Texas Health Science Center at San Antonio, San Antonio, Texas, USA. Objectives: To determine physical properties of the Digora® digital intra-oral ...
ã
Dentomaxillofacial Radiology (2000) 29, 28 ± 34 2000 Macmillan Publishers Ltd. All rights reserved 0250 ± 832X/00 $15.00 www.nature.com/dmfr
Physical properties of a photostimulable phosphor system for intra-oral radiography HC Stamatakis*,1, U Welander1 and WD McDavid2 1
Department of Oral Radiology, Karolinska Institutet, Stockholm, Sweden; 2Department of Dental Diagnostic Science, University of Texas Health Science Center at San Antonio, San Antonio, Texas, USA
Objectives: To determine physical properties of the Digora1 digital intra-oral radiographic system (Soredex Orion Corporation, Helsinki, Finland) for dierent calibration settings and beam energies. Methods: The line spread function (LSF) and the modulation transfer function (MTF) were determined from radiographs of a slit. Noise power spectra (NPS) were determined from radiographs exposed to homogeneous radiation ®elds at 10, 50 and 100% of the calibration exposure for three tube potentials. All calculations were performed using relative values of exposure comprised of gray level, the signal at the photomultiplier tube and the ampli®ed signal in order to con®rm agreement between these dierent approaches. Noise equivalent quanta (NEQ) were calculated from the one-dimensional NPSs and the MTF. Detective quantum eciencies (DQE) were determined from the NEQs and representative values of the photon ¯uence. Signal-to-noise ratios (SNR) were calculated for dierent signal contrasts applying the NEQs. Results: The MTF of the system exhibited typical characteristics and falls to a value close to zero at the Nyquist frequency of about 7 cycles/mm. Noise as expressed by the NPS was found to be relatively low, i.e. about 1075 to 1076 mm2 depending on exposure and frequency. There was no signi®cant dierence between data obtained at dierent beam energies. The NEQ and hence the DQE were relatively high. DQE decreased with increased exposure. For exposures in the clinical range of the DQE reached a peak value of about 25%. SNRs are favorable. Conclusion: The physical properties of the Digora1 intra-oral system indicate that it is suitable for digital intra-oral radiography. Keywords: radiography, dental; digital radiography, dental; technology, radiologic
Introduction The Digora1, manufactured by Soredex Orion Corporation, Helsinki, Finland, was the ®rst digital intra-oral radiographic system based on photostimulable phosphor (PSP) technology. Its clinical imaging characteristics and performance have been described by a number of authors.1 ± 7 The relationships between exposure and gray level values and Signal-to-Noise Ratios (SNR) have been published.8,9 Although the physical properties have been reported in detail,10 the Digora1 system has been modi®ed since then. The dose
*Correspondence to: Dr HC Stamatakis, Department of Oral Radiology, Karolinska Institutet, Box 4064, S-141 04 Huddinge, Sweden Received 4 June 1999; accepted 11 October 1999
response of the system has been described recently.11 A reason for the limited information on the physical properties of the Digora1 may be due to the fact that there is no obvious relationship between exposure and gray levels in the ®nal radiograph.8,9,11,12 In order to produce radiographs with an optimum dynamic range, the Digora1 system is calibrated to the maximum exposure that will be used clinically. Careful selection of the calibration exposure is essential since exceeding the exposure range will result in inferior image quality.3,8 The useful exposure range is between 10 and 100% of the calibration exposure.8 During calibration a high voltage (HV) value is set on the photo-multiplier tube (PMT) of the reading device which is subsequently used in scanning the PSPs.10,11 Furthermore a pre-scan is performed before an
Photostimulable phosphor HC Stamatakis et al
exposed PSP is actually processed.11 In the pre-scan, the reading device sets a gain value on the ampli®er that will depend on the amount of energy stored in the phosphor plate. This is intended to ensure optimum gray levels in the ®nal radiograph. As a consequence of the calibration and pre-scanning settings, any direct relationship between exposure and gray levels is lost. When assessing the physical properties of any digital radiographic system quantitatively, the relationship between actual exposure and gray levels stored in the image ®le must be known. In the Digora1 this can be achieved if the internal HV value at the PMT and the gain value set at the pre-scanning are known. These values cannot be read by the commercial Digora software. However we have recently shown that using special software provided by the manufacturer, the gray levels, G, in the ®nal radiograph are linear functions of exposure.11 Thus, since 8 bit data are employed, 255-G may be used within any one radiographic image as a relative measure of exposure. Using this approach, the aim of the present work was to determine physical properties of the Digora system. Material and methods The Digora1 system comprising a processor board P4000-1 DXR-40 and a P4000 scanner connected to a PC and operating with software version 2.10 was used for the present study. This Digora system has an active area of 4166560 pixels each about 71 mm