Dentomaxillofacial Radiology (1999) 28, 1±5 ã 1999 Stockton Press All rights reserved 0250 ± 832X/99 $12.00 http://www.stockton-press.co.uk/dmfr Determination of the optimal conditions for dental subtraction radiography using a storage phosphor system DS Brettle 1 R Ellwood 2 and R Davies 2 1 NW Medical Physics, and 2 Dental Health Unit, Manchester, UK Objectives: To determine the optimal kvp and exposure conditions for digital subtraction radiography system using a storage phosphor system. Method: Signal-to-noise (SNR) measurements were acquired using a Digora system (Soredex, Helsinki, Finland) of large area, low contrast, clinically realistic details against varying degrees of background attenuation for both single exposure unsubtracted and subtracted images. These results were combined with a measure of estimated thyroid dose to derive a gure of merit (FOM) for the unsubtracted and subtracted images. Results: For both unsubtracted and subtraction radiography, an exposure at 50 kvp and 250 mgy produce the best overall FOM. However, using the system at the 60 kvp maximum a FOM at 1000 mgy for unsubtracted radiography and 500 mgy for subtraction radiography gave the best SNR performance. Conclusion: Operating parameters have been derived which allow the user to choose between optimising SNR and dose (50 kvp, 250 mgy for unsubtracted and subtracted radiography) or SNR alone (60 kvp, 1000 mgy for unsubtracted and 60 kvp, 500 mgy for subtracted), for the visualisation of clinically representative details using the Digora system. Keywords: radiography, dental; digital radiography, dental; subtraction technique; doseresponse relationship Introduction During the last two decades greater emphasis has been given to the early detection and prevention of oral disease. Preventive interventions in the early stages of a disease are often both clinically and nancially more e ective than those performed when it is more advanced. 1 The early detection of destructive periodontal disease and caries progression is, therefore, becoming increasingly important. One technique which has been widely investigated in dentistry, since it facilitates earlier detection of hard tissue pathology, is subtraction radiography. Anatomical structural noise is the predominant limiting factor in the diagnosis of small changes in dental hard tissues and its removal, using subtraction radiography, signi cantly improves the diagnostic yield compared with a single radiograph. 2 has Correspondence to: Mr DS Brettle, North West Medical Physics, Christie Hospital, Manchester M20 4BX, UK Received 2 June 1998; accepted 13 September 1998 been applied in a wide range of dental diagnostic tasks. For example, in periodontology it has been used to evaluate periodontal disease progression 3, periodontal surgery 4 and therapeutic agents. 5 Although subtraction radiography has also been used to detect changes in the status of caries 6 it has not found such widespread acceptance in cariology as in periodontology. One important factor in the implementation of subtraction radiography is the easy access to robust digital data. Direct digital systems that facilitate this are becoming increasing available. We have previously evaluated the clinical e ectiveness of a photostimulable phosphor (PSP) direct digital system compared with lm. 7 This type of system behaves in a di erent way to lm, most notably in respect to its wide exposure latitude. Therefore before it can be used for subtraction radiography an investigation into the optimal operating parameters is needed. Radiographic imaging is a noise-limited process where the quality of the radiographic image is determined not only by how well the signal is reproduced but also on the level of noise. Image quality improves with increased signal strength and
2 reduced noise. Therefore image quality should be quanti ed through some measure of the relationship between the two factors. Such a measure is the signalto-noise ratio (SNR) which in its simplest form is the ratio of signal amplitude to the standard deviation in the noise uctuations. 8 Signal-to-noise-ratio is preferable to subjective techniques for this type of study as it allows quantitative measurements which are intrinsically linked to the physical performance of the system. Additionally, details can be chosen which simulate the radiographic response of clinical features and as such re ect actual clinical performance. However, the physical performance of a system is only part of the overall picture and any system that is to be used clinically also needs to consider the risk to the patient. An acceptable measure of risk for an intra-oral radiographic examination would be the dose to the thyroid. The aim of this study was therefore to identify the kvp and radiation exposure for a digital subtraction radiography system using the Digora system which provides the optimal combination of image quality and patient safety. Methods Test object An aluminium step-wedge comprising six steps ranging from 0 ± 2 cm in 4 mm steps was constructed (Figure 1). A 2 mm thick lead stop, equal in area to a step, was included in the base of the step-wedge along one of the long sides. This, combined with the 0 cm step, ensured the detector received exposures ranging from 0 ± 100% of the entrance exposure. Four 5 mm diameter aluminium details on a 0.1 mm Al base plate were placed on each step. The thickness of the details was 0.1, 0.3, 0.5 and 0.7 mm. The test object was designed to cover the range and subtleties of densities that may occur in a clinical image whilst allowing measurement of SNR. Images were acquired for each combination of 50, 60 and 70 kvp and 60, 125, 250, 