Dentomaxillofacial Radiology (2003) 32, 390396 q 2003 The British Institute of Radiology http:/ /dmfr.birjournals.

org

RESEARCH

Comparison of standard and task-specic enhancement of Digoraw storage phosphor images for approximal caries diagnosis
ndahl3, A Wenzel4, FC van Ginkel2, B Kullendorff 5, A Mystad*,1, DB Svans1, PF van der Stelt2, H-G Gro H Hintze4 and TA Larheim1
Department of Maxillofacial Radiology, Faculty of Dentistry, University of Oslo, Norway; 2Department of Oral Radiology, Academic teborg University, Sweden; Centre for Dentistry, Amsterdam, The Netherlands; 3Department of Oral and Maxillofacial Radiology, Go 4 Department of Oral Radiology, Royal Dental College, University of Aarhus, Denmark; 5Department of Oral Radiology, University of Malm and Department of Radiology, Trelleborg County Hospital, Sweden
1

Objectives: To compare approximal caries detection on Digora storage phosphor images pre-enhanced with the automatic caries-specic Oslo enhancement procedure (Oslo-enhanced method) and storage phosphor images individually enhanced by observers particularly experienced in digital imaging using standard brightness and contrast functions of the Digora system (Digoraenhanced method). Methods: Seven staff members from four oral radiology departments rated 240 approximal surfaces for caries with regard to lesion depth in the inner and outer half of the enamel and dentine, using a 5-point condence scale. The observations were validated histologically. A receiver operating characteristic (ROC) analysis and an analysis of variance with three dependent variables (observer condence, observer signed error and observer absolute error) were performed. Results: The most evident difference between the methods as elucidated by the ROC analysis was the highly signicant smaller interobserver and intraobserver variance with the Oslo-enhanced method for all but one observer. The methods were not different with regard to average diagnostic accuracy (Az values) as tested with paired t-tests, and there was no correlation between Az across methods. The method and lesion main effects, as well as the lesion by method interaction effect, were multivariately signicant (P , 0.001) in favour of the Oslo-enhanced method. On a univariate level, the method main effect was not signicant for the absolute observer error (P 0.330). All other univariate effects were signicant (P , 0.001). Conclusions: The accuracy of approximal caries detection with Digora storage phosphor images pre-enhanced and images individually enhanced was similar, but interobserver and intraobserver variability improved with the Oslo-enhanced method. Since image manipulation is not performed with the Oslo-enhanced method, the increased speed of the diagnostic procedure combined with the improved observer variability would probably be even more pronounced for the general dental practitioner. Dentomaxillofacial Radiology (2003) 32, 390396. doi: 10.1259/dmfr/76382099 Keywords: radiographic image enhancement; diagnosis; dental caries; photostimulable phosphor; computer-assisted Introduction A number of studies have analysed the accuracy of radiographic approximal caries diagnosis comparing
*Correspondence to: Associate Professor, Dr odont. Anne Mystad, Department of Maxillofacial Radiology, Faculty of Dentistry, University of Oslo, Box 1109, Blindern, 0317 Oslo, Norway; E-mail: amoystad@odont.uio.no This study was supported by the Research Council of Norway, Ref. no. 115606/320. Received 21 February 2001; revised 13 May 2003; accepted 1 June 2003

conventional lm and digital techniques based on solidstate or photostimulable storage phosphor technologies, using images with optimal exposures. Basic image processing algorithms, mainly those altering brightness and contrast, have made it possible to achieve the same level of diagnostic accuracy in images initially of less than optimal contrast as in high quality images, but no studies have been able to demonstrate that approximal caries can

Digital caries diagnosis A Mystad et al

391

be better diagnosed on digital images than on conventional lm, with two exceptions (for a review see Wenzel 19981). In both these studies, Digora images were compared with lm.2,3 The digital images were transferred to another platform and were enhanced with an algorithm named the caries-specic Oslo enhancement procedure.3 This algorithm was applied to each image either manually2 or automatically3 before the observers made their assessment, leading to improved diagnostic accuracy as well as improved observer variability. The observers were not allowed to use ad libitum functions that are available with the digital systems in either of the studies. Images enhanced with the caries-specic Oslo-enhancement procedure have not been compared with those individually enhanced by observers using basic image processing functions (brightness and contrast) of the Digora system. Thus, the aim of this approximal caries study was to compare the diagnostic performance of Digora storage phosphor images pre-enhanced with the caries-specic Oslo enhancement procedure and images individually enhanced using standard brightness and contrast functions of the Digora system. The null hypothesis is that there is no difference between pre-enhanced and individually enhanced images. Materials and methods Extracted human teeth (59 premolars and 61 molars) with both sound and carious approximal surfaces were radiographed with the Digora storage phosphor system (Soredex Medical Systems, Helsinki, Finland). The teeth were placed with their approximal surfaces in contact according to a previously described technique.2 Exposures were the same as those for E-speed Plus lms and the irradiation geometry was individually adjusted for each pair of contacting approximal surfaces. A total of 120 images was obtained and then copied into two sets of images. One was kept in its original state (Figure 1a) while the other was