square. The exact size depends on calibration conditions.11 Two dental X-ray units were used to expose test radiographs, a Prostyle Intra (Planmeca Oy, Helsinki, Finland) operating at 60 and 70 kVcp and a Combex DX-907 (Takara-Belmont Co., Osaka, Japan) operating at 90 kVp. Half value layers were 1.92, 2.13, and 2.91 mm Al respectively. Exposures at the level of the PSP were measured with a calibrated ionization chamber (Radcal Model 1065 ± 6, Radcal Corp., Monrovia, CA, USA) and subsequently expressed in mC/kg. Signal and noise properties Calculations of the physical properties in the present work were performed using special software developed by one of the authors.13 ± 15 All signal and noise properties that are discussed and the respective calculation methods have been extensively described elsewhere.13 ± 27 Signal properties Line spread function and modulation transfer function Radiographs for determining the LSF and the MTF were exposed using a tantalum test object with a 10 mm wide slit (Atomic Products Corporation, Shirly, NY, USA). Exposures were made so that images of the slit covered the major part of the gray level range. The LSF and the MTF are signi®cantly dependent on pixel size but also on the scattering and absorption
properties of the phosphor layer and its thickness. The size of the pixels varies slightly between individual systems since the active area of the PSP employed at calibration is divided into ®xed numbers of rows and columns. Because of a small variation in the size of the active area of individual PSP plates, the pixel size will vary within about+5%. This variation will have an eect, although limited, on the LSFs and the MTFs that are valid for individual systems. It should be noted, however, that the pixel size set at calibration remains constant for the individual system and is valid for all successive radiographs irrespective of the size of the active area of exposed PSPs. The concept of the MTF assumes continuous data which is not the case with digital systems with a discrete pixel size. To circumvent the fact that the alignment between sine functions and pixels will aect their recorded amplitude, a so-called pre-sampling MTF is determined. This may be achieved by calculating the MTF from averaged LSF collected perpendicular along a substantial length of the image of the slit. The MTF thus obtained represents amplitudes of sine functions of dierent frequencies as they would appear before being sampled into pixels of the digital system. To determine representative LSFs and pre-sampling MTFs in dierent directions radiographs were exposed of the test object with a slit. The test object was positioned by hand with its slit inclined slightly o axis from a position approximately at 08, 458 and 908 to the long axis of the PSP. The focus-to-object distance was kept longer than one meter to minimize geometrical unsharpness. 255-G was employed as a relative measure of exposure. LSFs and MTFs for the three slit positions were determined as described previously.13,28
29
Noise properties Noise power spectrum The Digora1 system was calibrated at two exposures, 5.76 mC/kg and 28.82 mC/kg at the image plate at each of three tube potentials. Three sets of 20 radiographs were exposed to a homogeneous radiation ®eld at 10, 50 and 100% of the calibration exposure at each calibration. The radiographs were directly exposed to the homogeneous radiation ®eld without additional ®ltration. As in the case of slit images, 255-G was used as a relative measure of exposure. Gray levels were also transformed to signals at the PMT and also to ampli®ed signals in order to con®rm that all three approaches lead to identical results as would be expected.11 To transform gray levels into signals at the PMT and ampli®ed signals the following relationship was used:11 Sg dc where S is the signal, i.e. the relative light intensity incident on the PMT, g is the gain, i.e. the ampli®cation factor, d a constant set at the calibration, and c is a factor depending on calibration G 255 ÿ
Dentomaxillofacial Radiology
Photostimulable phosphor HC Stamatakis et al
30
conditions. Solving the above expression with respect to S and Sg gives the signal and the ampli®ed signal respectively. For each combination of the three tube potentials and the three exposure levels at the two calibrations, sets of image data were created using relative values of exposure comprised of gray level, the signal at the PMT and the ampli®ed signal. Two-dimensional NPSs were calculated over ten overlapping areas of 2566256 pixels or approximately 18.2618.2 mm in each of the 20 test radiographs resulting in ensemble averages determined from a total of 200 areas. These calculations were performed for all three exposures at each calibration and tube potential. Low spatial frequency was not removed. One-dimensional NPSs were calculated by means of rotational averaging of the two-dimensional functions.14 Combined signal and noise properties Noise equivalent quanta The NEQ was derived as described previously.14 Detective quantum eciency Representative values of the photon ¯uence at each tube potential and exposure were calculated from tabulated radiophysical data employing eective monoenergetic beam energies found from measured half value layers together with exposures in mC/kg.15 The photon ¯uences were 3.376105, 3.956105 and 6.366105 mC/kg.
