Digital subtraction radiography in artificial recurrent caries detection P.V. Nummikoski, T.S. Martinez, S.R. Matteson, W.D. McDavid and S.B. Dove Department of Dental Diagnostic Science, University of Texas Health Science Center at San Antonio, Texas, USA
Received 8 April 1991 and in final form 26 November 1991 The diagnostic accuracy of digital subtraction radiography in detection of artificial recurrent caries lesions was assessed in this project. The use of digital subtraction radiography has been shown to markedly increase the accuracy of the detection of destruction in the periodontal bone, but the method has not been evaluated in secondary caries detection. Defects of three different sizes, simulating recurrent caries, were sequential1y prepared in the interproximal cavity preparation margins of 28 teeth. Two composite restorative materials with different radiographic densities were used as posterior restorations, and a radiograph was obtained of each defect size and restorative material. The radiographs were digitized and subtracted from the reference images, and the conventional radiographs and the subtraction images were evaluated by seven observers. The data were analysed with ROC statistics. Subtraction radiography was found to be superior to conventional radiography in recurrent caries detection, mainly by reducing the false-positive diagnoses. The radiopacity of the restorative material had a significant effect on accuracy with conventional but not with subtraction radiography. / Keywords: Radiography; subtraction technic; dental caries; dental restoration, permanent Dentomaxillofac. Radial., 1992, Vol. 21, 59-64, May
Introduction Recurrent caries in previously restored teeth have been shown to account for from 42% to 88% of the need for restorative dentistry":'. Progression into the tooth is rapid and Leinfelder" reported that once detected clinically or radiographically, caries under a composite restoration can breach the pulpal chamber in 6-8 months. For these reasons, the early detection of caries associated with a previously restored tooth is an important component of the examination of dental patients. Present caries detection methods
In dental practice today, diagnostic methods for recurrent caries include clinical probing, transil1umination and intra-oral radiographs. The latter are considered necessary especially with posterior teeth, because of their wide faciolingual dimensions and the interruption of light flow at the margins of the restorations which make both the transillumination and probing unreliable. However, the effectiveness of standard radiographic methods in detecting recurrent caries has been questioned. Gratt et al. 5 and Douglass et al. 6 concluded that the sensitivity of dental radiographs in detecting early dental caries is only 60% or less. The actual depth of a carious lesion is consistently underestimated because of the relative insensitivity of the film-based imaging system and its inability to reveal early tissue demineralization among the structured and unstructured noise of the image. Silverstone? found that when a lesion appearing to be limited to the outer enamel is first detected on a bitewing radiograph,
the demineralization has already penetrated several hundred micrometers into the dentine. In view of the treatment ramifications of advanced recurrent caries, the use of sensitive diagnostic methods is critical to enhance early detection of the disease. Digital subtraction radiography method
Digital image subtraction (DSR) has been shown to be an accurate diagnostic tool in alveolar bone evaluation'"!", With this method, once the radiographic images are aligned and digitized into the computer memory, the image of the mask film is electronically subtracted from that of the subsequent film so that the resultant image represents the difference in density between the two films. The technique results in enhancement of the subtle changes that are difficult, if not impossible, to visualize from the original radiographs amongst the high structural (anatomical) noise. Ortman et al. II showed that 5% bone mineral loss can be visualized with the use of a wel1-control1ed DSR method. In comparison, more than 30% bone mineral loss is required before it can be detected visual1y from radiographic film. In addition, the subtraction technique, by altering the appearance of the pathological changes so that they are more conspicuous, results in a higher degree of interobserver agreernent'". The effect of dental restorative materials
The use of composite posterior restorative material is increasing in practice today 13. Several manufacturers Dentomaxillofac. Radiol., 1992, Vol. 21, May
59
Digital subtraction radiography in caries detection: P. V. Nummikoski et al. use radiopaque components in these materials in order to render them more visible on the radiograph, and this has been found to have an effect on the detectability of the adjacent caries. Matteson et al. 14 reported that for simulated recurrent caries in extracted teeth, the disease was detected on periapical films more effectively when a radiopaque composite was present instead of amalgam. On the other hand, the falsepositive rate for recurrent caries was significant, especially for teeth restored with radiolucent composites. Tveit and Espelid" studied the detection of recurrent caries adjacent to radiopaque composite and amalgam restorations and showed that only 53% of the lesions adjacent to the former and 44% adjacent to the latter were detected. Goshima and Goshima 16 found that the best restorative material for the detection of adjacent defects had a radiographic density similar to that of enamel. This study was therefore undertaken to assess the diagnostic accuracy of the DSR in detection of artificial recurrent caries lesions in posterior teeth, and to compare it with the diagnostic accuracy of the conventional radiographic (CR) film method. Also, the effect of the radiopacity of the composite restoration on the detection of adjacent recurrent caries was evaluated.
