Performance of methods for detecting occlusal caries lesions : ICDAS X radiological image

Objective: We investigated the precision and accuracy of methods to detect caries lesions on the occlusal surface in vitro using ICDAS and radiological image. Methodology: Human third molars (n=14) were placed in an acrylic resin base and maintained wet during the study. The occlusal surfaces were visually inspected by three examiners using the International Caries Detection and Assessment System (ICDAS) method. The condition of each tooth was registered by images obtained with Digital Radiography (DR), Microcomputed Tomography (μ-CT) and Histologic Specimen (HS). For each tooth and method utilized in the study, an image was selected based on the largest extension of caries found, wherein the three examiners attributed a score to the lesion in accordance with the visual description of each method. The Kappa index, Fisher’s exact test and the Spearman ́s correlation coefficient were used for evaluating reliability and accuracy, with a significance level of 5%. Results: Considerable interobserver reliability values were found for ICDAS (k = 0.701), almost perfect for μ-CT (k = 0.855) and HS (k = 0.920), and reasonable for DR (k = 0.221). Significant statistical difference was observed for ICDAS (p < 0.05), and for DR and μ-CT methods (p < 0.01). The correlation was moderate for ICDAS (r = 0.597), high for DR (r = 0.764) and perfect for μ-CT (1.000). Conclusions: the most reliability method for detecting caries lesions on occlusal surfaces in vitro was μ-CT, followed by ICDAS and DR. The most accurate method was μ-CT, followed by DR and ICDAS.


Introduction
The diagnosis of caries lesions on the occlusal surfaces of human teeth is commonly done by conventional methods-visual inspection and tactile exam-by which the dentist detects the presence of the disease with the unaided eye or in association with a blunt-tip dental explorer (Ismail, 2004). When associated with other techniques, the precision of these methods can be improved (Ko et al., 2012). Complementary methods for detecting caries lesions are currently available, such as the ICDAS and radiological technologies.
The efficacy of visual inspection methods, such as the International Caries Detection and Assessment System (ICDAS), for the early diagnosis of caries lesions can be observed in the literature (Pitts & Ekstrand, 2013;Ekstrand et al., 2007;Ekstrand et al., 2018). Developed in 2002, the ICDAS is the most recent effort by researchers to create a tool for diagnosing caries through scientifically based evidences about systems of clinical detection of caries (Banting et al., 2012). It is a clinical method for visual inspection that describes a standard for measuring a caries lesion, detecting it in different degrees of severity, and assessing its extension in a systematic manner-both on an individual and collective level-and classifying it (Braga et al., 2009;Zandoná & Zero, 2006). Moreover, this system improves diagnosis (Ekstrand et al., 2018)., treatment and prognosis, contributing to the ability of dentists to accompany the evolution of a patient's general oral health as well as assisting them with the oral health education of the patient (Brocklehurst et al., 2012).
The histologic test is frequently used as the gold standard for validating the detection of caries in vitro (Mitropoulos et al., 2012;Park et al., 2011;Hintze & Wenzel, 2006); however, it makes long-term study impossible because it destroys the sample. By comparison, radiographic technology has contributed to investigations that evaluate the ICDAS method by measuring the physical alterations that have occurred due to caries on the mineralized portion of the tooth (Zou et al., 2011). Interproximal digital radiography is the complementary method Research, Society and Development, v. 9, n. 10, e1859108490, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i10.8490 5 adopted by dentists for supporting conventional methods of detecting the caries lesion. It presents good sensitivity and high specificity in cavitated lesions (Neuhaus et al., 2011). It is effective for interproximal caries, but ineffective for the detection of occlusal caries lesions located in enamel. This is due to the quantity of healthy tissue that inhibits the X-ray beams (Rechmann et al., 2012). The radiographic measure is an estimate of the continual mineral loss and does not detect certain stages of the lesion or it's activity, except in situations in which it's possible to compare sequential radiographic images of the same caries lesion for a period of time (Young et al., 2015).
The µ-CT has been suggested as an approach to investigating in vitro alterations in mineralized samples of small dimensions. Analyses of dynamic processes-such as caries lesions-are possible because the technique permits non-destructive tridimensional analysis with high spatial resolution image quality, achieving up to 1 micrometer (µm) (Marques & Appoloni, 2015;Davis et al., 2013). It also permits the detection of caries lesions in the initial stage, at which point it is imperceptible with the unaided eye because it is sensitive to the concentration of the sample's minerals (Elfrink et al., 2013). It is a method of inspection that provides detailed transverse images of the internal region of the analyzed minerals (Özkan et al., 2015;Soviero et al., 2012). Such images enable the reconstruction of a succession of Xrays, also called projections, of a given volume of the sample. Each projection is a mapping of the linear attenuation coefficient of the region where the X-ray beam is transmitted (Marques & Appoloni, 2015).
Despite the complementary methods for detecting caries lesions are currently available, studies demonstrating the outcomes of the comparison between ICDAS and radiological imaging in vitro are scarce. Thus, this study was conducted to assess the reliability and accuracy of methods of detecting caries lesions on occlusal surfaces in vitro using ICDAS, digital radiography and µ-CT.