about 2 and 4 cycles/mm where the functions show a slight increase. Since this increase is related to exposure, it is most marked in NPSs as the high calibration exposure. At exposures of 50 and 100% of the high calibration exposure the functions are basically the same (Figure 2b). As expected, NPSs calculated from gray levels of the ®nal radiographs, from the signals at the PMT and from the ampli®ed signals were identical. Values calculated from radiographs exposed at 60, 70 and 90 kVp were also practically identical. Insigni®cant dierences may be attributed to rounding o during numerical integration. Therefore, all illustrations give one arbitrarily selected tube potential of 70 kVp. Combined signal and noise properties Noise equivalent quanta NEQs for the two calibration settings are illustrated in Figure 3a,b. The functions have valleys at about 2 and 4 cycles/mm. At the peak of 1 ± 1.5 cycles/mm the plate captures about 60 000 photons/mm2 at the low calibration setting and 10% of the calibration exposure (Figure 3a) increasing to about 580 000 photons/mm2 at the high calibration
a
Signal-to-noise ratio The SNR was calculated as described previously. Results Signal and noise properties Signal properties Line spread function and modulation transfer function The LSF calculated from a radiograph exposed with the slit inclined approximately 458 is illustrated in Figure 1a. There were no signi®cant dierences between the LSFs calculated with the slit inclined 08, 458 and 908. Figure 1b illustrates the presampling MTF corresponding to the LSF illustrated in Figure 1a. Below a frequency of about 3 cycles/mm the MTF is higher than 0.3. It falls to a value approaching zero at the Nyquist frequency of about 7 cycles/mm. As in the case of the LSF there were no signi®cant dierences between the MTFs calculated from images with the slit inclined 08, 458 and 908.
b
Noise properties Noise power spectrum One-dimensional NPSs for the two calibration settings are illustrated in Figure 2a and b. In general, the noise was relatively low. It will, however, be noticed that with increased exposure the NPSs gradually became more irregular, especially at Dentomaxillofacial Radiology
Figure 1 (a) LSF calculated from a radiograph of a 10 mm wide slit. Raw data are marked as dots. (b) MTF calculated from the raw data shown in (a)
Photostimulable phosphor HC Stamatakis et al
31
a
a
b
b
Figure 2 One-dimensional NPSs for (a) the low and (b) the high calibration setting. The three functions represent from top to bottom exposures of 10, 50 and 100% of the respective calibration exposure 5.76 and 28.82 mC/kg, respectively. ÐÐ 100%, ± ± ± ± 50%, - - - - 10%
Figure 3 NEQs for (a) the low and (b) the high calibration setting. The three functions represent from top to bottom exposures of 100, 50 and 10% of the respective calibration exposure. ÐÐ 100%, ± ± ± ± 50%, - - - - 10%
setting and 100% of the calibration exposure (Figure 3b). Since the NEQ is eected by the square of the MTF, all functions fall o relatively rapidly from the peaks and reach values approaching zero at the Nyquist frequency.
general, SNRs are relatively high. Frequencies that appear at the threshold SNR should be of particular interest. Assuming that this threshold is passed between SNRs of 3 ± 517,23, Table 1 gives interpolated data of the frequencies where these limits occur.