0.3 mm and 0.6 mm wide, the second 0.4 mm deep and 1.0 mm wide, and the largest 0.5 mm deep and 1.6 mm wide. The 28 teeth were arranged in groups of four to produce seven phantoms. The defects were enlarged in random order, exposing radiographs at successive stages until a total of 10 radiographs of each phantom was obtained. Each radiograph contained one or two intact surfaces. The final study material was 70 radiographs (7 phantoms x 10 radiographs), each with six approximal surfaces visible resulting in a total of 420 surfaces for evaluation. Out of the 420 surfaces, 40% were intact and 60% had a defect.
Restorative materials The effect of the radiopacity of the restorative material on the detection of recurrent caries with the subtraction method was studied by using two posterior composite filling materials with different radiation attenuation coefficients. Valux (3M, St Paul, MN, USA) was selected as a radiolucent material with an attenuation coefficient considerably lower than that of the enamel. A radiopaque material with an attenuation coefficient considerably higher than that of the enamel but less than that of amalgam was represented by FulFil (L.D. Caulk, Milford, DE, USA).
Imaging equipment
Materials and methods Twenty-eight freshly extracted, unrestored posterior teeth, 14 molars and 14 premolars, were obtained from the Department of Pathology at the University of Texas Health Science Center at San Antonio. Cavities were prepared similar to posterior MOD inlay restorations but without undercuts so that the restorations could be easily removed and replaced after the preparation of the simulated marginal recurrent caries. Each tooth was restored using two different composite restorative materials and reference (mask) radiographs of each tooth were made with both materials. The radiographic images were digitized and stored for future use.
The device shown in Figure 2 was fabricated so as to obtain reproducible serial periapical radiographs. A Gendex 1000 dental X-ray unit (Gendex Corporation, Milwaukee, WI, USA) was used for the film exposures with a 40-cm focus-film distance and 2-cm object-film distance. The exposures were made with a 0.5 s exposure time at 65 kVp and 15 rnA with 2.5 mm Al total filtration. The radiographs were recorded on Ultraspeed No 2 periapical film (Eastman Kodak, Rochester, NY, USA), and processed in 4.5 min with a temperature-controlled automatic processor at 20°C (Dent-X 9000, Dent-X Co., Elmsford, NY, USA).
Digitization of the radiographs Fabrication of artificial caries Caries was simulated by drilling wedge-shaped defects at the margin of the mesial and distal boxes of the cavity preparation (Figure 1) using a micrometerdriven stage that guided the high-speed drill. Three sizes of defects were prepared. The smallest was
Figure 1 Artificial caries lesions of three sizes used in the study: 0.3 x 0.6 mrn, 0.4 x 1.0 mm and 0.5 x 1.6 mm. a, Coronal view; b, sagittal view
60
Dentomaxillofac. Radiol., 1992, Vol. 21, May
The radiographic images were converted to a RS-170 video signal by a video camera (MTI Series 68; DageMTI, Inc., Michigan City, IN, USA). An eight-bit
Figure 2 A radiographic fixture to hold the block of teeth, X-ray tube-head and radiographic film in reproducible relationship, A 2.5-cm thick acrylic block simulating the soft tissue is positioned in front of the teeth
Digital subtraction radiography in caries detection: P. V. Nummikoski et al.
a
b
Figure 3 Examples of the baseline radiographs that were used in the image subtraction procedure as reference films; reference film of the block of teeth restored with FulFil (a) and Valux (b) restorative materials
analogue-to-digital conversion of the video signal was done with an image grabber board FG-lOO-AT (Imaging Technologies Inc., Woburn, MA, USA) interfaced with an IBM-AT compatible microcomputer JE 3020 (Jameco Electronics, Belmont, CA, USA) resulting in a range of 256 grey levels. The images were displayed on 48-cm Mitsubishi Color Display (Mitsubishi Model no. HA 3905, Signa Design Inc., Fremont, CA, USA) with 640(H) x 480(V) pixel screen. The image manipulations and subtractions were done with a software package!" developed on site with real-time subtraction capability. The reference images (Figure 3) were digitized and stored in an optical disk. Subsequent images were aligned with the appropriate reference image by monitoring a real-time subtraction image on the CRT display while shifting and rotating the subsequent image on a stage controlled by micrometers until the image cancellation was subjectively maximized. Once aligned, the subsequent image was digitized. A digital gamma correction algorithml'' was applied on the digitized images to standardize the radiographic density and contrast between the mask and subsequent images. Once the gamma correction was performed, the reference image was subtracted from the subsequent image and a pixel value of 128 was added to the subtraction image to restore its visibility (Figure 4).
