Effect of ambient light and monitor brightness and ... - BIR Publications

5 downloads 0 Views 113KB Size Report
settings on the detection of approximal caries in digital ... approximal carious lesions, ambient light levels less than 50 lux were significantly better than.
Dentomaxillofacial Radiology (2008) 37, 380–384 ’ 2008 The British Institute of Radiology http://dmfr.birjournals.org

RESEARCH

Effect of ambient light and monitor brightness and contrast settings on the detection of approximal caries in digital radiographs: an in vitro study K Helle´n-Halme*,1, A Petersson1, G Warfvinge2 and M Nilsson1 1 Department of Oral and Maxillofacial Radiology, Faculty of Odontology, Malmo¨ University, Malmo¨, Sweden; 2Department of Oral Pathology, Faculty of Odontology, Malmo¨ University, Malmo¨, Sweden

Objectives: The aim of this study was to investigate how brightness and contrast settings of the display monitor and ambient light level (illuminance) in the viewing room affect the clinician’s ability to diagnose carious lesions in digital radiographs. Methods: Standardized radiographs were taken of 100 extracted teeth. Seven observers evaluated the images for approximal carious lesions twice, once under 50 lux and once under 1000 lux room illumination. Monitor brightness and contrast were varied ¡50% and ¡6%, respectively, to mimic the normal limits of monitor adjustment by an inexperienced user and one optimal setting. This was done by adjusting radiograph brightness and contrast by ¡25%. Thus, five radiographs of each tooth were evaluated. Receiver operating characteristic (ROC) analyses were performed. Histological examinations of the teeth served as the criterion standard. A paired t-test was used to evaluate whether differences in the areas under the ROC curves were significant and kappa was used to evaluate intraobserver agreement. Results: When a monitor with optimal brightness and contrast settings was used to detect approximal carious lesions, ambient light levels less than 50 lux were significantly better than levels above 1000 lux (dentin and enamel lesions, P , 0.01; dentin lesions, P , 0.02). Increasing the contrast setting of the monitor by 6% did not change these results; 50 lux was still significantly better than 1000 lux (enamel lesions, P , 0.01; dentin and enamel lesions, P , 0.02) for evaluating radiographs. Intraobserver agreement differed from fair to good. Conclusions: Reducing ambient light to less than 50 lux significantly increased the accuracy of diagnosing approximal carious lesions on a monitor with an optimal brightness setting and an optimal or slightly higher than optimal contrast setting. Dentomaxillofacial Radiology (2008) 37, 380–384. doi: 10.1259/dmfr/26038913 Keywords: display monitor, digital radiography, dental caries, receiver operating characteristic curve

Introduction The quality of digital radiographic images is dependent on each part in the imaging chain being maintained at a high-quality level. One of the weaker links in the process, according to Samei,1 appears to be the monitor. Another factor that probably affects diagnostic outcome is the amount of ambient light *Correspondence to: Kristina Helle´n-Halme, Department of Oral and Maxillofacial Radiology, Malmo¨ University, SE-205 06 Malmo¨, Sweden; E-mail: [email protected] Received 1 February 2007; revised 21 August 2007; accepted 19 September 2007

(illuminance) in the operating room. The American Association of Physicists in Medicine (AAPM)2 recommends an illuminance of less than 50 lux when monitors with a luminance of less than 250 cd m22 are being used in the same room. Standard monitors in general dental practices usually have a luminance of 100–200 cd m22. In Sweden, the Swedish Standards Institute3 (SIS) sets standards for environmental conditions in different working situations. SSEN 12464-13 includes standards for ambient light conditions in a dental practice. According to these