Methodology
This research is characterized by an experimental study, quantitative, exploratory, and descriptive study, in which the search strategy was the international system for the detection and evaluation of caries injuries (ICDAS), carried out in several databases available at the time.
After receiving approval for the project from the Ethics in Research Committee (N º 811.736), 14 human third molars-extracted for therapeutic reasons-and affected by the caries Research, Society and Development, v. 9, n. 10, e1859108490, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i10.8490 6 disease at different levels and without sealant or restorative treatments, were included in the sample.
The teeth were disinfected for 7 days in a solution of chloramine-T (Vetec Química Fina Ltda.; Rio de Janeiro, Rio de Janeiro, Brazil) and 0.5% water at 4°C; cleaned-removing debris and soft tissue-by radicular scraping using periodontal curettes (Mcall; Hu-Fiedy Mfg. Co., LLC; Chicago, USA) and polished with pumice paste in association with a Robson brush attached to a Koncept slow-speed drill (KaVo of Brazil Ind. Com. Ltda., Joinville, Santa Catarina, Brazil). The clean teeth were stored in a glass jar containing ultrapure water to avoid dehydration (Elga; PurelabOption-Q DV25; São Paulo, Brazil), and later placed in an acrylic resin base and randomly coded from I to XIV.

ICDAS
Three examiners were calibrated to the criteria of the ICDAS visual inspection method.
The training occurred in one single step using the electronic system (www.icdas.org), which provides an educational software package-an e-learning program (Topping et al., 2008).
Immediately after the training, the teeth were cleaned by prophylaxis with pumice stone and water for 60 seconds (s), rinsed and dried for 5s with a triple syringe and placed on individual workbenches so that the occlusal surfaces of each tooth could be visually and independently inspected by the examiners. In cases of disagreement over the score number, the criterion of consensus was adopted. The ICDAS control group for evaluation was established this way.
The occlusal surface of each tooth was photographed at a 1:1 proportion for the purpose of registering the effect of caries on the 14 teeth assessed by the ICDAS. A digital camera (EOS Rebel T1i; Canon Inc.; Tokyo; Japan) equipped with a EF100mm f/2.8 Macro USM and a circular MR-14EX flash, all of the Canon brand, was used. The focal distance, the aperature and light sensitivity were, respectively: 100mm; f16; ISO 100. The images were adjusted-in relation to their size and exposure-and processed as a batch with the digital camera's processing, visualization and image editing software from Digital Photo Professional (Canon; Canon Inc., Tokyo; Japan).

