Faculty of Odontology, Gothenburg University, Gothenburg, Sweden. Received 8 September ... Keywords: X-ray film; dental radiography; dental caries; ROC analysis. Introduction .... Melbourne, Florida, USA) after processing in an automatic ...
Influence of film fog on radiographic caries diagnosis B. Svenson, H.-G. Grondahl· and A.-M. Lindvall· Department of Oral Radiology, Postgraduate Dental Education Center, Orebro and "Department of Oral Radiology, Faculty of Odontology, Gothenburg University, Gothenburg, Sweden
Received 8 September 1989 and in final form 9 March 1990 Increasing film speed may potentially jeopardize diagnostic quality because of a more rapid build up of fog and an ensuing decrease in image contrast. A standardized amount of fog was simulated by pre-exposing E-speed dental films before use to obtain images of extracted teeth with and without approximal caries. Three experiments were performed. In the first, no effort was made to compensate for the increased density caused by fog. In the other two, density was kept constant by reducing exposure and developing time respectively. Nine dentists recorded the presence and absence of caries using a rating scale to indicate diagnostic confidence. As a measure of diagnostic accuracy, the area under the binormal ROC curve was used and the values averaged across observers. The results showed that fog, up to a level of base plus fog of 0.6 optical density units, had no influence on the diagnostic accuracy in the absence of any compensation. When exposure was lowered to compensate for the increased density, significantly lower diagnostic accuracy was found. On the other hand, when developing time was decreased, no significant difference was found. Observers ranked radiographs with higher levels of fog as being of lower quality. However, no significant correlation was found between subjective ranking of image quality and diagnostic accuracy obtained from the same radiographs. It is concluded that relatively high levels of fog do not exert a negative influence on approximal caries diagnosis and no attempt should therefore be made to compensate for the increased density. Keywords: X-ray film; dental radiography; dental caries; ROC analysis
Introduction All X-ray films show a certain density above that caused by the base of the film. Different definitions of this extra density, which is referred to as fog, exist l - 3 • In this study we have followed Barr and Stephens" and defined it as the increase in film density from factors other than those due to patient exposure. Fogging is influenced by exposure to background radiation, which in tum depends on both storage conditions'"? and film speed6 ,7 ,1O, as well as by the type of developing solution ll , 12 and its temperature'". It has long been accepted that fogging of X-ray films above a certain level is undesirable because it lowers the radiographic contrast/ and may thus contribute to misleading diagnoses". An optical density (00) value of base plus fog in excess of 0.20 was considered objectionable by Ter-Pogossian", while Kaffe et al. 12 accepted a value of up to 0.25. The density caused by the base of the film can easily be compensated for by increasing the viewing light. Lundh", therefore, ignored this factor and suggested that a limit of 0.30 above base should be permissible. Since a typical value for base density is 0.07 1 , the fog levels considered acceptable by these authors can be seen to vary widely. Studies on the impact of fog levels on diagnostic accuracy in oral radiology are sparse I4 ,15 . The speed of dental X-ray films has increased
© 1990 Butterworth-Heinemann for 0250-832 x 90/020105-04
progressively since they were first introduced'" and has quadrupled since the mid-1950s17 • As a consequence, there has been an equivalent reduction in radiation dose to both patients and personnel l 8-20 . However, since an increase in fog is associated with an increase in film speed, it is possible that there has been a fall in diagnostic accuracy. At a fog level of 0.30, half the film already consists of exposed silver halide grains and is therefore unable to collect diagnostic informatiorr". The diagnosis of approximal caries is an aspect of oral radiology which is particularly contrast dependenr'". In addition, it is very relevant to general practice where it exerts a major influence on treatment decisions 23 , 24 . Therefore, the aim of this study was to evaluate the influence of fogging on the accuracy of radiographic caries diagnosis. Specifically, we wanted to answer the following questions: (1) Does increased film fog decrease the diagnostic accuracy of small approximal lesions: (a) at normal exposure and developing; (b) when exposure is reduced to compensate for increased film density due to fog; (c) when developing time is shortened to compensate for increased film density caused by fog? (2) Does the observers' subjective rating of the image quality correlate with the diagnostic accuracy of the same radiographs?