Detective quantum eciency Figure 4a,b illustrate DQEs at the two calibration settings. At both settings the DQE decreases with increased exposure. The ranges are relatively wide. At the low calibration setting the peak of the DQE varies between about 14% at an exposure of 100% of the calibration exposure to about 24% at an exposure of 10% of the calibration exposure (Figure 4a). The corresponding values for the high calibration setting are about 5 and 18% respectively (Figure 4b). DQEs exhibit the same characteristics as the NEQs. This is due to the fact that the DQE is calculated from the NEQ by division by a constant, the photon ¯uence. Signal-to-noise ratio Typical SNRs are shown in Figure 5a,b. The graphs represent functions valid for exposures of 50% of the two calibration exposures. The ®ve functions in each illustration represent signal contrasts of 2, 4, 6, 8 and 10% of the mean signal. In
Discussion An analysis of the dose response function of the Digora system11 and the results of the present work show that determination of both signal and noise properties may be performed by using 255-G as a relative measures of exposure. We have con®rmed in agreement with previously published papers.1 ± 7 that the Digora1 system has properties that makes it suitable for digital intra-oral radiography. The pixel size is relatively large compared with other intra-oral digital systems and the pre-sampling MTF approaches zero at a Nyquist frequency which is as low as about 7 cycles/ mm. On the other hand, the MTF exhibits typical properties in the frequency range up to about 3 cycles/ mm where most diagnostic information is likely to be represented, and object details characterized by spatial Dentomaxillofacial Radiology
Photostimulable phosphor HC Stamatakis et al
32
a a
b b
Figure 4 DQEs for (a) the low and (b) the high calibration setting. The three functions represent from top to bottom exposures of 10, 50 and 100% of the respective calibration exposure. Note that the DQE decreases with increased exposure. ÐÐ 100%, ± ± ± ± 50%, - - - 10%
frequencies approaching the Nyquist frequency, for example small objects or relatively distinct edges, are still perceptible in Digora1 radiographs (Table 1). This is even true for details with a very low signal contrast. In fact, the calculated SNRs seem to underestimate detectability. A study involving viewers indicates that signals can be detected in Digora radiographs at such a low SNR as 2 (Stamatakis H.C. submitted for publication). Noise in Digora radiographs is relatively low and, consequently, the DQE is relatively high compared with the CCD sensors. (Stamatakis H.C. unpublished data). At high exposures irregularities appear in the NPSs. This may be explained by system noise, mainly structure but also electronic noise, added to photon ¯uctuations in the radiation ®eld and seems to be a general characteristic of PSPs.25,27 Figure 6 shows an enlarged contrast-enhanced detail from one Digora1 radiograph exposed to a homogeneous ®eld at a high exposure. It will be noted that the most apparent pattern in the image corresponds to a frequency of about 2 cycles/mm, where the most dominant irregularity in NPSs, NEQs and DQEs appears. Dentomaxillofacial Radiology
Figure 5 SNRs valid for exposures of 50% of (a) the low and (b) the high calibration exposure. The ®ve functions represent from top to bottom signal contrast of 10, 8, 6, 4 and 2% of the mean signal. ÐÐ 2%, ± ± ± ± 4%, ± - ± 6%, Ð - Ð 8%, - - - - 10%
Table 1 Approximate limit frequencies in cycles/mm that should be detectable in Digora1 radiographs. The two values represent limit frequencies when the threshold SNR is assumed to be 3 and 5 respectively. It should be noted that the Nyquist frequency is about 7 cycles/mm Signal contrast
Exposure relative to calibration exposure 10% 50% 100%
High calibration setting 10% 7.0/6.4 8% 6.8/6.0 6% 6.4/5.5 4% 5.8/4.8 2% 4.4/3.4 Low calibration setting 10% 6.2/5.3 8% 5.8/4.9 6% 5.3/4.4 4% 4.6/3.6 2% 3.2/-
7.0/7.0 7.0/6.7 7.0/6.3 6.4/5.5 5.1/4.0
7.0/7.0 7.0/6.8 7.0/6.4 6.6/5.8 5.4/4.2
7.0/6.4 6.8/6.0 6.4/5.6 5.8/4.8 4.5/3.4
7.0/6.7 7.0/6.4 6.7/6.0 6.2/5.2 4.9/3.8