Caries detection K Helle´n-Halme et al

standards, about 1000 lux (a very bright light) should be used in a dental operating room. This light level is of course contradictory to the level recommended by the AAPM for monitor use. Cederberg et al4,5 found that the ambient light level had no significance in the diagnosis of artificial carious lesions, but Li et al6 found that the observers could determine the lengths of endodontic files in root canals more accurately in a room with dimmed light than in a room with ordinary artificial light. In a field study,7 dentists were visited at their clinics and their digital systems were evaluated. Of special interest were the monitors used to evaluate digital radiographs. It was found that many general practice dentists evaluated radiographs on monitors that were placed in a brightly lit operating room and that reflected ambient light to a high degree. The same study7 found that fewer low-contrast details could be observed in a room with an ambient light level recommended for dental operating rooms3 than in rooms with lower illuminance. Brightness and contrast can be adjusted on most modern monitors. It must be stressed that altering these parameters in an imageprocessing programme does not usually produce the same result. It was found that general practice dentists were unaware that these parameters could be changed and therefore did not realize that they also affected the final image perception. It was concluded that faint contrast objects were more easily detected when monitor brightness and contrast had been adjusted using standard test images.2 Bitewing examination is the principal method dentists use to diagnose approximal carious lesions. Few studies have investigated how ambient light affects the diagnosis of caries on digital radiographs,4,5 and in these studies the level of light was not specified. To our knowledge no studies on how display monitor brightness and contrast influence radiological diagnosis of approximal caries have been published. The aim of this study was to evaluate whether the ambient light level (according to both standard regulations for a dental operating room and recommendations given for a viewing room) influences the diagnosis of carious lesions in digital images. A further aim was to investigate in a standardized experimental set-up how display monitor brightness and contrast affect the diagnosis of approximal carious lesions. It was hypothesized that the diagnosis of carious lesions would improve when radiographs were viewed on an optimally adjusted display monitor and when the ambient light level was reduced below SIS recommendations for dental operating rooms.

Materials and methods 100 human teeth (40 premolars and 60 molars) were selected from a group of extracted teeth whose approximal surfaces on visual inspection were either

381

Figure 1 Device for placing the block of teeth, digital sensor (arrow) and rectangular collimator of the X-ray machine in a standardized way

intact or had carious lesions of different sizes. Half (50%) of the teeth chosen were visually healthy and half had carious lesions of different extensions in the approximal surfaces. Only teeth with small carious lesions with or without approximal cavities were included in the study. The teeth were mounted in 30 blocks made of PRESIDENT putty (Colte`ne Whaledent AG, Cuyahoga Falls, OH), with 3 or 4 teeth in each block. A plate of 1 cm thick PlexiglasH was used to simulate soft tissue in the geometry used for radiographs. Standardized radiographs were taken using a device that kept the block with the teeth, the soft tissue supplement, the digital sensor and the collimator in place (Figure 1). Each block was exposed once using a dental digital system (Schick CDR Wireless 2; Schick Technologies Inc. Long Island City, NY) coupled to a Prostyle Intra X-ray machine (Planmeca Oy, Helsinki, Finland). Exposure settings were 63 kVp, 8 mA and 0.1 s. The distance from the X-ray focus to the object was 22 cm. The diagonal measurement of the monitor used to view the images (Dell 456; Dell Inc., Round Rock, TX) was 48 cm. The monitor was visually adjusted with AAPM test images TG18-QC and TG18-CT.2,7 Brightness was set to 80% and contrast to 50% to obtain the best image subjectively. Test images TG18LN12-01, -09 and -182,7 were used to ensure uniform monitor luminance for all observations. The monitor was cleaned with a glass cleaner (Spectra; Nordex, Nilfisk-Advance, Stockholm, Sweden) before each observation session. The same monitor was used for all evaluations. To evaluate whether detection of carious lesions depended on the settings of the display monitor, the 30 images were altered in brightness and contrast in a standardized way with Paint Shop Pro 4 software (v4.12 Shareware; Jasc Inc., Eden Prairie, MN) and stored on disc. Changes were made to mimic radiographic differences that commonly occurred in a clinic. Brightness and contrast were varied by ¡25% each. Dentomaxillofacial Radiology