IADMFR
Dentomaxillofac. Radiol., 1990, Vol. 19, August
lOS
Film fog and radiographic caries diagnosis: B. Svenson et al. Materials and methods The study was performed on 100 extracted nonhypoplastic human premolars. The 200 approximal surfaces were examined visually and by probing and 126 were recorded as intact and 74 carious. Of the latter, cavitation was found in 29, while 45 showed discolourations only, indicating the presence of subsurface caries. The teeth were mounted in plaster blocks in groups of four and placed as close as possible to each other with the most prominent part of the approximal surfaces positioned at the same vertical level. The radiographs were taken with a Minray DC (Soredex, Orion Corp. Ltd, Helsinki, Finland) Dental X-ray machine operating at 60 kV and 2.5 mmAI filtration. Kilovoltage and exposure times were checked before the study by means of a Digi-X unit (RTI Electronics, Molndal, Sweden). To achieve reproducible radiographs each plaster block was placed in a specially designed holder connected to the X-ray machine with a metal tube. The focus-film distance was 40 em and the beam collimated to a field size of 3 x 4 em. A 1 em thick layer of soft tissue-equivalent material (courtesy Dr Richard L. Webber, School of Dentistry, University of Alabama, Alabama, USA) was placed in front of the teeth. Ektaspeed (Eastman Kodak Co., Rochester, NY, USA) dental X-ray films from the same batch were used. The unexposed films had a base plug fog density of 0.20 as measured by a digital densitometer (Victoreen Digital Densitometer II, Victoreen Inc., Melbourne, Florida, USA) after processing in an automatic processor (Siemens Pantomat PlO, Hope Industries, Willow Grove, PA, USA) with G153 developer and G353 fixer (AgfaGevaert NV, Mortsel, Antwerp, Belgium) at a temperature of 26°C. Processing time was 6 min. During the experiment the stability of the processing was checked at regular intervals. To obtain experimental films with varying amounts of fog, four sets of fresh dental X-ray films with the same emulsion number were pre-exposed for different exposure times. These films showed mean density values of base plus fog of 0.30, 0.40, 0.50 and 0.60. To obtain a clinically comparable density value for the premolar teeth in the study, 61 diagnostically acceptable bitewing radiographs were selected and the density measured with a densitometer (MacBeth TD 502, Kollmorgen Corp., Newburgh, NY, USA) in dentine close to the proximal enamel in one premolar from each radiograplr". The mean density was 0.89. The experimental films with a base plus fog of 0.20 were subsequently used to obtain radiographs of the experimental teeth so that the density of the dentine was as close as possible to 0.89. The exposure times required to achieve this are referred to below as the normal exposure. Three experiments were performed using films differing in fog. In the first, the normal exposure was used, leading to increasing density with increasing fog. In the second, exposure time was changed so as to achieve a consistent average density in the dentine of 0.89 irrespective of the amount of fog. Thus, the exposure time was decreased with increasing fog. The films with base plus fog values of 0.30, 0.40, 0.50 and 0.60 were used and those with a base plus fog density of 0.20 from the first experiment served as controls. All 106
Dentomaxillofac. Radiol., 1990, Vol. 19, August
films were again developed automatically. In the third experiment, normal exposure was employed and the films manually developed (Kodak Dental X-ray developer, Kodak-Pathe, Chalon-sur-Saone, France) at a temperature of 22°C. To obtain radiographs with a dentine density as close to 0.89 as possible, the films with fog levels of 0.20 were normally developed but those with levels of 0.30, 0.40, 0.50 and 0.60 underdeveloped. Thus, the greater the fog the shorter the developing time. All radiographs were mounted (Trollhatteplast AB, Trollhattan, Sweden) and nine dentists recorded caries with the aid of a magnifying viewer (X-Produkter, Malmo, Sweden) and viewbox with constant light intensity. Each observer was asked to select one of five ratings to represent his level of confidence that a lesion was present or not as follows: 1 = definitely not caries, 2 = probably not caries, 3 = questionable, 4 = probably caries, 5 = definitely caries. The observers were not calibrated before the reading of the radiographs and did not receive any information about the number of carious lesions. The diagnostic decisions were compared with the actual presence or absence of caries for each surface. The results were analysed by means of the ROC-curve technique". The index, A z , of diagnostic accuracy was used which ranges from a lower limit of 0.5 for an ROC lying along the diagonal, reflecting performance at chance level to an upper limit of 1.0, indicating perfect performance. The data were analysed using 'ROCFIT' (obtained from Charles Metz, Department of Radiology, The University of Chicago, Chicago, IL, USA). This computer program calculates maximum likelihood estimates of ROC areas as well as relevant parameters of performance. The area under the ROC curve was calculated for each fog level in each experiment by averaging the individual areas of the nine observers. Statistical analysis was performed by using Tukey's test, a test for multiple comparisons, to determine where significant differences existed between fog levels within the experimental groups and between fog levels across experimental groups. P values