The consistency of the laboratory, radiographic and digitizing procedures was assessed with a method introduced by Riittiman et al. 8 by measuring the standard deviation of the grey level distribution in the subtracted images. This evaluation value takes into account all the variables in a project that can result in imperfect subtraction of the mask and subsequent images: inaccuracies during the assembly and disassembly of the set of teeth, inaccuracies in the exposure of the radiographs, inaccuracies in development and digitization of the films, and in the subtraction procedure. For this evaluation, two sets of four teeth were assembled and disassembled five times. Radiographs were made of each assembled set and the images were digitized and subtracted from the reference image resulting in eight subtraction images. The standard deviation of the grey level distribution was measured for each subtraction image for quantitation of the non-ideal subtraction of the anatomical structures in the theoretically identical images.
Image evaluation and statistical treatment of the data A panel of seven trained dentists (two diplomates of the American Board of Oral and Maxillofacial Radiology and five trainees in oral and maxillofacial radiology) viewed the 70 conventional and 70 CRT subtraction images for the presence of caries on a fivepoint scale: 1 = caries definitely present 2 = caries probably present 3 = cannot tell 4 = caries probably not present 5 = caries definitely not present.
Figure 4 Example of digital subtraction image on the monitor screen with several caries-simulating defects in the interproximal restorative margins
The observers were informed that 40% of surfaces were intact and 60% were carious. They were shown a subset of images twice to measure the intra-observer and interobserver agreement. The evaluation data was stored in the VAX 8600 mainframe computer (Digital Equipment Corporation, Merrimack, NH, USA) for the statistical analyses. ROC analysis!" of the data was performed using a computer program, modified by Dove et al,z°, to determine the sensitivity and specificity values for both Dentomaxillofac. Radiol., 1992, Vol. 21, May
61
Digital subtraction radiography in caries detection: P. V. Nummikoski et al. Valux
FulFil
0.6
a
l-Specificity
Valux
FulFil
0.8
1.0
Figure 5 ROC curves of the conventional radiographic method - - 0 - - and the digital image subtraction method - - . - with two restorative materials: a. FulFil restorative material; b, Valux restorative material
diagnostic imaging modalities with both restorative materials. The trapezoidal method of curve fitting was used to obtain the representative ROC areas. The effects of the imaging modalities and the restorative materials were evaluated with two-factor analysis of vanance.
0.4
0.8
a
l-Specificity
0.6
0.8
1.0
l-Specificity
l-Specificity
Figure 6 The ROC curves of different size lesions with FulFil (a) and Valux (b) restorative materials when a conventional radiographic method was used: - - . - - , all lesion sizes included; - - D - , medium and small lesion sizes included; --A--, only small lesion sizes included
Valux
FulFil 1.
0c-:~§~~~'7r
0.9
0.8
>- 0.7 '~ 0.6 :~ 0.5 V1
~
Results
(f)
The average standard deviation of the grey level distribution of the testing set of eight subtracted images without lesions was 2.5 ± 0.7 pixel values. Figure 5 shows the ROC curves for the diagnostic accuracy of the CR and the DSR when using the two restorative materials. The ROC areas for individual observers and the average areas are shown in Table I. The two-factor ANOVA of the ROC areas (Table II), and the subsequent individual r-tests (Table III), show that the tested imaging methods, restoration materials
Table I Individual ROC areas for both imaging methods and restorative materials
Imaging method Conventional
Subtraction
Observer
FulFil
Valux
FulFil
Valux
\
0.932 0.878 0.827 0.9\ \ 0.899 0.86\ 0.936
0.597 0.606 0.626 0.533 0.564 0.48\ 0.630
0.958 0.954 0.908 0.955 0.968 0.920 0.975
0.957 0.914 0.934 0.91 \ 0.9\5 0.924 0.960
0.892 ±0.040
0.577 ±0.054
0.948 ±0.025
0.93\ ±0.020
2 3 4 5 6 7 Mean s.d.
