Dentomaxillofacial Radiology (2008) 37, 458463 2008 The British Institute of Radiology http://dmfr.birjournals.

org

RESEARCH

A comparison of the diagnostic accuracy of bitewing, periapical, unfiltered and filtered digital panoramic images for approximal caries detection in posterior teeth
ZZ Akarslan*,1, M Akdevelioglu1, K Gungor1 and H Erten2
1 Department of Oral Diagnosis and Radiology, Gazi University School of Dentistry, Turkey; 2Department of Operative Dentistry and Endodontics, Gazi University School of Dentistry, Turkey

Objectives: The aim of this study was to investigate the diagnostic accuracy of conventional bitewing, periapical radiographs, unfiltered and filtered digital panoramic images for the assessment of approximal carious lesions in posterior teeth. Methods: 20 patients who required bitewing, periapical and panoramic radiographs participated in the study. The digital panoramic images were treated with the sharpen, smooth and emboss filters. Three experienced observers evaluated the bitewing and periapical radiographs, and the unfiltered and filtered digital panoramic images for the presence or absence of approximal carious lesions according to a 5-point confidence scale. Intra- and interobserver reliability was analysed with the Cronbach alpha test and diagnostic accuracy was evaluated using receiver operating characteristic (ROC) analysis. Results: According to the results, the intra- and interobserver agreement levels were found to be almost perfect. The highest diagnostic accuracy was found for the bitewing, followed by the periapical and panoramic images. Compared with the unfiltered panoramic images, the filtered images had an almost equal or slightly higher diagnostic accuracy. The largest increase in diagnostic accuracy was observed for those images filtered with the emboss filter. Conclusions: The accuracy of unfiltered and filtered digital panoramic images is lower than conventional bitewing and periapical radiographs. The filtration of the digital panoramic images, especially the emboss filter, may have some value for the detection of approximal carious lesions. Dentomaxillofacial Radiology (2008) 37, 458463. doi: 10.1259/dmfr/84698143 Keywords: approximal caries, bitewing, periapical, digital panoramic

Introduction Bitewing, periapical and panoramic radiographic imaging techniques are routinely used in dentistry. Although the bitewing and periapical radiographs have been shown to be superior to conventional panoramic images due to the projection geometry and image resolution, it has been proposed that the image resolution of panoramic radiography has increased with technological improvements and become comparable with intraoral radiography for the diagnosis of dental caries.1,2
*Correspondence to: Dr Zuhre Zafersoy Akarslan, Gazi Universitesi Dis Hekimligi Fakultesi Oral Diagnoz ve Radyoloji BD, 8 Cadde, 84 Sok, Emek, Ankara, Turkiye; E-mail: dtzuhre@yahoo.com Received 31 August 2007; revised 11 December 2007; accepted 11 December 2007

Panoramic images could be obtained with film-based conventional techniques and direct digital imaging systems. Compared with conventional film-based techniques, direct digital panoramic imaging systems have some advantages, such as image enhancement, which may enhance their diagnostic value or facilitate the diagnostic interpretation.36 Controversial results have been reported for image enhancement facilities for caries detection in intraoral digital radiography. Some of the authors have reported that contrast enhancement and filtering increase the diagnostic accuracy for the detection of caries,7,8 while some have reported the contrary.9,10 Studies have evaluated the efficacy of bitewing, periapical and panoramic images in the diagnosis of

Approximal caries detection and radiography ZZ Akarslan et al

459

approximal carious lesions,1,11 but to our knowledge a comparison between conventional intraoral radiographic techniques, unfiltered and filtered digital panoramic images has not been assessed to date. As radiographic techniques are widely used for approximal caries diagnosis on posterior teeth and there is a rapid development and increase in adopting digital panoramic systems, we aimed to evaluate the efficacy of digital images and the effect of different filters on caries detection in posterior teeth, and compare them with conventional film-based bitewing and periapical radiographic techniques.