DIGITAL RADIOGRAPHY (DR)
The radiographic image of the coronal portion of each tooth was obtained with a size 2 phosphorus plate digitizer-3X4 centimeters (cm), from the VistaScan Plus digital radiography system (Dürr Dental of Brazil; Porto Alegre; Rio Grande do Sul; Brazil), with an X-ray exposure time of 0.2s. The system was connected to the analogic X-ray equipment by Heliodente Plus (Sirona Brasil; São Paulo, São Paulo, Brazil) with a voltage of 70kV and a current Research, Society and Development, v. 9, n. 10, e1859108490, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i10.8490 of 7 milliamperes (mA), which was used to employ a technique similar to intraoralinterproximal radiography. The images were processed with a real resolution of 22 lines pairs per millimeter (Pl/mm) using the digital radiography system's DBSWIN software (Dürr Dental of Brazil; Porto Alegre; Rio Grande do Sul; Brazil).

IMAGE PROCESSING
For each assessed tooth, independently of the method used, images were obtained that were adjusted in relation to the size and exposure, and processed as a batch with the processing, visualization and image editing software Digital Photo Professional (Canon; Canon Inc., Tokyo; Japan) which was provided with the digital camera. In Figure 1, images of each tooth are represented, being selected according to the greatest degree to which caries affected each tooth. This was determined by the deepest extension of color alteration and demineralization of the sample, as compared with healthy tissue.

SCORES
The images obtained from each method for each assessed tooth were analyzed independently by the three examiners of the ICDAS method using a HP Pavilion Ultrabook computer (Hewlett-Packard Brasil Ltda.; São Paulo; São Paulo; Brazil), with a 14-inch screen.
For each assessed image, a score was attributed in accordance with the description in Table 1.  The scores were recodified in accordance with the presence of caries and without caries = 0 (ICDAS, DR, µ-CT, and Histologic Specimen = 0) and with caries = 1 (ICDAS ≥ 1 and ≤ 6 and DR, µ-CT and Histologic Specimen ≥ 1 and ≤ 4) and in accordance with the localization of the caries and without caries = 0 (ICDAS, DR, µ-CT and Histologic Specimen = 0); with caries in enamel = 1 (ICDAS ≥ 1 and ≤ 3 and DR, µ-CT and Histologic Specimen = 1) and with caries in dentin = 2 (ICDAS ≥ 4 and ≤ 6 and DR, µ-CT and Histologic Specimen ≥ 2 and ≤ 4) (Al-Khatrash et al., 2011). In cases of disagreement among the examiners about the number of the score, the criterion of consensus was adopted to determine the score of the tooth for each method. The histologic specimen method was adopted as the gold standard.

STATISTICAL ANALYSIS
The ordinal categorical data were compiled with Microsoft Office Excel 2010 and transferred to the program Statistical Package for Social Sciences (SPSS), version 20.0. In order to express the reliability of the methods tested in this study, the Kappa Index (k) was used. In order to test possible associations and correlations between the methods, Qui-Quadrado Tests, Fisher's exact test and Spearman's coefficient correlation (p) were used. The significance level was set at 5%. Table 2 shows the interobserver reproducibility for ICDAS (k = 0.701), DR (k = 0.221), µ-CT (k = 0.855) and Histologic Specimen (k = 0.920). Significant statistical differences were observed upon comparing the diagnosis of caries between the histologic specimen method (gold standard) and the other methods (Table 3).

Results
Greater difference was found using ICDAS, followed by DR. ICDAS underestimated the diagnosis of seven teeth with carie in dentin and DR underestimated two teeth when compared to the gold standard. The methods of detecting caries lesions on occlusal surfaces correlated with the HS (Table 4). Significant statistical difference was observed for ICDAS (p < 0.05), and for DR and µ-CT methods (p < 0.01). The correlation was moderate for ICDAS (r = 0.597), high for DR (r = 0.764) and perfect for µ-CT (1.000).