Table II Two-factor analysis of variance of the ROC areas with both imaging methods and restorative materials
Source Imaging method Restorative material Interaction Error
62
df
\
24
Sum of squares
Mean square F-test
0.\94
0.194
140.4
Received 8 April 1991 and in final form 26 November 1991 The diagnostic accuracy of digital subtraction radiography in detection of artificial recurrent caries lesions was assessed in this project. The use of digital subtraction radiography has been shown to markedly increase the accuracy of the detection of destruction in the periodontal bone, but the method has not been evaluated in secondary caries detection. Defects of three different sizes, simulating recurrent caries, were sequential1y prepared in the interproximal cavity preparation margins of 28 teeth. Two composite restorative materials with different radiographic densities were used as posterior restorations, and a radiograph was obtained of each defect size and restorative material. The radiographs were digitized and subtracted from the reference images, and the conventional radiographs and the subtraction images were evaluated by seven observers. The data were analysed with ROC statistics. Subtraction radiography was found to be superior to conventional radiography in recurrent caries detection, mainly by reducing the false-positive diagnoses. The radiopacity of the restorative material had a significant effect on accuracy with conventional but not with subtraction radiography. / Keywords: Radiography; subtraction technic; dental caries; dental restoration, permanent Dentomaxillofac. Radial., 1992, Vol. 21, 59-64, May
Introduction Recurrent caries in previously restored teeth have been shown to account for from 42% to 88% of the need for restorative dentistry":'. Progression into the tooth is rapid and Leinfelder" reported that once detected clinically or radiographically, caries under a composite restoration can breach the pulpal chamber in 6-8 months. For these reasons, the early detection of caries associated with a previously restored tooth is an important component of the examination of dental patients. Present caries detection methods
In dental practice today, diagnostic methods for recurrent caries include clinical probing, transil1umination and intra-oral radiographs. The latter are considered necessary especially with posterior teeth, because of their wide faciolingual dimensions and the interruption of light flow at the margins of the restorations which make both the transillumination and probing unreliable. However, the effectiveness of standard radiographic methods in detecting recurrent caries has been questioned. Gratt et al. 5 and Douglass et al. 6 concluded that the sensitivity of dental radiographs in detecting early dental caries is only 60% or less. The actual depth of a carious lesion is consistently underestimated because of the relative insensitivity of the film-based imaging system and its inability to reveal early tissue demineralization among the structured and unstructured noise of the image. Silverstone? found that when a lesion appearing to be limited to the outer enamel is first detected on a bitewing radiograph,
the demineralization has already penetrated several hundred micrometers into the dentine. In view of the treatment ramifications of advanced recurrent caries, the use of sensitive diagnostic methods is critical to enhance early detection of the disease. Digital subtraction radiography method
Digital image subtraction (DSR) has been shown to be an accurate diagnostic tool in alveolar bone evaluation'"!", With this method, once the radiographic images are aligned and digitized into the computer memory, the image of the mask film is electronically subtracted from that of the subsequent film so that the resultant image represents the difference in density between the two films. The technique results in enhancement of the subtle changes that are difficult, if not impossible, to visualize from the original radiographs amongst the high structural (anatomical) noise. Ortman et al. II showed that 5% bone mineral loss can be visualized with the use of a wel1-control1ed DSR method. In comparison, more than 30% bone mineral loss is required before it can be detected visual1y from radiographic film. In addition, the subtraction technique, by altering the appearance of the pathological changes so that they are more conspicuous, results in a higher degree of interobserver agreernent'". The effect of dental restorative materials
The use of composite posterior restorative material is increasing in practice today 13. Several manufacturers Dentomaxillofac. Radiol., 1992, Vol. 21, May
59
Digital subtraction radiography in caries detection: P. V. Nummikoski et al. use radiopaque components in these materials in order to render them more visible on the radiograph, and this has been found to have an effect on the detectability of the adjacent caries. Matteson et al. 14 reported that for simulated recurrent caries in extracted teeth, the disease was detected on periapical films more effectively when a radiopaque composite was present instead of amalgam. On the other hand, the falsepositive rate for recurrent caries was significant, especially for teeth restored with radiolucent composites. Tveit and Espelid" studied the detection of recurrent caries adjacent to radiopaque composite and amalgam restorations and showed that only 53% of the lesions adjacent to the former and 44% adjacent to the latter were detected. Goshima and Goshima 16 found that the best restorative material for the detection of adjacent defects had a radiographic density similar to that of enamel. This study was therefore undertaken to assess the diagnostic accuracy of the DSR in detection of artificial recurrent caries lesions in posterior teeth, and to compare it with the diagnostic accuracy of the conventional radiographic (CR) film method. Also, the effect of the radiopacity of the composite restoration on the detection of adjacent recurrent caries was evaluated.