were taken, imported from the clipboard to MS paint (MicrosoftH Paint v5.1; Microsoft Corporation, Redmond, WA) and saved in files corresponding to the patients bitewing and periapical radiographs. Observation sessions The bitewing and periapical radiographs were masked and evaluated by each observer separately in the same room, on a view box under subdued lighting, in different viewing sessions. Approximately 2 weeks separated the viewing sessions of both radiographs. The digital panoramic images were displayed on a 17 inch Super VGA monitor with a screen resolution of 10246768 pixels. The contrast and brightness of the images were set to 100 and 0, respectively. The computer was an Intel PentiumH having 256 MB of RAM. The operating system of the computer was Windows XP (Microsoft, v2002) and the digital imaging software used for the Orthoralix DDE images was VixWin Pro (Gendex Dental Systems, v1.5). All unfiltered and filtered digital panoramic images were evaluated under subdued lighting conditions and the viewing distance was kept at approximately 70 cm. All of the conventional radiographs, unfiltered and panoramic images were evaluated by three experienced radiologists. All observers were instructed on the definition of the rating scale before the examination sessions. The observers examined the approximal surfaces of the teeth according to a 5-point confidence scale as follows: 1 5 caries definitely absent; 2 5 caries probably absent; 3 5 equal chance of caries being present or absent; 4 5 caries probably present; 5 5 caries definitely present. This made it possible to perform receiver operating characteristics (ROC) analysis. A carious lesion was defined as any amount of decalcification present on the approximal surface of a tooth.12 The observers were blinded to each other during evaluation sessions of the bitewing, periapical and panoramic images. The observers did not know which bitewing, periapical, unfiltered digital and filtered images belonged to which patient. All images were evaluated randomly. Approximately 2 weeks elapsed between the examination sessions of the bitewing, periapical, unfiltered and filtered panoramic images. Intraobserver reliability was assessed 1 month later with the examination of all radiographic images taken from ten of the patients. 2 months later, after all of the evaluation sessions were completed, the true disease status of the teeth was determined with the consensus of the three observers simultaneous interpretation of all evaluated bitewing, periapical, unfiltered and filtered panoramic digital images. Statistical analysis Intra- and interobserver agreement levels were analysed with the calculation of Cronbachs alpha values. The diagnostic accuracy of each imaging modality was
Dentomaxillofacial Radiology

Materials and methods Image acquisition 20 patients who had no missing posterior or canine teeth in all quadrants (except third molars), requiring posterior bitewing, periapical and panoramic radiographic investigation were included in this study. The bitewing and periapical radiographs were taken with double-packed Ektaspeed Plus films on a radiograph machine (Trophy CCX, Vincennes, France) operating at 70 kVp and 8 mA, having 2.5 eq aluminium filtration and a 0.860.8 mm focal spot, according to the manufacturers exposure recommendations, with the bisecting angle technique to obtain duplicate radiographs. The radiographs were processed in an automatic roller transport processor machine (Velopex Extra-X Medivance Instruments Ltd, London, UK) with fresh chemicals. The digital panoramic images were taken with the Orthoralix 9200 DDE machine (Gendex Dental Systems, Milan, Italy). The digital panoramic images were taken at 7074 kVp, 410 mA and 12 s according to patient size in order to maintain consistent radiographic density from the unit. The digital images were taken at 16-bit greyscale levels and saved as TIFF files. One set of the bitewing and periapical radiographs was placed in the patients dental charts and the second set was used in the study. The posterior bitewing, periapical radiographs and digital panoramic images were taken by one trained investigator to provide consistency in the radiographic techniques. The digital panoramic images consisted of one unfiltered and three filtered images with the sharpen, smooth and emboss filters. A total of four kinds of digital panoramic images were evaluated. The filters are present as both pull-down menus and as icons localized above the images. The use of the filter icons is easy as they can be applied to the images using the mouse. The sharpen filter is a high pass filter which enhances the edges, with an increased grainier appearance, and the smooth filter is a low pass spatial filter providing smooth blurring. The emboss filter is a filter providing a simulated three-dimensional (3D) image. After the application of each of the filters to the unfiltered image, screenshots using the Windows screen print function

460

Approximal caries detection and radiography ZZ Akarslan et al

Figure 1 All images belong to the same site in the same patient. Arrows show the approximal caries. (a) Bitewing radiograph used in the study. (b) Periapical radiograph used in the study. (c) Cropped unfiltered panoramic image. (d) Cropped panoramic image, filtered with the sharpen filter. (e) Cropped panoramic image, filtered with the smooth filter. (f) Cropped panoramic image, filtered with the emboss filter. The bitewing and periapical radiographs were obtained by scanning conventional radiographs, and the panoramic images were captured from the computer screen using a digital camera

analysed with the ROC analyses (a test in the diagnostic efficacy of an imaging method is compared with chance) and asymptotic 95% confidence intervals. The mean of the area under ROC curves (AUC) was calculated for each observer for each imaging modality, using SPSS 10.0 statistical program.

