure 1. DGZfP Proceedings BB 67-CD 123 Paper 22
CR systems have been in use for medical radiology since the early 1980 s and more recently for industrial imaging. This approach employs an imaging screen containing photostimulable storage phosphor crystals. The phosphor grains are coated on a flexible substrate and store the incident radiation energy in the form of trapped electrons - thus the name storage phosphor. The trapped electrons can be excited by laser stimulation to produce visible light from small pixel elements of the phosphor media. The light is detected electronically, digitized, and the digital signal level stored in computer memory. Once all available elements on the phosphor media have been stimulated and signal levels stored, the information is computed into a 2- dimensional image. As recent work has shown 1, CR systems are primarily limited by the blurring within the storage phosphor media and produce image quality comparable to coarse grain NDT radiographic films. Another method introduced over the past two years employs a fluorescent media, such as a medical phosphor intensifying screen or CsI crystal, to convert x-rays to light, which are then captured by an a-si photodiode array. As with CR, this process is primarily limited by the blurring within the phosphor or CsI, and though it will produce image quality somewhat better than CR systems, it still falls short of NDT film. The a-si arrays can exhibit a wide latitude as well as wide energy response, and are suitable for applications where lower image quality than medium speed NDT film is required. Figure 1. Comparison of indirect and direct digital radiography methods. DGZfP Proceedings BB 67-CD 124 Paper 22
III. a-se array description As shown in Figure 1, the primary difference between indirect methods and the a-se array is that x-rays are directly converted to charge in the latter. Due to the application of an electric field across the Se layer, it can also be seen that charge transport is constrained along vertical field lines, significantly reducing lateral information spread. This means that the detector array output closely matches the input radiation signal shape, with array resolution limited by the device geometry. The way in which the a-se array achieves the sharp input-output relationship results from its unique device structure 2,3, shown in Figures 2 and 3. The device employs a uniform and continuous a-se layer vapor deposited over a charge collection and field-effect transistor (FET) array forming a full 35 cm x 43 cm active area. The Se layer has a dielectric and top electrode layers applied to form a capacitor structure where electrons generated during exposure are collected under an applied bias. The charge collection and integration is accomplished with a 129 micron square charge collection (or so called mushroom ) electrode deposited over the FET array. The resulting pixel pitch is 139 microns, yielding a geometric fill factor of 86%; this is an important device parameter, which helps it achieve high charge collection efficiency, and thereby high image signal-to-noise ratios. The charge collection electrode is attached to a signal storage capacitor, which is in turn attached to the drain side of the FET. Figure 2. Cross section of the multi-layer a-se structure. DGZfP Proceedings BB 67-CD 125 Paper 22
Figure 3. Detail of pixel structure. Another key feature of the present device is that its structure, while producing a very high collection efficiency of the generated charge, prevents damage to the device at high exposure levels. As the device is exposed to radiation, the collection of charge reduces the applied bias. Charge collection efficiency is bias dependent, meaning that the electron-hole pair generation and separation efficiency is reduced. This prevents over charging of the signal storage capacitors to prevent damage to the structure at high exposures. IV. Image formation and readout As the array is exposed to ionizing radiation, electron-hole pairs are created within the a-se layer. The applied bias produces a 10V/micron field across the Se layer to separate the charge such that the holes are propagated to the charge collection electrode and the electrons collect at the top electrode interface. Additionally, the applied field prevents significant lateral spreading of the generated charge during exposure. Theoretical and experimental analysis of a-se layers under an applied field show that extremely high intrinsic spatial resolutions can be achieved 4. Unlike phosphor coupled a-si and CR systems whose resolution is phosphor limited, in the present case, the resolution is limited by the pixel geometry, not the a-se. Thus, very high modulation transfer function (MTF) would be expected in a-se arrays. During exposure, the FET s are in the off mode while charge is collected at the signal storage capacitor. At the end of the exposure, a positive pulse is applied to the FET gates starting with the first gate, G1. The data line, D1, connected to the FET source collects the charge from the drain, as shown in Figure 4. The signal charge is propagated to charge amplifiers connected to the data lines. The signals from the row are multiplexed and propagated to a 14 bit analog-to-digital converter and stored in computer memory. The readout continues row by row until all pixels have been read. DGZfP Proceedings BB 67-CD 126 Paper 22
