Effect of various digital processing algorithms on the measurement accuracy of endodontic file length

Transcription

1 Effect of various digital processing algorithms on the measurement accuracy of endodontic file length Betül İlhan Kal, DDS, a B. Güniz Baksı, PhD, b Nesrin Dündar, PhD, c and Bilge Hakan Şen, PhD, d Izmir, Turkey EGE UNIVERSITY Objective. The aim of this study was to compare the accuracy of endodontic file lengths after application of various image enhancement modalities. Study Design. Endodontic files of three different ISO sizes were inserted in 20 single-rooted extracted permanent mandibular premolar teeth and standardized images were obtained. Original digital images were then enhanced using five processing algorithms. Six evaluators measured the length of each file on each image. The measurements from each processing algorithm and each file size were compared using repeated measures ANOVA and Bonferroni tests (P 0.05). Paired t test was performed to compare the measurements with the true lengths of the files (P 0.05). Results. All of the processing algorithms provided significantly shorter measurements than the true length of each file size (P 0.05). The threshold enhancement modality produced significantly higher mean error values (P 0.05), while there was no significant difference among the other enhancement modalities (P 0.05). Decrease in mean error value was observed with increasing file size (P 0.05). Conclusions. Invert, contrast/brightness and edge enhancement algorithms may be recommended for accurate file length measurements when utilizing storage phosphor plates. (Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007;103:280-4) Digital intraoral radiography has been accepted as an alternative to film-based radiography in endodontics. An advantage of digital imaging over conventional radiography is that images are not static, but may be manipulated by image processing to improve image quality. 1 Several image-processing algorithms, mainly adjustments for contrast and brightness have been evaluated for various diagnoses and treatment options. 1-3 It has been suggested that proper application of image processing and filtering may improve both the diagnostic quality and validity of digital images by delineating subtle differences in radiodensity and sharpening the distinction between neighboring tissues. 4-6 Many studies have investigated the effect of image enhancement on the diagnostic quality of digital images obtained either for the detection of periapical lesions and/orprognosisofendodontictreatment. 7,8 Ithasbeen a Assistant professor, Department of Oral Diagnosis and Radiology, b Professor, Department of Oral Diagnosis and Radiology, School of Dentistry, Ege University. c Associate professor, Department of Oral Diagnosis and Radiology, d Professor, Department of Restorative Dentistry and Endodontology, Received for publication Apr 20, 2006; returned for revision May 29, 2006; accepted for publication Jun 1, /$ - see front matter 2007 Mosby, Inc. All rights reserved. doi: /j.tripleo proved that enhanced digital images were diagnostically comparable to film images for the interpretation of root and canal morphology as well as final verification of the outcome of root canal treatment. In studies comparing the image quality of different imaging systems, 1,9,10 ithas been established that digital image enhancement and filtering may increase diagnostic visibility of endodontic file tips. On the contrary, it has been reported by some authors that even the enhancement and filtering cannot aid for the delineation and visibility of thin files. 2,11 Most of these studies have compared the subjective image quality of digital images based on observer performances; and therefore, their major limitation is reliance on subjective assessment of the image. 2,9,12,13 However, it has been advocated that it may be more important to use a computable objective measure to predict diagnostic accuracy rather than subjective assessments. 14 This is particularly important in endodontics because it will influence the final verification of the outcome of root canal treatment as it determines an accurate working length and consequently a possible successful treatment. It is, however, interesting to note that in studies comparing the validity of file length measurements using different imaging systems, the investigators compared the mean error values of the systems only in respect to conventional film measurements. 11,15-17 Itwas clearly stated that when evaluating the accuracy of a diagnostic method, a solid validation method is required to express the true state of disease 280