transferred to a different platform where image processing was performed automatically with Image/Pro Plus, version 1.2 for Windows (Media Cybernetics, Silver Spring, MD). The caries-specic Oslo enhancement procedure uses a combination of image processing operations: sharpening, contrast stretching, gamma correction and horizontal image distortion.3 The enhancement procedure was empirically tried out in a laboratory at the Department of Maxillofacial Radiology, University of Oslo, and a series of investigations were planned. The complete formula of the algorithm will not be released until all the planned investigations have been accomplished. Therefore, details of the caries-specic Oslo enhancement procedure have not been given here or in previously published studies.2,3 The automatically enhanced images are hereafter named Oslo enhanced images. They were sent to the observers on CD-ROMs and could not be altered (Figure 1b). The images in their original state (Figure 1a) were distributed by diskette and were imported into a Digora database. They were then enhanced ad libitum by each observer with the aid of Digora software 2.0 to a quality considered optimal for caries diagnosis (Figures 1c f ). The individually enhanced images are hereafter named Digora-enhanced images. Seven staff members from oral radiology departments in teborg, Malmo and Aarhus with experiAmsterdam, Go ence in digital radiology and caries diagnosis were used as observers. They evaluated the two methods, i.e. the two sets of images, with an interval of at least 2 weeks between the sessions to prevent any learning effect. The images in their original state were always evaluated before the automatically enhanced images, because we did not want the observers to have the possibility of using the automatically enhanced images for comparison. All images were displayed on 17" monitors, 1024 768 pixel high colour resolution, for which contrast and brightness were adjusted with the aid of a provided test image. Detailed instructions were given before each viewing session.

Figure 1 An example of an original image (a), the same image processed with the caries-specic Oslo enhancement procedure (b), and examples of the image individually enhanced with the Digora software 2.0 (c f )
Dentomaxillofacial Radiology

392

Digital caries diagnosis A Mystad et al

In each set of images, 240 approximal surfaces were rated for caries on a 5-point condence scale: 1, denitely not carious; 2, probably not carious; 3, questionable; 4, probably carious; 5, denitely carious. If a rating of 4 or 5 was registered, the deepest part of the lesion had to be described as being in: 1, outer half of enamel; 2, inner half of enamel; 3, outer half of dentine; 4, inner half of dentine. For scoring the Oslo-enhanced images, a mouse-driven program was constructed by one of the authors (DBS) using Authorware professional version 2.0 for Windows (Macromedia Inc.600 000 Station, San Francisco, CA) that automatically stored the data entered by the observers. Histological examination of all surfaces was made as described elsewhere.2,4 Of the 240 surfaces, 17 had to be excluded from the study because the histological sections were destroyed during preparation. The distribution of sound surfaces and carious lesions conned to enamel and dentine is illustrated in Table 1. Data analysis A receiver operating characteristic (ROC) analysis5 was performed using the ROCFIT computer program (obtained from Charles Metz, Department of Radiology, The University of Chicago, IL). The area under the binormal ROC curve (A z) was calculated for each observer independent of lesion depth. The variance reported by the ROC analysis, i.e. mean-squared distances from the 223 observation points to the ROC curve, were used in F-tests of intraobserver variance between methods. The Az values for each method were used in a paired t-test, correlation test and in a F-test of interobserver variance between methods. As a ROC analysis does not give information about lesion depth determination, the data set was further analysed with the aid of the Statistical Package for the Social Sciences (SPSS), version 8.2 for Windows. From this package the General Linear Model (GLM) program was used. Results were considered signicant at P , 0.05. The observers radiographic caries data were compared with the histological reference data to obtain three variables, namely observer condence, observer signed error and observer absolute error. Observer condence was scored as 0, 1 or 2 depending on the observers condence rating related to the histological presence or absence of caries. The condence score became 0 when the observer had given a condence rating of 3, presence of caries questionable, independent of the result of the histological examination. The condence score became 2 for the extreme condence ratings of 1 and 5, denitely not carious and denitely carious. The intermediate
Table 1 Number of lesions and percentages of approximal caries surfaces (n 223) of extracted premolars and molars, with histology as gold standard Number Caries Caries Caries Caries Sound in outer half in inner half in outer half in inner half surfaces of of of of enamel enamel dentine dentine 54 27 38 18 86 Percentage 24% 12% 17% 8% 39%