Figure 4. Top view of device. At the end of the readout cycle, a charge erase cycle is applied to prepare the array for the next exposure. This removes any residual charge from the interfaces and any charge in deep trapping states of the a-se, thus preventing ghost images between successive exposures. In order to maximize the signal to noise, and thus image quality, the device also performs background corrections. The first involves creating a correction look up table (LUT) which stores a reference image of the detector array response to a flat x- ray exposure. This LUT stores the resulting image containing variations in x-ray source uniformity, as well as slight variations in array response at different energies and exposure levels. These variations are due to slight differences in the gain of each pixel and charge amplifier output. Generally, it has been found that a calibration LUT can be created once for a given set of exposure conditions, and updated if the exposure conditions change significantly. In addition to the LUT, the system also acquires dark field reference images before and after each exposure. Amorphous Se, like many semiconductors, generates a small dark current at room temperature. Since the amount of dark current will vary with time during and between exposures and array temperature, the acquisition of dark field images before and after exposure is employed to simulate the background noise during exposure. The background correction task is programmed into the device firmware and is performed automatically. The two reference images are averaged to estimate the magnitude of the background during the exposure and the exposure image is corrected. The result of this correction is to improve system contrast sensitivity and consistency between exposures. DGZfP Proceedings BB 67-CD 127 Paper 22
V. a-se array performance In order to evaluate the usefulness of the a-se device for non-destructive testing applications, the following performance characteristics were evaluated: - Exposure and energy response - MTF - Penetrameter sensitivity - Speed - Throughput Figures 5 and 6 show the response of the array to different metal thicknesses and energies at 150 cm source to detector distance and approximately 12 ma-second exposure. The results are presented in digital counts as a function of metal thickness at 50, 100, and 200 kvp. They indicate that the system produces a large change in response for the different metal thickness, with each digital count equal to approximately 2000 charges. Thus, the system can simultaneously achieve both wide latitude and density contrast. Data for energies up to 450 kvp have also been generated and reported in this conference s proceedings by Schneberk, et. al. 100000 Digital Counts 10000 1000 100 kv 50 kv 100 0 0.2 0.4 0.6 0.8 1 1.2 Al Step Thickness (inches) Figure 5. Aluminum step wedge. Characteristic curve generated at 150 cm source to detector distance, 24 ma-second exposure. DGZfP Proceedings BB 67-CD 128 Paper 22
100000 10000 Digital Counts 1000 200 kv 100 100 kv 10 0 0.2 0.4 0.6 0.8 1 1.2 SS Step Thickness (inches) Figure 6. Stainless steel step wedge. Characteristic curve generated at 150 cm source to detector distance, 24 ma-second exposure. The system MTF was determined by exposing the array to a 50 micron thick Pb foil at 100 kvp, 36 ma-seconds, and 150 cm source to detector distance; 0.5 mm of Pb filtration was also applied to the x-ray tube to attenuate its intensity. The edge response of the foil was determined by averaging 33 adjacent lines across the foil. The line spread functions (LSF) was calculated by taking the derivative of the edge response, as shown in Figure 7. The MTF was obtained by taking the Fourier transform of the LSF. Digital Counts 13000 12500 12000 11500 11000 10500 10000 0 0.5 1 1.5 2 2.5 Position (mm) dc/dx 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0-0.2 0.5 1 1.5 2 2.5 Position (mm) Figure 7. Edge spread response and LSF for a-se array. In order to compare the MTF of the a-se array with other digital radiography methods, similar data were generated for our CR system, and published data obtained for amorphous silicon photodiode arrays. The results are shown in Figure 8. The data for a-si are from an array employing fine grain fluorescent phosphor, which is expected to exhibit near the highest MTF possible for indirect arrays 5. The data of Figure 8 clearly show that the a-se exhibits a very high contrast, as compared to CR DGZfP Proceedings BB 67-CD 129 Paper 22