Caries detection K Helle´n-Halme et al

382

The percentage denotes per cent units selectable in the software program. Only one parameter, brightness or contrast, was altered in each radiograph. 5 images of each radiograph, 150 images in total, were made: standard brightness and contrast, +25% brightness/ standard contrast, 225% brightness/standard contrast, standard brightness/+25% contrast, and standard brightness/225% contrast. To evaluate how well the altered radiographs corresponded to actual changes in monitor brightness and contrast, luminance (luminance meter LS-100; Konica-Minolta, Langenhagen/ Hannover, Germany) was measured on test images TG18-LN12-01, -09 and -18.2 For the monitor used in this study, the brightness change in the radiographs corresponded to an alteration of ¡50% in monitor brightness and the contrast change to an alteration of ¡6% in monitor contrast. Seven observers, one specialist in oral and maxillofacial radiology and six general practice dentists well experienced in caries diagnostics, evaluated the radiographs on two different occasions. All observers were familiar with digital radiography. The first evaluation was made in ambient light less than 50 lux (range 42– 49 lux) and the second in ambient light more than 1000 lux (range 1005–1025 lux). Ambient light levels were measured with a light meter (Light-O-Meter, P-11; Unfors, Billdal, Sweden). The observers were allowed to change the magnification of the radiographs, but not to display monitor settings for brightness and contrast or image brightness and contrast. The 150 radiographs were randomly mixed so that neither the 5 renditions of a single image nor all images with the same brightness and contrast occurred consecutively. The radiographs were displayed in the same way for each observer. The observers were asked to score their level of confidence about whether caries was present on the different approximal surfaces with this 5-point scale: 1 5 definitely not caries 2 5 probably not caries 3 5 questionable caries 4 5 probably caries 5 5 definitely caries Intraobserver agreement was determined by asking each observer to evaluate 60 approximal surfaces twice at an interval of at least 14 days. Both evaluations were made in an evaluation room with an ambient light level of less than 50 lux. The same monitor used in the other evaluations in this study was used to determine intraobserver agreement. After the observers completed their evaluations, the teeth were histologically examined.8,9 Each tooth was cut in 1 mm slices with a low speed saw and diamond blade (IsoMetH II-1180 Low Speed Saw and IsoMet, Diamond Wafering Blade, 460.012 (10.2 cm60.3 mm); Buehler Ltd, Greenwood, IL). The sliced teeth were attached to a microscope glass with transparent glue and evaluated for caries by two observers (KH-H, GW) under a light microscope (magnification 10640). A consensus was reached in cases of disagreement. Dentomaxillofacial Radiology

Caries was defined as present when demineralization was observed as opaque-white to dark-brown colour changes. The 200 approximal surfaces of the teeth were graded according to a scale from 0 to 3: 0 5 sound, no visible lesion; 1 5 lesion confined to the enamel; 2 5 lesion involving the enamel and enamel–dentin border but not the body of the dentin; 3 5 lesion involving the enamel and undisputedly the body of the dentin. Statistical analysis Receiver operating characteristic (ROC) curves were used to analyse all radiographic evaluations.10 The criterion standard was the histological validation.8 The ROC curves depict the combined data from the seven observers. The areas (Az) under the curves were calculated. A paired t- test11 was used to analyse the effects of different monitor settings with ambient illumination of 50 lux or 1000 lux. The hypothesis was that diagnoses of carious lesions differed non-significantly in ambient levels of 50 lux and 1000 lux with different monitor settings for brightness and contrast settings. Differences were considered significant when P , 0.05. Intraobserver agreement was estimated using kappa12 statistics. The values were interpreted according to the guidelines of Landis and Koch13 adapted by Altman.11 The number of surfaces with intraobserver agreement was also calculated.

Results Histological evaluation of the 200 approximal surfaces of the teeth showed that 100 surfaces were sound and 100 surfaces had a carious lesion. The two observers disagreed in 31% of the cases. Disagreement was confined to grades 0 and 1. Table 1 presents the distribution of carious lesions by depth. All ROC curves represent combined data for all observers. Table 2 presents the mean areas under the ROC curves for all simulated monitor settings at illuminances of 50 lux and 1000 lux. This result indicates that for all monitor settings, mean areas (Az) under the ROC curves were larger in ambient light less than 50 lux. Differences in diagnostic accuracy between the two ambient light levels were significant for all carious lesions (P , 0.01) and for dentin carious lesions (grade 3) (P , 0.02) with an optimally adjusted monitor. Differences were also significant for all

Table 1 Results of the histological evaluation of the approximal surfaces of the teeth. 100 teeth in total Lesions (grade)

No. of surfaces

Per cent

Sound (0) Enamel caries (1) Enamel–dentin border (2) Dentin caries (3)

100 75 14 11

50.0 37.5 7.0 5.5

Caries detection K Helle´n-Halme et al

383

Table 2 Areas (Az) under the receiver operating charateristic curves, calculated as a mean for all seven observers at an illuminance of ,50 lux and .1000 lux. Approximal surface lesion type: enamel lesions 5 grade 1 and 2, dentin lesions 5 grade 3 Ambient light ,50 lux

.1000 lux

Lesion type

Lesion type

Simulated monitor settings

Enamel

Dentin

All caries

Enamel

Dentin

All caries

Optimal Brightness 250% Brightness +50% Contrast 26% Contrast +6%

0.551 0.557 0.549 0.536 0.564

0.725 0.682 0.684 0.705 0.685

0.598 0.594 0.586 0.587 0.600

0.530 0.548 0.534 0.523 0.540

0.667 0.676 0.671 0.675 0.673

0.567 0.584 0.569 0.564 0.581

carious lesions (P , 0.02) and for enamel carious lesions (grades 1 and 2) (P , 0.01) with a monitor contrast setting increased by 6%. Differences between the two ambient light levels with the other simulated monitor settings were non-significant. Values for intraobserver agreement were good for two of the observers (0.66, 0.65), moderate for one observer (0.57) and fair for four observers (0.38, 0.35, 0.30, 0.30). The mean number of surfaces for which the observers’ duplicate readings agreed was 72% (range 53–92%).