condence ratings of 2 and 4, probably not carious and probably carious, were converted into a condence score of 1. Then, a sign was added to scores 1 and 2 when a carious lesion was present, as observed histologically, while a 2 sign was added in the histological absence of a lesion. Thus, a penalty was given for a wrong rating, which was larger when the observer was more condent about his assessment; similarly, a reward was given for correct observations, which was larger when the observer was more sure of his assessment. The signed observer error was calculated by subtracting the histologically true lesion depth (0 4) from the radiographically observed lesion depth (0 4). Positive error values were thus obtained when an observer overestimated lesion depth, and negative values were obtained when lesion depth was underestimated. Value 0 was obtained when radiographic and histological depth estimates were in agreement. The signed observer error does not always show how large the average error of an observer is, because the positive values of overestimations and the negative values of underestimations can annul each other to a larger or smaller extent. Therefore, the absolute observer error was calculated as the absolute values of the signed observer error. The absolute observer error indicates how big the error made by the observer is, irrespective of overestimation or underestimation. Signed and absolute observer errors together provide a good insight into observer performance. In an analysis of variance (MANOVA), observer condence, and signed and absolute error were entered as dependent variables. Lesion depth (0 4) was treated as a between-subject factor. Image method (Digora-enhanced vs Oslo-enhanced) and observer identity (1 to 7) were entered as within-subject factors. Results The most evident difference between the methods as elucidated by the ROC analysis was the highly signicant smaller interobserver and intraobserver variance with the Oslo-enhanced method compared with the Digoraenhanced method for all but one observer (Table 2). The methods were not different regarding Az values as tested with paired t-tests, and there was no correlation between Az values across methods. From the analysis of variance, supplementary results were obtained. Image method and lesion main effects, as well as the lesion by method interaction effects, were multivariately signicant (P , 0.001) in favour of the Oslo-enhanced method. On a univariate level, the method main effect was not signicant for the absolute observer error (P 0.330). All other univariate (main and interaction) effects were signicant (P , 0.001). To illustrate the image method effect, the mean values for observer condence scores, and signed and absolute errors associated with the two methods are presented in Table 3. On average, the Oslo-enhanced method gave rise to a signicantly greater observer condence and

Dentomaxillofacial Radiology

Digital caries diagnosis A Mystad et al

393

Table 2 Diagnostic accuracy of approximal caries expressed as Az values and standard deviation (SD) in receiver operating characteristic (ROC) analyses for seven observers reading 223 surfaces in Digora storage phosphor images, either individually enhanced with Digora software 2.0 (Digora-enhanced) or automatically enhanced with the caries-specic Oslo enhancement procedure (Oslo-enhanced). Tests of differences between the two methods Digora-enhanced Observer Az 1 2 3 4 5 6 7 0.8797 0.9195 0.7930 0.7793 0.8215 0.8608 0.7964 SD (n 223) 0.0430 0.0491 0.0491 0.0323 0.0473 0.0270 0.1107 (n 7) 0.0522 Az 0.8502 0.8191 0.8515 0.8272 0.8336 0.8508 0.8562 0.8412 SD (n 223) 0.0302 0.0352 0.0261 0.0278 0.0273 0.0264 0.0504 (n 7) 0.0144 F 2.02732 1.94571 3.53901 1.34994 3.00191 1.04597 4.82430 13.1406 t 0.25 r 2 0.298 P (two-sided) , 0.0001 , 0.0001 , 0.0001 0.0258 , 0.0001 0.7380 (NS) , 0.0001 0.0064 0.81 (NS) 0.52 (NS) Oslo-enhanced

Mean 0.8357 Paired t-test Correlation between Az Digora and Az Oslo NS, not signicant