2 Volume 103, Number 2 Kal et al. 281 and/or measurement. 18,19 Accordingly, they should have compared also the measurements in respect to the true length in order to demonstrate the absolute accuracy of the systems. When those studies about the validity of file length measurements using digital enhancements were evaluated, it was surprising to observe that only one study had evaluated the measurements in respect to the true length of the files. 10 Consequently, it is possible to conclude that the amount and the type of processing that can improve the accuracy of endodontic file length measurements are still debatable. Therefore, the aim of this study was to compare the validity of various processing algorithms with regard to the measurement of endodontic file lengths using Digora storage phosphor plate (SPP) images. MATERIALS AND METHODS Twenty extracted permanent lower premolar teeth with single roots and root canals and endodontic files with ISO sizes 08, 10, and 15 were used for the study. One investigator (B.I.K.) performed the technical procedures for all teeth. Standard access cavities were prepared using a water-cooled diamond fissure bur in a high-speed hand piece. Gates-Glidden drills #2 and #3 were used to enlarge the coronal part of the root canals. Size 08 Hedstrom files (Dentsply Maillefer, Ballaigues, Switzerland) were inserted into each canal until the tip of the file was seen just at the apical foramen. Following the fixation of the files using resin composite material (Filtek Z250, 3M ESPE, St Paul, MN) and a file stopper, the experimental teeth were mounted in acrylic blocks with original premolar and molar teeth at either end creating natural contact points. A 10-mm rectangular orthodontic wire was also mounted in each acrylic block in the same plane as the endodontic file and placed parallel to the root axis. Because there is a radiographic magnification to some extent on any digital image, the wire served as a reference for the measurement of file lengths and as a calibration control for each original and enhanced digital image. The acrylic blocks were then placed on a supporting post (Pentamix, 3M ESPE) with the x-ray cone to allow exact parallel alignment. In order to keep the receptor perpendicular to the beam at all exposures and to provide consistent and reproducible exposition geometry, a Rinn-Endo-ray film holder (Dentsply/Rinn Corporation, Elgin, IL) was used. The standard geometric configuration was fixed at 30 cm source-to-object distance, and zero degrees vertical and horizontal angulations of the x-ray beam. A plexiglass with thickness of 15 mm was inserted between the x-ray tube and the acrylic block to simulate the effect of soft tissue during all exposures. Radiographic images of each experimental Fig. 1. Original and enhanced digital images of file size 08 (O Original; C Contrast/brightness; S Sharpen; E Find edges; I Invert; T Threshold). tooth were obtained with the Digora SPPs (Soredex, Orion Corporation, Helsinki, Finland) with the x-ray unit operating at 65 kvp and 10 ma for 0.16 seconds (Trophy Radiologie, Vincennes, France). Following the exposures, the Digora plates were scanned immediately in the Digora fmx-scanner previously calibrated for a highest exposure of 1 second and the resulting images were transferred as 8-bit TIFF files to a personal computer (LiteOn, Dong Guang, China) with operating system of Microsoft Windows XP (Microsoft Corp, Redmond, WA). Thereafter, the file in each canal was removed and the distance from the file stop to the tip was measured 3 times using a digital caliper to the nearest mm by the same researcher and the mean was recorded as the true length. The same procedure was repeated for sizes 10 and 15 files. Sixty original images were obtained from 20 teeth exposed with 3 different size files. After taking each original radiograph, one researcher manipulated images using the Image J 1.34 software package (National Institutes of Health, Bethesda, MD). Image J was chosen as the processing software because different softwares having similar functions with common names for various processing techniques may have different effects on the image 20 and because original Digora software did not contain sharpen, edge enhancement, and threshold algorithms. Five different processing algorithms, namely contrast and brightness adjustment, find edges, invert, threshold and sharpen were applied as provided by the