A possible explanation for the fact that the NPSs calculated for the two highest exposures were practically identical may be that system noise contributes signi®cantly to the total noise. An eect of this property is that the DQE decreases with
Photostimulable phosphor HC Stamatakis et al
Figure 6 Enlarged, contrast-enhanced detail of a Digora1 radiograph exposed to a homogeneous ®eld at 100% exposure at the high calibration setting, i.e. 28.82 mC/kg. The most apparent pattern in the image corresponds to a frequency of about 2 cycles/mm where the most dominant irregularity in NPS, NEQ and DQE data occurs
increased exposure10,11,28 although the DQE is relatively high, especially when low exposures are employed. This should be the case in clinical work. Compared with Espeed ®lm, with a DQE in the range of 1 to 2% depending on exposure conditions,10 it should be possible to reduce the patient dose substantially by upto 90%. It is true that a reduction of the exposure by 90% compared with E-speed ®lm will result in increased noise. However, this noise will be dominated by quantum ¯uctuations while system noise will be limited. It seems likely that the diagnostic information will only be aected to a limited extent when low exposures are employed.3 In this context it is an advantage that the Digora1 system compensates for underexposure and displays radiographs with a suitable gray level range. However, the Digora1 system also compensates for overexposure which from a radiation
protection point of view may be a drawback. When the calibration exposure is not exceeded, the user has no absolute indication that an overexposure has been made. The present data on the MTF is consistant with previous results.10 However, although the NPS and NEQ in the study are of the same order of magnitude they do not show the irregularities we have found possibly because they smoothed their data. Furher DQEs in our study are better, probably re¯ecting the improved eciency of the newer version of the system. During the ®nal preparation of this study, Soredex Orion Corporation launched a further modi®cation of the Digora1 system with smaller pixels of about 63 mm compared with the model we used. This will improve the resolution, and the Nyquist frequency will increase to about 8 cycles/mm. Noise at comparable dose levels may be increased and therefore should be investigated further. Since there is no direct relationship between the technical properties of radiographic systems and their diagnostic yield, studies that include viewer performance are necessary in order to evaluate their clinical applicability. Nevertheless, objective measures of physical properties, such as those presented in this work, are always required for comparisons with other systems. The results of the present work on the Digora1 indicate that it has technical properties that make it suitable for intra-oral radiography. This view is borne out by clinical experience.
33
Acknowledgements The authors wish to express their sincere thanks to Heikki Kanerva and Olli Ojala, Research and Development Department, Soredex Orion Corporation, Turku, Finland, for providing us with the software that made this study possible, their kind assistance in interpreting data and their constructive advice regarding the manuscript.
References 1. Wenzel A, Borg E, Hinze H, GroÈndahl HG. Accuracy of caries diagnosis in digital images from charge-coupled device and storage phosphor systems: an in vitro study. Dentomaxillfac Radiol 1995; 24: 250 ± 254. 2. Borg E, GroÈndahl HG. Endodontic measurements in digital radiographs acquired by a photostimulable storage phosphor system. Endod Dent Traumatol 1996; 12: 20 ± 24. 3. GroÈndahl HG, Wenzel A, Borg E, Tammisalo E. An image plate system for digital intra-oral radiography. Dent Update 1996; 23: 334 ± 337. 4. Versteeg KH, Sanderink GC, Velders XL, van Ginkel FC, van der Stelt PF. In vivo study of approximal caries depth on storage phosphor plate images compared with dental x-ray ®lm. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1997; 84: 210 ± 213. 5. Borg E, GroÈndahl K, GroÈndahl HG. Marginal bone level buccal to mandibular molars in digital radiographs from chargecoupled device and storage phosphor systmes. An in vitro study. J Clin Periodontol 1997; 24: 306 ± 312.