Discussion
The ICDAS is among the alternatives to the visual and tactile methods of inspection recommended by the World Health Organization because of its high level of reproducibility and accuracy for assessing primary coronal caries lesions (Ekstrand et al., 2018). This method was studied in this research with the purpose of comparing it to the methods based on X-rays and to the histologic specimen method (gold standard) so as to assess its performance in detecting caries lesions on the occlusal surfaces of third molars. Research, Society andDevelopment, v. 9, n. 10, e1859108490, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i10.8490 13 In this study, the interobserver reproducibility on the ICDAS method was considerable (0.701). The results concurred with those found by Rodrigues et al. (2013) in which sixteen post-graduate students assessed the occlusal surfaces of 67 exfoliated deciduous teeth after only one training on a mannequin.
When correlated to the histologic specimen method and µ-CT, the visual inspection system ICDAS was moderate because a large number of teeth was classified with caries in only the enamel, being that the caries process had already affected dentin in accordance with the gold standard (Table 3). This could have occurred due to the visual limitation of the method, principally for score 3 where the enamel showed microcavitation, although without visualization of the dentin. In this approach, it is difficult to precisely assess the degree of the lesion's damage when no stains have occurred on the assessed surface. In Figure 1 The results show that the methods that are more easily interpreted produced the best outcomes. This was the case with the µ-CT method, which after the statistical analyses showed a similar precision and a perfect correlation with the histologic specimen method. However, this method implies a high dose of radiation and requires extensive time for digitalization, reconstruction and processing of the images, making the cost-benefit relationship prohibitive (Özkan, et al., 2015).
In our study, the tridimensional data obtained with the µ-CT method showed more information about the condition of the occlusal tissue than digital radiography using the phosphorus plate method, which presented images with superimposition of structures on the occlusal surface, thus making diagnosis of caries in enamel more difficult (Krzyżostaniak et al., 2014).
Mineral density also has been considered as a parameter for the determination of demineralization and of remineralization in dental caries. This is because it offers insight into the dynamic alterations associated with the tridimensional spatial distribution of the mineral in caries lesions (Zou et al., 2011). In this study, the µ-CT method provided tridimensional images of the assessed teeth, which made it possible to generate precise assessment of the caries lesion within a tooth without needing to destroy the sample. Thus, the method permits reassessment of a sample when necessary in order to detect episodes of demineralization or remineralization in natural caries lesions (Zou et al., 2011). However, the longitudinal studies with repeated measures of samples with more than one series of demineralization or remineralization episodes Research, Society and Development, v. 9, n. 10, e1859108490, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i10.8490 14 are not practical because of the limited availability of equipment and the operational costs. The variation in and the uncertainty of determining the composition of the enamel, be it healthy or decayed, limits the precision with which mineral concentrations of the enamel can be established based on reference values (Zou et al., 2011). It is difficult, however, to simulate a standard of mineral composition of the enamel and dentin structures since the quantification of minerals varies between individuals (Soviero et al., 2012).
In this study, the histologic section that was adopted as the reference for assessment was the section that was most visually affected by the caries lesion. Having obtained the histologic specimens in intervals of 1 mm, it was difficult to isolate a lesion that most represented the severity of the disease. This was due to the tridimensional nature of caries propagation as determined by the complex anatomy of occlusal surface. A lesion could originate in the same location on the surface of the tooth, but spread obliquely and asymmetrically below the surface of the enamel (Jablonski- Momeni et al., 2008). This makes diagnosis of a lesion's extension even more complex, and at the same time emphasizes the importance of using supplemental exams to diagnose the disease.
It is important to emphasize the difficulty the dentist encounters when detecting initial stage caries lesions, especially when they occur in the enamel of the occlusal surface. The methods commonly adopted for this purpose depend on visual acuity and are limited by subjectivity. Therefore, to associate these methods with the radiographic technology could improve the precision of detecting this type of lesion by means of quantifying the degree of mineral loss in the structure of the enamel, thus guaranteeing that the dentist adopts the appropriate intervention.

Conclusions
In summary, from the methods used during this research, the most reliability method for detecting caries lesions on occlusal surfaces in vitro was µ-CT, followed by ICDAS and DR.
The most accurate method was µ-CT, followed by DR and ICDAS.
It is necessary to carry out clinical research with longitudinal monitoring, to evaluate the treatment of caries lesions according to the criteria of the ICDAS method, since the scientific evidence of the diagnostic method is confirmed by numerous studies.