Table 3 Estimated averages for observer condence, and signed and absolute observer error, with standard deviation (SD), of approximal caries diagnosis for seven observers reading 223 surfaces in Digora storage phosphor images. (For Digora-enhanced and Oslo-enhanced, see text for explanation of variables) Image method Condence
a

Mean 0.616 0.980 2 0.677 2 0.535 0.783 0.759

SD 0.059 0.073 0.046 0.050 0.039 0.040

Signed error

Absolute error
a

Digora-enhanced Oslo-enhanced Digora-enhanced Oslo-enhanced Digora-enhanced Oslo-enhanced

Signicantly different (P , 0.001) Figure 2 Observer condence by lesion depth for the two methods: storage phosphor images individually enhanced with the Digora software 2.0 (Digora-enhanced) and automatically enhanced with the cariesspecic Oslo enhancement procedure (Oslo-enhanced)

a signicantly smaller amount of underestimation than the Digora-enhanced method. The difference in absolute observer error, however, was not large enough to become signicant. Figures 2 4 show the effects responsible for the differences between the Digora-enhanced and the Osloenhanced method. Figure 2 shows that for both methods the observers reported with some condence the absence of a lesion, but with Digora-enhanced images the condence of the observers was greater than with the Oslo-enhanced images. In contrast, the condence with regard to outer enamel lesions resulted in more wrong decisions in the Digora-enhanced images than in the Oslo-enhanced images. The mean observer condence value for outer enamel lesions was not below 0 in the Oslo-enhanced images, indicating that, on average, the observers questioned whether a lesion was present or not. In the Digora-enhanced images, however, the observers were condent to a certain degree that a lesion was not present when in fact there was a true outer enamel lesion. Figure 3 shows that both methods underestimated lesion size, expressed as signed observer error, which was to a lesser extent in the Oslo-enhanced images than in the Digora-enhanced images, except for lesions in the inner dentine. Whereas the results shown in Figures 2 and 3 were signicant, the absolute observer error as shown in Figure 4 was not signicant between the methods.

Figure 3 Signed observer error by lesion depth for the two methods: storage phosphor images individually enhanced with the Digora software 2.0 (Digora-enhanced) and automatically enhanced with the cariesspecic Oslo enhancement procedure (Oslo-enhanced)

All interaction effects in which the observer is involved were multivariately signicant. On a univariate level, the lesion by method by observer interaction effect was the only one not signicant (P 0.201).
Dentomaxillofacial Radiology

394

Digital caries diagnosis A Mystad et al

Discussion This is the third study in which Digora storage phosphor images, pre-enhanced with the caries-specic Oslo enhancement procedure on a different platform, have been analysed with regard to accuracy for approximal caries diagnosis, validated against histology. In both previous studies, the pre-enhanced storage phosphor images were compared with conventional lms.2,3 In the study by Mystad et al,2 the diagnostic accuracy was found to be signicantly superior with the digital images with regard to both enamel and dentine, evaluated separately. This was supported in the more recent study by Svans et al.3 When pre-enhanced images and lms in that study were analysed with regard to diagnostic accuracy of caries lesions in outer and inner enamel, and in outer and inner dentine, separately, only the accuracy of outer enamel evaluation was signicantly different. Additionally, a signicantly lower interobserver variability was found with the pre-enhanced images. In the present study in which pre-enhanced images were compared with storage phosphor images individually enhanced by the observers using Digora software, the diagnostic accuracy was also analysed with ROC statistics. However, the differences between the averaged Az values obtained with pre-enhanced and individually enhanced storage phosphor images were not signicant. What the ROC analysis did show was the signicantly smaller observer variability with the pre-enhanced images, substantiating the study by Svans et al.3 The value of obtaining better agreement between observers with regard to caries diagnosis has been emphasized by others in a recent review.6 It has been claimed that if digital diagnostic systems are to be replaced by a conventional system, there should not only be better diagnostic accuracies but also oberserver variabilities should be at least comparable.7 Digital subtraction radiography was introduced early as a method for approximal caries detection.8 However, in a recent subtraction study the interobserver variability did

Figure 4 Absolute observer error by lesion depth for the two methods: storage phosphor images individually enhanced with the Digora software 2.0 (Digora-enhanced) and automatically enhanced with the cariesspecic Oslo enhancement procedure (Oslo-enhanced)

The way the two methods worked for different observers is presented in Figure 5. The Oslo-enhanced and Digoraenhanced methods are presented separately with regard to observer condence by lesion depth. With the Digoraenhanced images, not one observer managed, on average, to assess outer enamel lesions correctly. With the Osloenhanced images some observers (Observers 3, 4 and 5) were able to spot the outer enamel lesions correctly more often than they did with the Digora-enhanced images. The other observers were convinced that there was nothing to be seen in the outer enamel, although the strength of their conviction was less than with the Digora-enhanced images. The Oslo-enhanced images tended to make sound surfaces less obvious to Observers 3, 4 and 5, whereas the other observers were equally condent in their decisions with both methods. The curves in Figure 5 show that observers with the lowest condence values when deciding that outer enamel lesions were present had the highest values when deciding that carious lesions were absent.