0.3 mm and 0.6 mm wide, the second 0.4 mm deep and 1.0 mm wide, and the largest 0.5 mm deep and 1.6 mm wide. The 28 teeth were arranged in groups of four to produce seven phantoms. The defects were enlarged in random order, exposing radiographs at successive stages until a total of 10 radiographs of each phantom was obtained. Each radiograph contained one or two intact surfaces. The final study material was 70 radiographs (7 phantoms x 10 radiographs), each with six approximal surfaces visible resulting in a total of 420 surfaces for evaluation. Out of the 420 surfaces, 40% were intact and 60% had a defect.
Restorative materials The effect of the radiopacity of the restorative material on the detection of recurrent caries with the subtraction method was studied by using two posterior composite filling materials with different radiation attenuation coefficients. Valux (3M, St Paul, MN, USA) was selected as a radiolucent material with an attenuation coefficient considerably lower than that of the enamel. A radiopaque material with an attenuation coefficient considerably higher than that of the enamel but less than that of amalgam was represented by FulFil (L.D. Caulk, Milford, DE, USA).
Imaging equipment
Materials and methods Twenty-eight freshly extracted, unrestored posterior teeth, 14 molars and 14 premolars, were obtained from the Department of Pathology at the University of Texas Health Science Center at San Antonio. Cavities were prepared similar to posterior MOD inlay restorations but without undercuts so that the restorations could be easily removed and replaced after the preparation of the simulated marginal recurrent caries. Each tooth was restored using two different composite restorative materials and reference (mask) radiographs of each tooth were made with both materials. The radiographic images were digitized and stored for future use.
The device shown in Figure 2 was fabricated so as to obtain reproducible serial periapical radiographs. A Gendex 1000 dental X-ray unit (Gendex Corporation, Milwaukee, WI, USA) was used for the film exposures with a 40-cm focus-film distance and 2-cm object-film distance. The exposures were made with a 0.5 s exposure time at 65 kVp and 15 rnA with 2.5 mm Al total filtration. The radiographs were recorded on Ultraspeed No 2 periapical film (Eastman Kodak, Rochester, NY, USA), and processed in 4.5 min with a temperature-controlled automatic processor at 20°C (Dent-X 9000, Dent-X Co., Elmsford, NY, USA).
Digitization of the radiographs Fabrication of artificial caries Caries was simulated by drilling wedge-shaped defects at the margin of the mesial and distal boxes of the cavity preparation (Figure 1) using a micrometerdriven stage that guided the high-speed drill. Three sizes of defects were prepared. The smallest was
Figure 1 Artificial caries lesions of three sizes used in the study: 0.3 x 0.6 mrn, 0.4 x 1.0 mm and 0.5 x 1.6 mm. a, Coronal view; b, sagittal view
60
Dentomaxillofac. Radiol., 1992, Vol. 21, May
The radiographic images were converted to a RS-170 video signal by a video camera (MTI Series 68; DageMTI, Inc., Michigan City, IN, USA). An eight-bit
Figure 2 A radiographic fixture to hold the block of teeth, X-ray tube-head and radiographic film in reproducible relationship, A 2.5-cm thick acrylic block simulating the soft tissue is positioned in front of the teeth
Digital subtraction radiography in caries detection: P. V. Nummikoski et al.
a
b
Figure 3 Examples of the baseline radiographs that were used in the image subtraction procedure as reference films; reference film of the block of teeth restored with FulFil (a) and Valux (b) restorative materials
analogue-to-digital conversion of the video signal was done with an image grabber board FG-lOO-AT (Imaging Technologies Inc., Woburn, MA, USA) interfaced with an IBM-AT compatible microcomputer JE 3020 (Jameco Electronics, Belmont, CA, USA) resulting in a range of 256 grey levels. The images were displayed on 48-cm Mitsubishi Color Display (Mitsubishi Model no. HA 3905, Signa Design Inc., Fremont, CA, USA) with 640(H) x 480(V) pixel screen. The image manipulations and subtractions were done with a software package!" developed on site with real-time subtraction capability. The reference images (Figure 3) were digitized and stored in an optical disk. Subsequent images were aligned with the appropriate reference image by monitoring a real-time subtraction image on the CRT display while shifting and rotating the subsequent image on a stage controlled by micrometers until the image cancellation was subjectively maximized. Once aligned, the subsequent image was digitized. A digital gamma correction algorithml'' was applied on the digitized images to standardize the radiographic density and contrast between the mask and subsequent images. Once the gamma correction was performed, the reference image was subtracted from the subsequent image and a pixel value of 128 was added to the subtraction image to restore its visibility (Figure 4).