Results Out of 640 surfaces, 160 were excluded as they had a restoration; therefore, 480 surfaces were used in the present study. A total of 282 sound surfaces and 198 carious surfaces were detected according to the true disease status, determined by the consensus of the three observers. The depth of the lesions localized on the enamel, on the outer half of the dentin and the inner half of the dentin were determined as 60 (30.2%), 113 (56.8%) and 26 (13.1%), respectively. The frequencies and percentages of the sound and carious surfaces are shown in Table 1. Cronbachs alpha values for intraobserver reliability were calculated as 0.918, 0.929 and 0.957 for the first, second and third observers, respectively. The Cronbachs alpha value for interobserver agreement was calculated as 0.940, indicating an almost perfect agreement level. All of the approximal surfaces of the maxillary, mandibular molar and premolar teeth which had no
Dentomaxillofacial Radiology

restoration could be assessed with both bitewing and periapical radiographs. Some of the surfaces could not be assessed with the normal and filtered panoramic radiographs. A high number of the surfaces of the maxillary premolar teeth (n 5 72) could not be evaluated with the panoramic radiographs due to superimposition of the approximal surfaces. Most of the approximal surfaces of the premolar teeth which could be evaluated with the panoramic images were rated to have carious lesions. Following this dental region, 32 surfaces of the mandibular premolars, 6 surfaces of the maxillary molars and 4 surfaces of the mandibular molars could not be assessed. Details are shown in Table 2. The highest mean AUCs were calculated for the bitewing, followed by periapical radiographs for all dental regions. The AUCs calculated for the normal and filtered panoramic radiographs were lower than the bitewing and periapical radiographs. The dental regions affected the diagnostic accuracy of all imaging modalities. According to the mean of the

Table 1

Numbers and percentages of sound and carious surfaces Sound n (%) Carious n (%) 44 51 52 51 198 (36.7) (42.5) (43.3) (42.5) (41.3) 76 69 68 69 282 (63.3) (57.5) (56.7) (57.5) (58.8)

Maxillary premolar Maxillary molar Mandibular premolar Mandibular molar Total

Approximal caries detection and radiography ZZ Akarslan et al

461

Table 2 Number of proximal surfaces that could and could not be examined with each imaging modality in all dental regions Maxillary premolar Imaging modality Bitewing Periapical Panoramic E 120 120 48 NE 0 0 72 Maxillary molar E 120 120 114 NE 0 0 6 Mandibular premolar E 120 120 88 NE 0 0 32 Mandibular molar E 120 120 116 NE 0 0 4

E, number of surfaces that could be evaluated; NE, number of surfaces that could not be evaluated

AUCs, the highest diagnostic accuracy was calculated for the mandibular molar teeth with the bitewing, periapical, unfiltered and filtered digital panoramic images. The lowest diagnostic accuracy was calculated for the maxillary molar region with bitewing, periapical and panoramic images treated with the sharpen filter, and for the mandibular premolar region with the unfiltered panoramic and panoramic image treated with the smooth and sharpen filter. According to the mean AUCs, the application of the sharpen and smooth filters slightly changed the diagnostic accuracy of the unfiltered panoramic images. On the other hand, the emboss filter led to a high increase in the diagnostic accuracy in the maxillary molar region. Details are shown in Table 3. The highest AUCs were calculated for the bitewing, followed by periapical radiographs for all observers. The AUCs decreased for the unfiltered panoramic radiograph for all observers compared with bitewing and periapical radiographs. Other than the second observer, the diagnostic accuracy of the unfiltered and filtered panoramic images was statistically significant from chance (P , 0.05), indicating that they were all suitable imaging modalities for the detection of approximal carious lesions. For the second observer, the unfiltered panoramic images diagnostic accuracy decreased in the mandibular premolar region and this was not statistically significant from chance (AUC 0.585; P . 0.05), indicating that it had no value for approximal caries diagnosis in this region. With the use of the sharpen and emboss filters, the AUCs increased significantly (AUC 0.647; AUC 0.628; P , 0.05) and

with the use of the smooth filter the AUCs significantly decreased compared with the bitewing, periapical, panoramic image treated with the sharpen and emboss filters for this observer (AUC 0.605; P . 0.05). With the application of the emboss filter to the unfiltered panoramic images, an increase in the AUCs was observed for most of the regions for all observers.