3 282 Kal et al. February 2007 Table I. Absolute mean error standard deviation and minimum-maximum values (mm) of endodontic file lengths for various processing algorithms* File size 08 File size 10 File size 15 Mean SD Min-Max Mean SD Min-Max Mean SD Min-Max Original Threshold Sharpness Invert Find edges Con./bright *The numbers in bold show the lowest mean error values for each file size. software (Fig. 1). Thus, there were 360 original and enhanced images to be assessed. Three radiologists and 3 endodontists with a mean age of 35.5 (range 28-56) and mean clinical experience of 11.8 (range 6-28) acted as evaluators and were asked to measure the distance (length) of each file from the apical-most tip of the file to the stopper on each original and enhanced image. The viewing sessions were performed in a darkened room to minimize glare. Digital images were displayed on a 17-inch SVGA color monitor ( pixels; LiteOn) and the measurements were performed using the distance measurement tool provided by the above-mentioned software. All measurements done by observers were first calibrated to the known reference length (10 mm) of the rectangular wire using the above-mentioned software s distance calibration tool. Image J software provided measurements to the nearest mm as well. A mean (from all 6 observers) was calculated for measurements obtained from each image (original and enhanced using 5 different algorithms) and file size. To simplify statistical analysis, the mean error in each reading was calculated as the absolute value of the difference between an observer s measurement and the true length in case negative or positive values would cancel each other out. The mean absolute errors were then compared for all observers for each image processed using different algorithms and each file size using repeated measures analysis of variance (ANOVA) and Bonferroni tests (P.05). Paired t test was performed to compare the observers measurements with the true length for each image processed using different algorithms and each file size with the statistical significance set at.05 for all statistical analyses (SPSS 3.0 for Windows, SPSS Inc, Chiago, IL). RESULTS Repeated measures ANOVA showed significant differences among processing algorithms and 3 file sizes (P.05), but not among the observers (P.05). The results presented in Table I demonstrate the mean absolute error values of 3 endodontic files measured using various enhancement modalities. The inverted images presented the lowest mean error values for size 08 files, while contrast/brightness adjusted images for size 10 files and edge enhanced images for size 15 files (Table I). This means that above-mentioned enhancements provided file lengths that were quite similar to the true length of each file size. However, no significant differences were observed among original, sharpened, inverted, edge enhanced, and contrast/ brightness adjusted images with regard to the mean error value for each of the 3 file sizes (P.05). The threshold enhancement modality produced significantly higher mean error values than other image enhancement modalities for all file sizes, indicating the least accurate measurements (P.05). When the accuracy of the measurements within each image-processing algorithm was compared with respect to the file sizes, a decrease in mean error value was observed with increasing file size for all processing algorithms used in this study (P.05). In other words, the accuracy of measurements increased with the increasing file size. The paired t test revealed that the observers measurements for file sizes 08, 10, and 15 on all images were significantly shorter than the true length (P.05), except for the inverted, contrast/brightness, and edge enhanced images for size 15 file (P.05). DISCUSSION The main aim of image processing is the process of producing images with sufficient detail to reveal the information that is already there but not visible to the naked eye. It is undisputed that diagnostic impact depends on the task, the quality of source data, and the kind of image processing applied. In studies testing the effect of image processing for the evaluation of endodontic length measurements, either contrast and brightness adjustments and/or magnification (zoom) algorithms were evaluated and images were mostly ac-

4 Volume 103, Number 2 Kal et al. 283 quired using charge-coupled device (CCD) sensors. 8,10,21 The present study is the first for evaluating the influence of multiple image-processing algorithms on the measurement accuracy of endodontic file lengths and first for comparing the effect of various processing algorithms on images acquired using only Digora SPPs. In the present study, the enhancement modalities were applied as originally provided by the software to avoid adverse effect of the variety of observer experience and individual image preferences in digital radiography. The use of automatic settings for each enhancement gave an equal chance of comparable performance to all observers. Since every other parameter including the radiation source, display, and measurement conditions and software were kept constant, the results reflect the