The consistency of the laboratory, radiographic and digitizing procedures was assessed with a method introduced by Riittiman et al. 8 by measuring the standard deviation of the grey level distribution in the subtracted images. This evaluation value takes into account all the variables in a project that can result in imperfect subtraction of the mask and subsequent images: inaccuracies during the assembly and disassembly of the set of teeth, inaccuracies in the exposure of the radiographs, inaccuracies in development and digitization of the films, and in the subtraction procedure. For this evaluation, two sets of four teeth were assembled and disassembled five times. Radiographs were made of each assembled set and the images were digitized and subtracted from the reference image resulting in eight subtraction images. The standard deviation of the grey level distribution was measured for each subtraction image for quantitation of the non-ideal subtraction of the anatomical structures in the theoretically identical images.
Image evaluation and statistical treatment of the data A panel of seven trained dentists (two diplomates of the American Board of Oral and Maxillofacial Radiology and five trainees in oral and maxillofacial radiology) viewed the 70 conventional and 70 CRT subtraction images for the presence of caries on a fivepoint scale: 1 = caries definitely present 2 = caries probably present 3 = cannot tell 4 = caries probably not present 5 = caries definitely not present.
Figure 4 Example of digital subtraction image on the monitor screen with several caries-simulating defects in the interproximal restorative margins
The observers were informed that 40% of surfaces were intact and 60% were carious. They were shown a subset of images twice to measure the intra-observer and interobserver agreement. The evaluation data was stored in the VAX 8600 mainframe computer (Digital Equipment Corporation, Merrimack, NH, USA) for the statistical analyses. ROC analysis!" of the data was performed using a computer program, modified by Dove et al,z°, to determine the sensitivity and specificity values for both Dentomaxillofac. Radiol., 1992, Vol. 21, May
61
Digital subtraction radiography in caries detection: P. V. Nummikoski et al. Valux
FulFil
0.6
a
l-Specificity
Valux
FulFil
0.8
1.0
Figure 5 ROC curves of the conventional radiographic method - - 0 - - and the digital image subtraction method - - . - with two restorative materials: a. FulFil restorative material; b, Valux restorative material
diagnostic imaging modalities with both restorative materials. The trapezoidal method of curve fitting was used to obtain the representative ROC areas. The effects of the imaging modalities and the restorative materials were evaluated with two-factor analysis of vanance.
0.4
0.8
a
l-Specificity
0.6
0.8
1.0
l-Specificity
l-Specificity
Figure 6 The ROC curves of different size lesions with FulFil (a) and Valux (b) restorative materials when a conventional radiographic method was used: - - . - - , all lesion sizes included; - - D - , medium and small lesion sizes included; --A--, only small lesion sizes included
Valux
FulFil 1.
0c-:~§~~~'7r
0.9
0.8
>- 0.7 '~ 0.6 :~ 0.5 V1
~
Results
(f)
The average standard deviation of the grey level distribution of the testing set of eight subtracted images without lesions was 2.5 ± 0.7 pixel values. Figure 5 shows the ROC curves for the diagnostic accuracy of the CR and the DSR when using the two restorative materials. The ROC areas for individual observers and the average areas are shown in Table I. The two-factor ANOVA of the ROC areas (Table II), and the subsequent individual r-tests (Table III), show that the tested imaging methods, restoration materials
Table I Individual ROC areas for both imaging methods and restorative materials
Imaging method Conventional
Subtraction
Observer
FulFil
Valux
FulFil
Valux
\
0.932 0.878 0.827 0.9\ \ 0.899 0.86\ 0.936
0.597 0.606 0.626 0.533 0.564 0.48\ 0.630
0.958 0.954 0.908 0.955 0.968 0.920 0.975
0.957 0.914 0.934 0.91 \ 0.9\5 0.924 0.960
0.892 ±0.040
0.577 ±0.054
0.948 ±0.025
0.93\ ±0.020
2 3 4 5 6 7 Mean s.d.
Table II Two-factor analysis of variance of the ROC areas with both imaging methods and restorative materials
Source Imaging method Restorative material Interaction Error
62
df
\
24
Sum of squares
Mean square F-test
0.\94
0.194
140.4