6. Barbat J, Messer HH. Detectability of arti®cial periapical lesions using direct digital and conventional radiography. J Endod 1998; 24: 837 ± 842. 7. Wenzel A, Hinze H, HoÈrsted Bindslev P. Accuracy of radiographic detection of residual caries in connection with tunnel restorations. Caries Res 1998; 32: 17 ± 22. 8. Hayakawa Y, Farman AG, Kelly MS, Kuroyanagi K. Intraoral radiographic storage phosphor image mean pixel values and signal to noise ratio: eects of calibration. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1998; 86: 601 ± 615. 9. Huda W, Rill LN, Benn DK, Pettigrew JC. Comparison of photostimulable phosphor system with ®lm for dental radiology. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1997; 83: 725 ± 731. 10. Brettle DS, Workman A, Ellwood RP, Launders JH, Horner K, Darsen RM . The imaging performance of a storage phosphor system for dental radiography. Br J Radiol 1996; 69: 256 ± 261.
Dentomaxillofacial Radiology
Photostimulable phosphor HC Stamatakis et al
34
11. Stamatakis HC, Welander U, McDavid WD. Dose response of a storage phosphor system for intraoral radiography. Dentomaxillofac Radiol 1999; 28: 272 ± 276. 12. Freedman M, Steller D, Jafroudi H, Mun SK. Quality control of storage phosphor digital radiography systems. J Digit Imaging 1995; 8: 67 ± 74. 13. Welander U, Nelvig P, Tronje G, McDavid WD, Dove SB, Morner A-C, et al. Basic technical properties of a system for direct acquisition of digital intraoral radiographs. Oral Surg Oral Med Oral Pathol 1993; 75: 506 ± 516. 14. Welander U, McDavid WD, Morner A-C, Tronje G, Tokuoka O, Fuchihata H, et al. Absolute measures of image quality for the Sens-A-Ray direct digital intraoral radiography system. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1995; 80: 345 ± 350. 15. Yoshiura K, Stamatakis HC, Welander U, McDavid WD, Shi XQ, Ban S, et al. Physical evaluation of a system for direct digital intraoral radiography based on a charge-coupled device. Dentomaxillofac Radiol 1999; 28: 277 ± 283. 16. Rose A. The sensitivity performance of the human eye on an absolute scale. J Opt Soc Am 1948; 38: 196 ± 208. 17. Doi K, Holje G, Loo L-N, et al. MTF's and Wiener spectra of radiographic screen-®lm systems. Health and Human Services publication No. 82-8187. Rockville, MD:Food and Drug Administration, 1982. 18. Sandrik JM, Wagner RF. Absolute measures of physical image quality: measurements and application to radiographic magni®cation. Med Phys 1982; 9: 540 ± 549.
Dentomaxillofacial Radiology
19. Nishikawa RM, Yae MJ. Signal-to-noise properties of mammographic ®lm-screen systems. Med Phys 1985; 12: 32 ± 39. 20. Munro P, Rawlinson JA, Fenster A. Therapy imaging: a signalto-noise analysis of metal plate/®lm detectors. Med Phys 1987; 14: 975 ± 984. 21. Wagner RF, Brown DG. Uni®ed SNR analysis of medical imaging systems. Phys Med Biol 1985; 30: 489 ± 518. 22. Barten PGJ. Physical model for the contrast sensitivity of the human eye. SPIE III 1992; 1666: 57 ± 72. 23. Workman A, Cowen AR. Signal, noise and SNR transfer properties of computed radiography. Phys Med Biol 1993; 38: 1789 ± 1808. 24. ICRU Report 54. Medical imaging ± the assessment of image quality 1996, Bethesda, MD, USA. 25. Workman A, Brettle DS. Physical performance measures of radiographic imaging systems. Dentomaxillofac Radiol 1997; 26: 139 ± 146. 26. Bushberg JT, Seibert JA, Leindholt EM, Boone JM. The Essential Physics of Medical Imaging. William & Wilkins: Baltimore, 1994. 27. Bradford CD, Peppler WW. Performance characteristics of a Kodak computed radiography system. Med Phys 1999; 26: 27 ± 37. 28. Welander U, McDavid WD, Sanderink GCH, Tronje G, MoÈrner AC, Dove SB. Resolution as de®ned by line spread and modulation transfer functions for four digital intraoral radiographic systems. Oral Surg Oral Med Oral Pathol 1994; 78: 109 ± 115.