Discussion This study found that the ambient light level in the evaluation room is important and should be kept low when approximal carious lesions are being evaluated on digital radiographic images. It was also found that proper adjustment of monitor brightness and contrast is essential. It is difficult to compare studies that evaluate carious lesions. Bader et al14 found in a systematic review that many problems are linked to this kind of study. They concluded that descriptions in the literature are problematic concerning, for example, reporting of methods, selection criteria for the sample of teeth and histological validation method. One important issue that can influence the diagnostic outcome of radiographically detectable carious lesions is related to the type of lesion being evaluated. Extensive carious lesions are rarely misdiagnosed in a radiograph. None of the extracted teeth used in this study had extensive carious lesions. The proportions of sound and carious surfaces were equal and the number of dentin lesions was low. In 2002, around 60% of the approximal surfaces in Swedish 19-year-olds were caries free, but there are no statistics for the adult population.15 In vitro studies have several limitations, according to Bader et al,14 and generalizations from such studies are difficult to apply to clinical situations for many reasons. Extracted teeth are not representative of the dentition of a patient and the radiographic examination is performed under ideal conditions. In our study, we tried to simulate the clinical situation as closely as possible. Radiographic quality and viewing conditions were suboptimal. It is

our clinical impression that bright ambient light hampers radiographic caries diagnoses. When planning studies of this kind, the number of observers needed and the size of the object sample necessary to obtain statistical power must be considered. The number of observers in our study was seven, with several years of experience of caries diagnosis. To ensure credible generalization of a study result, the use of many observers is of course preferable, but according to Swets and Pickett16 little is gained by using more than seven observers in ROC studies such as this. In a previous study, Hintze et al17 showed that given the same number of evaluations the standard error of difference between ROC curve areas as a function of the number of surfaces and observers was slightly larger for study designs with few observers than with many observers. The same study17 found that the numbers of observers and evaluated surfaces reached a level where increases in either variable ceased to affect the uncertainty of the results. The number of tooth surfaces and number of observers we chose to use was based on information in the report of Hintze et al.17 We chose to use 200 approximal surfaces and 7 observers to evaluate them. All observers were experienced in the diagnosis of dental caries. No attempt to calibrate the observers before the test was made, because the study was designed to mimic the clinical situation. Overall performance of the observers was compared with other studies18–20 that reported the same range of true positive and false positive diagnoses of carious lesions, especially for enamel caries. The level of intraobserver agreement was similar to other studies regarding caries diagnosis.18,19 The histological validation method can also affect results in this kind of study. It is generally accepted that histological evaluation of serial tooth sections is a reliable method for determining the diagnosis of dental caries.8,9,20 The method used in this study complies with the accepted method for establishing a criterion standard. In studies such as this, preparations of the material for and production of the criterion standard are tasks that very much depend on the skills of the persons involved in this process. We were two observers and disagreed in cases limited to differences between healthy surfaces and small enamel lesions. That the areas under the ROC curves, especially for enamel caries, were close to 0.5 reflects well-known difficulties in radiographic caries diagnoses. Our use of Dentomaxillofacial Radiology

Caries detection K Helle´n-Halme et al

384

a microscope to evaluate approximal surfaces histologically, which makes it possible to detect very small carious lesions, is one factor that can improve accuracy. Such lesions are almost impossible to see in a radiograph because a tooth must suffer loss of mineral substance before the carious lesion is detectable. No previous studies of effects of altered viewing conditions on the diagnostic outcome of digital dental radiographs were found in the literature. This may be due to the numerous interacting factors and variables that affect digital imaging compared with traditional film radiography. The entire digital viewing situation is difficult to grasp and requires thorough and systematic analysis to understand the importance of each individual step in the process. A previous study7 found that dentists were able to detect more objects on a lowcontrast phantom when the light in the room was dimmed. The same study also found that more objects were observed in the test images after monitor brightness and contrast were adjusted. These results indicate a need for more research. This study found a significant improvement in the accuracy of caries diagnoses on a