Discussion According to the mean AUCs, the highest diagnostic accuracy was calculated for the bitewing, followed by periapical radiographs for all dental regions. Compared with the bitewing and periapical radiographs, the unfiltered and filtered digital panoramic images had a lower accuracy in all dental regions. It is reported that conventional film-based bitewing radiography is superior to conventional film-based panoramic radiography for the detection of approximal carious lesions.1,13 Our results show consistency with these results. Compared with the unfiltered digital panoramic images, the diagnostic accuracy of the filtered images for approximal carious lesion detection was almost equal to or only slightly increased in some cases. According to the mean AUCs, the highest increase was observed for the emboss filter. The appearance of the images treated with the sharpen and smooth filters are similar to each other and the unfiltered image, but the appearance of the image treated with the emboss filter is different from the unfiltered image. We think that the

Table 3 The mean, minimum and maximum (range) of the area under ROC curves values and the upper and lower borders of 95% confidence interval (CI)for each imaging modality Region Maxillary premolar Maxillary molar Mandibular premolar Mandibular molar Mean Range CI Mean Range CI Mean Range CI Mean Range CI Bitewing 0.979 0.9660.989 0.9281.013 0.923 0.9010.946 0.8380.993 0.928 0.8890.949 0.8330.990 0.986 0.9830.991 0.9681.002 Periapical 0.956 0.9290.969 0.8751.005 0.905 0.8850.929 0.8220.979 0.907 0.8580.943 0.8710.992 0.970 0.9640.976 0.9361.003 Panoramic 0.738 0.7020.787 0.7040.998 0.693 0.6080.843 0.4930.922 0.682 0.5850.760* 0.4660.861 0.838 0.7970.914 0.6990.982 Panoramic sharpen 0.746 0.7230.785 0.7761.006 0.698 0.6200.839 0.5060.919 0.708 0.6470.760 0.5330.861 0.853 0.7970.924 0.6990.990 Panoramic smooth Panoramic emboss 0.736 0.7260.745 0.6250.898 0.717 0.6460.815 0.5330.900 0.678 0.6050.761* 0.4870.862 0.845 0.7980.911 0.6990.981 0.776 0.7510.795 0.7150.896 0.828 0.7900.886 0.7190.954 0.693 0.6280.771 0.5110.870 0.859 0.8250.912 0.7360.982

*P . 0.05, indicating not significant from chance

Dentomaxillofacial Radiology

462

Approximal caries detection and radiography ZZ Akarslan et al

higher increase in accuracy is due to the different appearance of the image, leading to a better visualization. We could not compare our results directly as we could not find a similar study made with digital panoramic imaging systems, but studies have shown that digital contrast enhancement and filtering may increase diagnostic accuracy for the detection of carious lesions in digital intraoral radiography.7,8,14 In the present study, the observers were not permitted to change the contrast of the images as we wanted to standardize the experiment conditions, letting only the assessment of the effect of the filters. In further studies, the effect of contrast enhancement and filters could be evaluated. There is no standardized terminology in digital software programs. This makes it very difficult to compare the efficacy of different kinds of filters used in different studies. In the future, standardized terminology of image processing should be provided in dentistry to allow a more accurate comparison between the results of different studies.15 The diagnostic accuracy of digital panoramic images differed according to the examined dental region. The accuracy for molar teeth was higher from premolar teeth. Akkaya et al1 reported that the diagnostic accuracy of conventional panoramic images for the molar region in approximal carious lesion detection was higher than that for premolar teeth. Our results are similar to that study. In rotational panoramic radiography it is generally accepted that marked overlapping of the tooth crowns occurs frequently, resulting in difficulties in the assessment of approximal carious lesions.16 This situation is particularly prominent in the premolar region. In our study, nearly twothirds of the examined maxillary premolar teeth could not be assessed due to overlapping of the approximal surfaces from the digital panoramic radiograph; therefore, digital panoramic radiography is not a suitable tool for the examination of the approximal surfaces of premolar teeth. In most of the cases, as the depth of the lesions increased, they could be detected in the digital panoramic images even when overlapping was present. The true disease status should be stated according to histological examination in studies in which the

comparison of different methods for caries detection is assessed. However, in some clinical studies, the style of the clinical setting does not make it possible for histological evaluation.17 Although assessment of the radiographs without clinical examination is not an appropriate procedure, the consensus diagnosis obtained from experienced observers is used for the comparison of different radiographic methods.1,18 We therefore used the consensus of the three experienced observers for the determination of the actual disease status of the examined teeth. In such a study design, it was not possible to perform histological examination of the teeth, therefore the AUCs calculated for the bitewing and periapical radiographs were over 0.90 and greatly inflated due to the gold standard as they provided most of the evidence of the disease compared with the panoramic images. The depth of the lesions is reported to have an impact on the radiographic detectability, so lesions localized on the inner half of dentin are detected with radiographs more often than those localized on the enamel. The accuracy of the detection of small approximal lesions has been reported to be almost equal to chance.19 In the present study, the majority of the lesions were localized on the inner half of dentin. ROC analysis provides the most meaningful approach to compare the diagnostic performance of two or more different radiographic imaging modalities because it distinguishes between the inherent capacities of the observers to under- and over-read when interpreting imaging and is used in many studies.1,9 The analysis is made by comparing significant differences between the areas under the ROC curves that represent the competing modalities.20 For these reasons, we used the ROC analysis for the comparison of the methods. In conclusion, the diagnostic accuracy of the digital panoramic images was lower than conventional film based bitewing and periapical radiographs, but it did have a value in the detection of posterior approximal carious lesions, especially for the mandibular molar teeth. The filtration of the digital panoramic images with the emboss filter has some value for the detection of approximal carious lesions in posterior teeth.