500, 1000 and 2000 mgy entrance exposures. The X-ray units and the exposure settings are shown in Table 1. The exposure was measured with a 4000 M+ dosemeter (Victoreen Inc., Cleveland, Ohio, USA). The operating range of the Digora (Soredex, Helsinki, Finland) system was de ned before use by reading an image plate uniformly exposed to the highest exposure and kvp to be used clinically. For this study the operating range was set at 70 kvp and 7 mas as determined by the Radiology Department, Manchester Dental Hospital. The same image plate was used for all exposures. Three identical images, containing both the stepwedge and the test details, were acquired for each setting. For the subtraction images a single base image i.e. an image of the stepwedge with the test details removed, was acquired at the same exposure settings as the original unsubtracted images and subtracted from the corresponding original image. All the subtraction images were aligned using the software described elsewhere. 9 Figure 1 Diagram of the test object and a corresponding radiograph at 60 kvp. For details of the test object see Methods Signal to noise ratio The signal-to-noise-ratio (SNR) measurements were calculated using a sampling aperture covering a 21621 pixel region. This region was large enough to cover the majority of the detail without including the edges. An identical aperture was also used for measurements on the background. The SNR was calculated using equation 1 where S Sig =mean grey level for the detail, Table 1 Systems and parameters used to achieve the range of radiographic factors Exposure times required to achieve indicated entrance dose(s) kvp Unit Filtration (mm Al) FOD (cm) ma 60 mgy 125 mgy 250 mgy 500 mgy 1000 mgy 2000 mgy 50 60 70 Oralix 50 a Heliodent MD b Heliodent MD b 2.13 2.3 2.3 a Gendex, Monza, Italy. b Sirona, Bensheim, Germany 25 35 35 7.5 7 7 ± 0.02 0.02 0.063 0.05 0.04 0.125 0.1 0.08 0.25 0.2 0.16 0.5 0.4 0.32 0.8 0.8 0.64
S BG =mean grey level for the background and s BG =standard deviation for the background SNR ˆ SSig S BG BG 1 The signal-noise ratio was measured for the 0.5 mm detail thickness at three di erent step thicknesses 4, 8 and 12 mm Al. This detail was chosen after all the images had been acquired because it was visible on all images and representative of subtle clinical changes. The step thicknesses were similarly chosen to represent a range of bone presentations which covered the typical clinical range. Eight separate SNR measurements were made for each detail choosing di erent background regions each time to include background variability. This was done for each of the three repeat images, so that 24 measurements were made in total for each detail. The mean SNR was recorded along with the standard deviation of the 24 SNR measurements as a measure of spread. FOM ˆ SNR 4SNR 12 2 Thyroid dose Where: SNR 4 =the SNR of the 0.5 mm detail on the 4 mm step SNR 12 =the SNR of the 0.5 mm detail on the 12 mm step This gure thus re ects the performance of each set of operating parameters across the range of potential clinical appearances relative to risk to the patient. 3 Thyroid doses To estimate the relative risk to the thyroid gland for each of the exposures the NRPB 10 organ data was used in conjunction with the XDOSE software (v2.1) (National Radiation Laboratory, Christchurch, NZ). The examination chosen was the lateral head radiograph because the thyroid gland was not directly irradiated in this view. Calculations were made for each of the imaging conditions used. Figure 3 50 kvp SNRs against entrance exposure for the 0.5 mm Figure of merit calculation We now have two measures, SNR and thyroid dose, to describe the performance of the subtraction system under simulated clinical conditions. The use of a gure of merit (FOM) allows a range of data to be presented in a single gure which is designed to increase with increasing SNR and decrease with increasing thyroid dose. The highest FOM is therefore the optimum balance between bene t and risk. The FOM used is developed from the FOM de nition given by Chakraborty and Barnes 11 and is de ned as: Figure 4 60 kvp SNRs against entrance exposure for the 0.5 mm Figure 2 Relative thyroid dose against entrance exposure for a lateral head radiograph without the thyroid gland in the primary beam Figure 5 70 kvp SNRs against entrance exposure for the 0.5 mm
4 Figure 6 The same measurements as Figure 3 after subtraction of Figure 9 Relationship of gure of merit (FOM) to thyroid entrance dose for unsubtracted images Figure 7 The same measurements as Figure 4 after subtraction of Figure 10 Relationship of gure of merit (FOM) to thyroid entrance dose for subtracted images Discussion Figure 8 The same measurements as Figure 5 after subtraction of Results Thyroid doses Figure 2 shows the relative thyroid doses for each examination calculated using the NRPB data. SNR measurements The results of the SNR study are shown in Figure 3 ± 8. The combination of the SNR and the thyroid dose data is shown as the FOM in Figures 9 and 10. The aim of this study was to determine the optimal radiographic parameters for subtraction dental radiography using a photostimulable phosphor direct digital system. To constrain the study, it was decided to restrict the parameters that could be varied to kvp and exposure. SNR was chosen as the parameter to optimise as it is a quantitative measure which has a strong basis in imaging theory. 8 The requirement to measure SNR a ected the design of the test object. More speci cally, the details had to be large enough to get relevant measurements, yet not so large as to be clinically unrealistic or prevent the acquisition of background measurements. Using a step-wedge ensured a range of radiographic contrasts consistent with those achievable in dental radiography. Similarly, the thickness of the details were chosen to be clinically relevant and span the threshold of detection. Aluminium was used for the details and the steps because it has a similar attenuation coe cient to tissues containing calcium. The operation of the Digora system is not as straightforward as a conventional radiographic system. It requires the operating range to be set before use and manipulates the data range accordingly. This process, combined with the complex mechanisms of photostimulable phosphor luminescence, makes the