Figure 5 Observer condence by lesion depth by seven observers for the two methods: (a) storage phosphor images individually enhanced with the Digora software 2.0; and (b) storage phosphor images automatically enhanced with the caries-specic Oslo enhancement procedure
Dentomaxillofacial Radiology

Digital caries diagnosis A Mystad et al

395

not improve compared with lm evaluation,9 in contrast to our nding with the pre-enhanced images. The ROC technique used in both the two previous studies and in the present study produces general statistics on each observer, and may well fail to detect the individual contribution of a specic factor involved in the study.10,11 Therefore, an analysis of variance was also used, enabling assessment of different parameters independently (between- and within-subject factors). Some differences between the methods were found with this analysis. When viewing pre-enhanced images, the observers underestimated caries lesion depth to a lesser extent than when viewing individually enhanced images, except in inner dentine. However, with both methods it was difcult for the observers to correctly separate approximal surfaces with carious lesions in the outer enamel from those that were sound. Some observers were able to assess outer enamel lesions correctly with the pre-enhanced images, but expressed less condence regarding their assessment of sound surfaces. The opposite was found with the individually enhanced images. One explanation might be that when the observers were asked to enhance the original Digora images ad libitum it is possible that some observers might have focused on the detection of sound surfaces more than on the detection of enamel lesions. From a clinical point of view it is important to improve the detection of approximal lesions in general as well as to detect sound surfaces correctly, particularly in patients with a low caries prevalence and a slow caries progression. Thus, further efforts are needed in the development of methods that might be able to correctly discriminate between sound surfaces and surfaces with caries lesions. In the present study we did not register the manipulation facilities used by the observers, nor did we record the time they spent on individual image enhancement, but certainly all the observers spent some time doing their manipulations ad libitum. Shrout et al12 suggested that digital enhancement was benecial, but selection of the proper procedure was time consuming since the clinicians are required to browse through potential enhancement techniques subjectively.12 Many authors have focused on the advantages of real-time imaging and time reduction with digital imaging,13 16 but very few have studied time use. In a recent questionnaire study among general dental practitioners in Norway, 79% stated that they saved time, which was estimated as 36 min per day for storage phosphor users and 25 min per day for charge-coupled device (CCD) sensor users.17 The average time spent on gamma/contrast and brightness manipulations for general dental practitioners
References
1. Wenzel A. Digital radiography and caries diagnosis. Dentomaxillofac Radiol 1998; 27: 3 11. ndahl H-G. 2. Mystad A, Svans DB, Risnes S, Larheim TA, Gro Detection of approximal caries with a storage phosphor system. A comparison of enhanced digital images with dental X-ray lm. Dentomaxillofac Radiol 1996; 25: 202 206.

has been reported to be approximately 24 s per image.18 We are convinced from our study that dentists with less experience in digital radiology than our observers would benet when evaluating pre-enhanced images. This concerns both the speed of the diagnostic procedure and the observer agreement. Our impression is that many dentists do not consider that they have time enough to manipulate each image before the diagnostic asssessment, and we doubt that those without digital experience would have obtained a similar diagnostic accuracy as our observers did. In a study on digitized radiographs, a signicant decrease in diagnostic accuracy of incipient approximal caries was observed when participating observers manipulated brightness/contrast and image size of optimally adjusted images.19 Since we have previously shown that image magnication may inuence diagnosis of approximal caries,20,21 the observers were not allowed in the present study to change the size of either the preenhanced images or the individually enhanced images. On the other hand, we do not know whether our pre-enhanced images could have been improved even further if the observers were allowed to manipulate the already optimized images. It might then be a possibility that the observers degraded these images. This will be the subject for a future study. It also remains to be studied whether our caries-specic enhancement procedure may yield similar results with CCD-based systems. In conclusion, the average accuracy of approximal caries detection obtained with Digora storage phosphor images either pre-enhanced on a different platform with the caries-specic Oslo enhancement procedure or individually enhanced with the Digora software was similar, but an improved interobserver and intraobserver variability was obtained with the Oslo-enhanced method. Since image manipulation is not performed with the Oslo-enhanced method, an increased speed of the diagnostic procedure combined with an improved observer variability would probably be even more pronounced for the general dental practitioner.