References
1. Akkaya N, Kansu O, Kansu H, Cagrankaya LB, Arslan U. Comparing the accuracy of panoramic and intraoral radiography in the diagnosis of proximal caries. Dentomaxillofac Radiol 2006; 35: 170174. 2. Lundeen RC, McDavid WD, Branwel GM. Proximal surface caries detection with direct exposure and rare earth screen/ film imaging. Oral Surg Oral Med Oral Pathol 1988; 66: 734 745. 3. Shrout MK, Russell CM, Potter BJ, Powell J, Hildebolt CF. Digital enhancement of radiographs: can it improve diagnosis? J Am Dent Assoc 1996; 127: 469473. 4. Analoui M. Radiographic image enhancement. Part I: spatial domain techniques. Dentomaxillofac Radiol 2001; 30: 19.
Dentomaxillofacial Radiology

5. Analoui M. Radiographic image enhancement. Part II: transform domain techniques. Dentomaxillofac Radiol 2001; 30: 6577. 6. Mol A. Image processing tools for dental applications. Dentomaxillofac Radiol 2000; 44: 299318. 7. Wenzel A. Digital radiography and caries diagnosis. Dentomaxillofac Radiol 1998; 27: 311. 8. Moystad A, Svanaes DB, Risnes S, Larheim TA, Grondahl HG. Detection of approximal caries with a storage phosphor system. A comparison of enhanced digital images with dental X-ray film. Dentomaxillofac Radiol 1996; 25: 202206. 9. Tyndall DA, Ludlow JB, Platin E, Nair M. A comparison of Kodak Ektaspeed Plus film and the Siemens Sidexis digital imaging system for caries detection using receiver operating

Approximal caries detection and radiography ZZ Akarslan et al

463

10.

11.

12.

13.

14.

characteristics analysis. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1998; 85: 113118. Eickholz P, Kolb I, Lenhard M, Hassfeld S, Staehle H. Digital radiography of interproximal caries: effect of different filters. Caries Res 1999; 33: 234241. Flint DJ, Paunovich E, Moore WS, Wofford DT, Hermesch CB. A diagnostic comparison of panoramic and intraoral radiographs. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1998; 85: 731735. Galal A, Manson-Hing L, Jamison H. A comparison of combinations of clinical and radiographic examinations in evaluation of a dental clinic population. Oral Surg Oral Med Oral Pathol 1985; 60: 553561. Douglass CW, Valachovic R, Wijesinha AW, Chauncey HH, Kapur KK, McNeil BJ. Clinical efficacy of dental radiography in the detection of dental caries and periodontal disease. Oral Surg Oral Med Oral Pathol 1986; 62: 330339. Wenzel A. Computer-aided image manipulation of intraoral radiographs to enhance diagnosis in dental practice: a review. Int Dent J 1993; 43: 99108.

15. Lehmann TM, Troeltsch E, Spitzer K. Image processing and enhancement provided by commercial dental software programs. Dentomaxillofac Radiol 2002; 31: 264272. 16. Scarfe WC, Nummikoski P, McDavid WD, Welander U, Tronje G. Radiographic interproximal angulations: Implications for rotational panoramic radiography. Oral Surg Oral Med Oral Pathol 1993; 76: 664672. 17. Halse A, Molven O. A strategy for the diagnosis of periapical pathosis. J Endod 1986; 12: 534538. 18. Valachovic RW, Douglass CW, Reiskin AB, Chauncey HH, McNeil BJ. The use of panoramic radiography in the evaluation of asymptomatic adult patients. Oral Surg Oral Med Oral Pathol 1986; 61: 289296. 19. Hintze H, Wenzel A, Jones C. In vitro comparison of D and E speed film radiography, RVG and Visualix digital radiography for the detection of enamel appoximal and dentinal occlusal caries lesions. Caries Res 1994; 28: 363367. 20. Kantor ML, Zeichner SJ, Valachovic RW, Reiskin AB. Efficacy of dental radiographic practices: options for image receptors, examination selection and patient selection. J Am Dent Assoc 1989; 119: 259268.

Dentomaxillofacial Radiology