interpretation of the Digora results more complex to analyse than a conventional lm-based system. The subtracted results are not as easy to interpret as the unsubtracted. Normally, when two temporally separated images containing small signal changes are subtracted, the signal-to-noise ratio will reduce due to the additive nature of uncorrelated noise sources. This is clearly not happening in the subtracted images. A possible explanation of this could be that the system is only weakly quantum limited, with stronger nonstochastic noise sources. This is indicated in the unsubtracted image by the large deviations in measured SNR. Sources of this noise could be scatter from adjacent steps or dirt on the image plates and/or the test object. Clearly, when the removal of anatomical noise is also considered, the subtracted system may well produce greater clinical SNRs than the unsubtracted images. The SNR results on their own do not give a total picture of the ideal exposure conditions. Obviously, increasing the exposure will increase the detail SNR but it will also increase patient dose. To determine the optimum exposure conditions for clinical use, this factor needs to be considered. This was achieved by estimating the dose to the thyroid for each of the exposure levels used. These doses were not absolute measurements but relative indicators. The results achieved were predictable, with higher exposures giving higher doses and greater penetration, i.e. higher kv, increasing the number of photons reaching the thyroid, and consequently the dose. Combining the results of the SNR measurements with the estimates of thyroid dose allowed a gure of merit to be derived which was sensitive to maximum SNR across the greatest range of steps (clinical representations) and risk to the patient. The results for FOM show that the best overall results for single unsubtracted exposure were obtained at 50 kvp and 250 mgy. However at 60 kvp and 1000 MGy the SNRs are better across the range, albeit for seven times the dose (Figure 9). The results at 70 kv were considerably worse than those at 50 and 60 kvp, suggesting that this may not be a suitable technique to chose. A similar case exists for the subtracted results with an optimum maximum at 50 kvp and 250 mgy (Figure 10). However signi cantly better SNRs were obtained at 60 kvp at the cost of increases in dose of up to four times. For both unsubtracted and subtraction radiography it appears that if optimum SNR/dose performance is desirable then 50 kvp should be used. If optimum SNR is required, then 60 kvp is preferable. In addition to the e ects of varying the exposure parameters, other factors also need to be considered. These include geometric alignment and its e ect on quanti cation of detail volume. Future investigations could also consider the e ects of dual energy subtraction which not only utilises the advantages of subtraction but increases relative radiographic contrast by exploiting the energy-dependent absorption properties of pathological features. Acknowledgements The authors would like to thank the sta of the Radiology Department, Manchester Dental Hospital, for their help 5 References 1. The Department of Health for England. An oral health strategy for England. London. Department of Health 1994 2. GroÈ ndahl K, GroÈ ndahl H-G, Webber RL. Digital subtraction radiography for diagnosis of periodontal bone lesions with simulated high-speed systems. Oral Surg Oral Med Oral Pathol 1983; 55: 313 ± 318. 3. Je coat MK, Reddy MS. A comparison of probing and radiographic methods for detection of periodontal disease progression. Current Opinion in Dentistry 1991; 1: 45 ± 51. 4. BraÈ gger U, Pasquali L, Kornman KS. Remodelling of interdental alveolar bone after periodontal ap procedures assessed by means of computer-assisted densitometric image analysis (CADIA). J Clin Periodontol 1988; 15: 558 ± 564. 5. Je coat MK, Reddy MS, Haigh S et al. A comparison of topical ketorolac systemic urbiprofen, and placebo for the inhibition of bone loss in adult periodontitis. J Periodontol 1995; 66: 329 ± 338. 6. Wenzel A., Halse A. Digital subtraction radiography after stannous uoride treatment for occlusal caries diagnosis. Oral Surg Oral Med Oral Pathol. 1992; 74: 824 ± 828. 7. Brettle DS, Workman A, Ellwood RP, Launders JH, Horner K, Davies RM. The imaging performance of a storage phosphor system for dental radiography. Br. J. Radiol. 1996; 69: 256 ± 261. 8. Workman A, Brettle DS. Review Article: Physical performance measures of radiographic imaging systems. Dentomaxillofac Radiol. 1997; 26: 139 ± 146. 9. Ellwood RP, Davies RM, Worthington HV. Evaluation of a dental subtraction radiography system. J Periodont Res. 1997; 32: 241 ± 248. 10 Hart D, Jones DG, Wall BF. Normalised organ doses for medical X-ray examinations calculated using Monte Carlo techniques. London HMSO. 1994 NRPB - SR262. 11. Chakraborty DP, Barnes GT. An energy sensitive cassette for dual-energy mammography Medical Physics. 1989; 16: 7 ± 13.