and indirect arrays. The MTF data for a-se are much closer to that of NDT film than any of the other digital methods shown. These results confirm that both the spatial and contrast resolution of indirect methods are severely limited by the optical blurring which take place in the fluorescent screens, whereas a-se exhibits no optical blurring. 1 MTF 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 fluorescent phosphor storage phosphor a -Se array 0 0 1 2 3 4 5 frequency (Lp/mm) Figure 8. MTF for medium grain CR, fine grain medical fluorescent phosphor screen, and amorphous selenium array. In addition to MTF, the penetrameter sensitivity of the a-se array was also evaluated. This was done by exposing a 25 mm thick stainless steel bock to x-rays at 200 kvp with 1%, 2%, and 4% plaque type penetrameters. The exposure was varied and the resulting image optimized to evaluate the visibility of the penetrameter holes. The exposure at which various penetrameter holes could be seen was noted. These results are summarized in Figure 9 for a-se, along with data for CR, medical fluorescent phosphors employed with a-si indirect arrays, and NDT film 1. These show that the a-se array achieves penetrameter sensitivities comparable to fine grain NDT film, but at about 1/50 th the exposure. Furthermore, the data show that a-se exhibits superior performance relative to both CR and indirect arrays employing fluorescent phosphor screens. VI. Areas of application It is important to note that even though it is possible to achieve better than 2% penetrameter sensitivity with both CR and indirect arrays, these methods still fall short of the sensitivity achieved with conventional NDT film, as well the a-se array. It is also important to note that simply achieving a 2% penetrameter sensitivity does not necessarily mean that flaws can be detected with equal probability. In fact, penetrameters were not designed to provide a measure of flaw detection efficiency, but rather to ensure the quality of the radiographic procedure when using film. However, since film has intrinsically high resolution which exceeds what is needed in many imaging applications, parameters such as film MTF do not need to be considered. Unlike film, however, digital detectors do have a finite resolving DGZfP Proceedings BB 67-CD 130 Paper 22
capability, which may be on the order of the flaw size being investigated. Therefore, simply employing a 2% sensitivity to assess the performance of a digital system is not sufficient. The system MTF must be evaluated along with penetrameter sensitivity to assess the performance of different digital systems and their ability to detect flaws of a particular size and density. The MTF data of Figure 8 clearly show very significant differences in resolving capability between the various systems, which directly impacts the ability to detect flaws. Another interesting point is that the use of fluorescent screens which might otherwise improve exposure speed are typically forbidden for NDT use; the reason is the blurring which takes place, resulting in loss of both resolution and contrast needed to detect flaws. This means that care should be taken when employing indirect methods; only those applications where high spatial resolution and contrast sensitivity are not needed should be targeted for the use of indirect methods. Conversely, the a-se array offers a suitable choice for applications where high contrast sensitivity is needed, and resolution requirements do not significantly exceed the pixel pitch. 10 a -Se % EPS indirect digital NDT film 1 1 10 100 1000 10000 Relative Exposure Figure 9. Penetrameter sensitivity for different imaging methods; 200 kvp, 150 cm source to detector distance, 25 mm stainless steel. In summary, it has been found that though the pixel pitch of a-se direct conversion flat panel arrays is limited to 139 microns, the contrast sensitivity is very high and comparable to fine grain NDT film. Furthermore, it is found that because of the systems high signal-to-noise ratio, the system can detect the presence of flaws smaller than the pixel pitch as long as sufficient object contrast is produced. Overall, the a-se array offers significant performance improvements over indirect digital methods, making it useful for a wider range of film replacement applications. DGZfP Proceedings BB 67-CD 131 Paper 22
VII. References 1. R. Kochakian, B. Vassen, and P. Willems, Application Limitations for Digital Radiography, Proceedings of the 1 st Pan-American Conference, September 14-18, 1998, pp. 15. 2. D. L. Lee, L. K. Cheung, E. F. Palecki, and L. S. Jeromin, A Discussion of the Resolution and Dynamic Range of Se-TFT Direct Digital Radiographic Detector, SPIE Vol. 2708, Physics of Medical Imaging, pp. 511, 1996. 3. D. L. Lee, L. K. Cheung, L. S. Jeromin, A New Digital Detector for Projection Radiography, SPIE Vol. 2432, Physics of Medical Imaging, pp. 237, 1995. 4. W. Que and J. A. Rowlands, X-ray Imaging Using Amorphous Selenium: Inherent Spatial Resolution, Med. Phys. 22, 364-374 (1995). 5. R. L. Weisfield, R. A. Street, R. Apte, A. Moore, An Improved Page-Size 127 mm Pixel Amorphous Silicon Image Sensor for X-ray Diagnostic Medical Imaging Applications, SPIE Proceedings, Medical Imaging 1997, 22-28 February 1997. DGZfP Proceedings BB 67-CD 132 Paper 22