direct effect of processing algorithms on the measurement accuracy of endodontic files. Although different algorithms provided varying levels of accuracy for each different file thicknesses, all processing algorithms used in this study presented significantly shorter measurements than the true length for each of the file sizes. The only exception occurred with the use of inverted, edge enhanced, and contrast-brightness enhanced images for file size 15, which presented measurements closest to the true length. On the other hand, it was found that the threshold algorithm gave rise to significantly shorter measurements than the other 4 algorithms for each of the file sizes. It was shown that fine details may be lost in threshold enhanced images, particularly file tips easily fade from the image. 8 Accordingly, the threshold modality by rounding off the single image points (pixels) according to a threshold may have made the file tip appear less clear and thus the file length somewhat shorter on digital images. There is evidence that certain enhancements of digital images may improve accuracy of detection and quantification of carious lesions, for the detection of periapical bone lesions and the visualization of root canal anatomy. 7,22,23 However, none of the image-processing algorithms significantly improved the measurement accuracy of size 08 and 10 files in this study and only few of them increased the measurement accuracy of size 15 files. It has been stated that size 08 and 10 files are not adequate for working length radiographs because small file tips fade out and are usually not visible. 24 However, it is sometimes inevitable to use these files in small canals such as in mesiobuccal roots of maxillary and mesial roots of mandibular molars. There are 2 reasons why no difference was found with the processing algorithms for thin files. First, pixel size of Digora images as given by the manufacturer is mm 5%. 25 Onthe other hand, the thicknesses of the size 08, 10, and 15 files at the tip are given as 0.08, 0.10, and 0.15 mm respectively. According to the Nyquist frequency, the spatial resolution of the system must equal twice the spatial frequency of the smallest detail to be detected. 26 Therefore, the minimum size of the detectable detail with Digora images would be 0.14 mm, which is approximate to the tip of file size 15. As can be expected, detectability of the file tip and consequently the accuracy of length measurement increased as the file got thicker and covered more pixels of the Digora image. Another reason for the underestimation of file length measurements for thin files may be that the Digora storage phosphor plates when exposed with exposure times as recommended by the manufacturer, resultant original images exhibit less contrast between different image areas and increased amount of quantum noise. 27,28 This increased amount of noise appearing as a fog on the image cannot be eliminated; in other words, faint details of the thin file tips could not be delineated in images with originally impaired density even with the aid of processing algorithms. At this point, it should be emphasized that with the use of particular processing algorithms (invert and contrastbrightness adjustment) the mean error values were reduced to mm for size 08 and mm for size 10 files, respectively. Although statistically significant, mm divergence from the true length could be regarded as clinically insignificant for size 10 files. Since the performance of processing algorithms improved with the increasing file size, with the use of above-mentioned processing algorithms sizes 10 and above may be recommended when using SPPs for endodontic file length determinations. Exposing the Digora SPPs above the exposure times recommended by the manufacturer might have optimized the response of the resultant images to the digital processing but such exposures exceeded those needed clinically. Another possibility of using size 08 and thinner files for endodontic length determination using storage phosphor plates may be provided by using files of a highdensity metal alloys. In the present study, the measurement of file lengths using 5 image-processing filters were undertaken with the goal of finding the best enhancement algorithm that would reliably increase the measurement accuracy of endodontic file length. In the clinical situation, discrimination of small file tips may become more difficult because of problems in selection of the optimal exposure time and the effects of scatter radiation, as well as differences in bone density. Clinical studies to compare the effects of various processing filters with SPP images will be required for further evaluation of the clinical usefulness of various processing algorithms.