standard monitor when AAPM2 recommendations concerning ambient light (i.e. illuminance less than 50 lux) were followed compared with an ambient light level of 1000 lux. Even though carious lesions in general are difficult to diagnose radiologically,9,20 the study showed that for all monitor brightness and contrast settings tested the observers seemed to do better in a room with low illuminance. One factor that contributes to this result is the human eye. When the pupil is dilated, such as in a room with low illuminance, the rods contribute substantially to the image perceived. The rods make it possible to see small variations in light intensity or small differences in greyscale between pixels on the monitor. Cones, which are more active at higher light levels and mediate high spatial resolution and colour perception, dominate the central portion of the retina. Rods are less active at higher light levels, and small contrast differences are more difficult to observe. In conclusion, digital dental radiographs being evaluated on a standard monitor should be viewed under low ambient light. The monitor should be adjusted for optimal brightness and contrast.

References 1. Samei E. Advances in digital radiography, categorical course syllabus. Samei E (ed). Oak Brook, IL: Radiological Society of North America (RSNA) Publication, 2003; pp 109–121. 2. American Association of Physicists in Medicine (AAPM): Task Group 18. Assessment of display performance for medical imaging systems. [Updated 2006 Jan 10; cited 2006 Sept 29]. Available from: http://deckard.mc.duke.edu/,samei/tg18. 3. SIS, Swedish Standards Institute. Light and lighting – lighting of work places. Part 1: indoor work places. SS-EN 12464-1, 2003 (in English). 4. Cederberg RA, Frederiksen NL, Benson BW, Shulman JD. Influence of the digital image display monitor on observer performance. Dentomaxillofac Radiol 1999; 28: 203–207. 5. Cederberg RA, Frederiksen NL, Benson BW, Shulman JD. Effect of different background lighting conditions on diagnostic performance of digital and film images. Dentomaxillofac Radiol 1998; 27: 293–297. 6. Li G, Sanderink GC, Welander U, McDavid WD, Nasstrom K. Evaluation of endodontic files in digital radiographs before and after employing three image processing algorithms. Dentomaxillofac Radiol 2004; 33: 6–11. 7. Helle´n-Halme K, Nilsson M, Petersson A. Digital radiography in general dental practice. A field study. Dentomaxillofac Radiol 2007; 36: 249–255. 8. Wenzel A, Hintze H. The choice of gold standard for evaluating tests for caries diagnosis. Dentomaxillofac Radiol 1999; 28: 132–136. 9. Hintze H, Wenzel A. Diagnostic outcome of methods frequently used for caries validation. A comparison of clinical examination, radiography and histology following hemisectioning and serial tooth sectioning. Caries Res 2003; 37: 115–124.

Dentomaxillofacial Radiology

10. Metz CE. Basic principles of ROC analysis. Semin Nucl Med 1978; 8: 283–298. 11. Altman DG. Practical statistics for medical research (1st edn). London, UK: Chapman and Hall, 1991. 12. Cohen J. A coefficient of agreement for nominal scales. Educ Psych Meas 1960; 20: 37–46. 13. Landis J R, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159–174. 14. Bader JD, Shugars DA, Bonito AJ. A systematic review of the performance of methods for identifying carious lesions. J Public Health Dent 2002; 62: 201–213. 15. World Health Organization; Global Oral Health – CAPP. WHO Oral Health Country/Area Profile Programme. WHO Collaborating Centre, Malmo¨ University, Sweden. [Cited 2006 Dec 5]. Available from: http://www.whocollab.od.mah.se/euro/ sweden/sweden.html 16. Swets JA, Pickett RM. Evaluation of diagnostic systems: methods from signal detection theory. New York, NY: Academic Press, 1982. 17. Hintze H, Frydenberg M, Wenzel A. Influence of number of surfaces and observers on statistical power in a multiobserver ROC radiographic caries detection study. Caries Res 2003; 37: 200–205. 18. Grondahl HG. Radiographic caries diagnosis. A study of caries progression and observer performance. Swed Dent J Suppl 1979; 3: 1–32. 19. Haugejorden O. A study of the methods of radiographic diagnosis of dental caries in epidemiological investigations. Acta Odontol Scand Suppl 1974; 32: 1–269. 20. Wenzel A. Digital radiography and caries diagnosis. Dentomaxillofac Radiol 1998; 27: 3–11 (Review).