Acknowledgments The authors wish to acknowledge the contribution to this study by Steinar Risnes, DDS, PhD, Department of Oral Biology, Faculty of Dentistry, University of Oslo, Norway, who conducted the histological evaluation; and Eva Borg, DDS, PhD, Department of teborg University, Sweden, Oral and Maxillofacial Radiology, Go and GCH Sanderink, DDS, PhD, Department of Oral Radiology, Academic Centre for Dentistry, Amsterdam, The Netherlands, who participated as observers.

3. Svans DB, Mystad A, Larheim TA. Approximal caries depth assessment with storage phosphor versus lm radiography: evaluation of the caries-specic Oslo enhancement procedure. Caries Res 2000; 34: 448 453. 4. Risnes S. A rotating specimen holder for hard tissue sectioning. Stain Technol 1981; 56: 265 266.

Dentomaxillofacial Radiology

396

Digital caries diagnosis A Mystad et al

5. Swets JA, Pickett RM. Evaluation of diagnostic systems: methods from signal detection theory. New York, NY: Academic Press, 1982. 6. Wenzel A. Digital imaging for dental caries. Dent Clin North Am 2000; 44: 319 338. 7. Verdonshot EH, Kuijpers JMC, Polder BJ, De Leng-Worm MH, Bronkhorst EM. Effects of digital grey-scale modication on the diagnosis of small approximal carious lesions. J Dent 1992; 20: 44 49. ndahl H-G, Gro ndahl K, Okano T, Webber RL. Statistical contrast 8. Gro enhancement of subtraction images for radiographic caries diagnosis. Oral Surg Oral Med Oral Pathol 1982; 53: 219 223. 9. Wenzel A, Antonisen PN, Juul MB. Reproducibility in the assessment of caries lesion behaviour: a comparison between conventional lm and subtraction radiography. Caries Res 2000; 34: 214 218. 10. Metz CE. Basic principles of ROC analysis. Semin Nucl Med 1978; 8: 283 298. 11. Wilson PWF, DAgostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation 1998; 97: 1837 1847. 12. Shrout MK, Russel CM, Potter BJ, Powell J, Hildebolt CF. Digital enhancement of radiographs: can it improve caries diagnosis? J Am Dent Assoc 1996; 127: 469 473. 13. Versteeg CH, Sanderink GCH, van der Stelt PF. Efcacy of digital intra-oral radiography in clinical dentistry. J Dent 1997; 25: 215 224.

14. Lavelle CL. The role of direct intraoral sensors in the provision of endodontic services. Endod Dent Traumatol 1999; 15: 1 5. 15. Berkhout WER, Sanderink GCH, van der Stelt PF. A comparison of digital and lm radiography in Dutch dental practices assessed by questionnaire. Dentomaxillofac Radiol 2002; 31: 93 99. 16. Parks ED. Digital radiography: an overview. J Contemp Dent Pract 2002; 3: 1 13. 17. Wenzel A, Mystad A. Experience of Norwegian general dental practitioners with solid state and storage phosphor detectors. Dentomaxillofac Radiol 2001; 30: 203 208. ndahl H-G. Observers use of image 18. Gotfredsen E, Wenzel A, Gro enhancement in assessing caries in radiographs taken by four intraoral-digital systems. Dentomaxillofac Radiol 1996; 25: 34 38. 19. Ohki M, Okano T, Nakamura T. Factors determining the diagnostic accuracy of digitized conventional intraoral radiographs. Dentomaxillofac Radiol 1994; 23: 77 82. ndahl H-G. Effect of image 20. Mystad A, Svans DB, Larheim TA, Gro magnication of digitized bitewing radiographs on approximal caries detection: an in vitro study. Dentomaxillofac Radiol 1995; 24: 255 259. 21. Svans DB, Mystad A, Risnes S, Larheim TA. Intraoral storage phosphor radiography for approximal caries detection and effect of image magnication: comparison with conventional radiography. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1996; 82: 94 100.

Dentomaxillofacial Radiology