5 284 Kal et al. February 2007 REFERENCES 1. Li G, Sanderink GCH, Welander U, McDavid WD, Näsström K. Evaluation of endodontic files in digital radiographs before and after employing three image processing algorithms. Dentomaxillofac Radiol 2004;33: Friedlander LT, Love RM, Chandler NP. A comparison of phosphor-plate digital images with conventional radiographs for the perceived clarity of fine endodontic files and periapical lesions. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2002;93: Hintze H, Wenzel A, Frydenberg M. Accuracy of caries detection with four storage phosphor systems and E-speed radiographs. Dentomaxillofac Radiol 2002;31: Wenzel A. Computer-aided image manipulation of intraoral radiographs to enhance diagnosis in dental practice: a review. Int Dent J 1993;43: Shrout MK, Russell CM, Potter BJ, Powell BJ, HildeboLt CF. Digital enhancement of radiographs: can it improve caries diagnosis? J Am Dent Assoc 1996;127: Eickholz P, Riess T, Lenhard M, Hassfeld S, Staehle HJ. Digital radiography of interproximal bone loss: validity of different filters. J Clin Periodontol 1999;26: Kullendorff B, Nilsson M. Diagnostic accuracy of direct digital dental radiography for the detection of periapical bone lesions. II. Effects on diagnostic accuracy after application of image processing. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1996;82: Shearer AC, Horner K, Wilson NH. Radiovisiography for length estimation in root canal treatment: an in-vitro comparison with conventional radiography. Int Endod J 1991;24: Ellingsen MA, Harrington GW, Hollender LG. Radiovisiography versus conventional radiography for detection of small instruments in endodontic length determination. Part 1. In vitro evaluation. J Endod 1995;21: Woolhiser GA, Brand JW, Hoen MM, Geist JR, Pikula AA, Pink FE. Accuracy of film-based, digital and enhanced digital images for endodontic length determination. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2005;99: Lozano A, Forner L, Liena C. In vitro comparison of root-canal measurements with conventional and digital radiology. Int Endod J 2002;35: Ellingsen MA, Hollender LG, Harrington GW. Radiovisiography versus conventional radiography for detection of small instruments in endodontic length determination. Part 2. In vivo evaluation. J Endod 1995;21: Leddy BJ, Miles DA, Newton CW, Brown CE. Interpretation of endodontic file lengths using Radiovisiography. J Endod 1994;20: Cosman PC, Gray RM, Olshen RA. Evaluating quality of compressed medical images: SNR, subjective rating and diagnostic accuracy. Proceedings of the IEEE 1994;82: Versteeg KH, Sanderink GCH, van Ginkel FC, van der Stelt PF. Estimating distances on direct digital images and conventional radiographs. J Am Dent Assoc 1997;128: Vandre RH, Pajak JC, Abdel-Nabi H, Farman TT, Farman AG. Comparison of observer performance in determining the position of endodontic files with physical measures in the evaluation of dental x-ray imaging systems. Dentimaxillofac Radiol 2000;29: Eikenberg S, Vandre R. Comparison of digital dental x-ray systems with self-developing film and manual processing for endodontic file length determination. J Endod 2000;26: Wenzel A, Hintze H. The choice of gold standard for evaluating tests for caries diagnosis. Dentomaxillofac Radiol 1999;28: Wenzel A, Verdonschot EH. Some considerations in the evaluation of diagnostic tests in dentistry. Dentomaxillofac Radiol 1994;23: Lehmann TM, Troeltsch E, Spitzer K. Image processing and enhancement provided by commercial dental software programs. Dentomaxillofac Radiol 2002;31: Hedrick RT, Dove SB, Peters DD, McDavid WD. Radiographic determination of canal length direct digital radiography versus conventional radiography. J Endod 1994;20: Naoum HJ, Chandler NP, Love RM. Conventional versus storage phosphor-plate digital images to visualize the root canal system contrasted with a radiopaque medium. J Endod 2003;29: Møystad A, Svanaes DB, Risnes S, Larheim TA, Gröndahl HG. Detection of aproximal caries with a storage phosphor system. A comparison of enhanced digital images with dental x-ray film. Dentimaxillofac Radiol 1996;25: Walton R, Torabinejad M. Principles and practice of endodontics. Philadelphia: WB Saunders; p Digora FMX, technical data. Available at: Accessed April 19, Versteeg CH, Sanderink GCH, van der Stelt PF. Efficacy of digital intra-oral radiography in clinical dentistry. J Dent 1997;25: Borg E, Gröndahl HG. Endodontic measurements in digital radiographs acquired by a photostimulable, storage phosphor system. Endod Dent Traumatol 1996;12: Chotas HG, Floyd CE Jr, Dobbins JT 3rd, Ravin CE. Digital chest radiography with photostimulable storage phosphors: signal-to-noise ratio as a function of kilovoltage with matched exposure risk. Radiology 1993;186: Reprint requests: Bilge Hakan Şen, PhD Ege Universitesi Dishekimligi Fakultesi Endodonti Bilim Dali Bornova, Izmir, Turkey bilge.hakan.sen@ege.edu.tr