Accuracy of visual and image-based ICDAS criteria compared with a micro-CT gold standard for caries detection on occlusal surfaces

Improving accuracy of early dental carious lesions detection using deep learning-based automated method

OBJECTIVE

To investigate the effectiveness of a convolutional neural network (CNN) in detecting healthy teeth and early carious lesions on occlusal surfaces and to assess the applicability of this deep learning algorithm as an auxiliary aid.

MATERIALS AND METHODS

A total of 2,481 posterior teeth (2,459 permanent and 22 deciduous teeth) with varying stages of carious lesions were classified according to the International Caries Detection and Assessment System (ICDAS). After clinical evaluation, ICDAS 0 and 2 occlusal surfaces were photographed with a professional digital camera. VGG-19 was chosen as the CNN and the findings were compared with those of a reference examiner to evaluate its detection efficiency. To verify the effectiveness of the CNN as an auxiliary detection aid, three examiners (an undergraduate student (US), a newly graduated dental surgeon (ND), and a specialist in pediatric dentistry (SP) assessed the acquired images (Phase I). In Phase II, the examiners reassessed the same images using the CNN-generated algorithms.

RESULTS

The training dataset consisted of 8,749 images, whereas the test dataset included 140 images. VGG-19 achieved an accuracy of 0.879, positive agreement of 0.827, precision of 0.949, negative agreement 0.800, and an F1-score of 0.887. In Phase I, the accuracy rates for examiners US, ND, and SP were 0.543, 0.771, and 0.807, respectively. In Phase II, the accuracy rates improved to 0.679, 0.886, and 0.857 for the respective examiners. The number of correct answers was significantly higher in Phase II than in Phase I for all examiners (McNemar test;P<0.05).

CONCLUSIONS

VGG-19 demonstrated satisfactory performance in the detection of early carious lesions, as well as an auxiliary detection aid.

CLINICAL RELEVANCE

Automated detection using deep learning algorithms is an important aid in detecting early caries lesions and improves the accuracy of the disease detection, enabling quicker and more reliable clinical decision-making.

Prevention of non-cavitated lesions with fluoride and xylitol varnishes during orthodontic treatment: a randomized clinical trial

OBJECTIVE

Fixed orthodontic appliances impair oral hygiene increasing the risk of non-cavitated lesions (NCLs) and tooth decay. The aim of this study was to compare the outcomes of fluoride and xylitol varnishes in preventing NCLs during comprehensive orthodontic treatment.

MATERIALS AND METHODS

The sample comprised 55 volunteers from 15 to 20 years of age under orthodontic treatment that were randomly divided into three groups: Fluoride Group (FG; n=17), Xylitol Group (XG; n=19), and Placebo Group (PG; n=19). The patients in each group received two applications of the following varnishes: Duraphat^TM (5% NaF), 20% xylitol, and placebo (no F/Xylitol) in the three groups, respectively. The varnishes were applied in the first appointment (T0) and 3 months later (T1). Clinical examinations were carried out at T0 and 6 months after (T2) using the ICDAS index and the QLF system (fluorescence difference). The intergroup comparisons were performed by ANOVA/Tukey's or Kruskal-Wallis/Dunn's tests (P<0.05).

RESULTS

There was no significant intergroup difference regarding ICDAS index changes from T0 to T2. Fluoride varnish produced significantly greater increase in fluorescence of NCLs (mean change of -0.65 +0.78 and -0.56 +0.83, for maxilla and mandible, respectively) in comparison to the other groups. The majority of non-cavitated lesions improved in the fluoride and xylitol varnish groups.

CONCLUSIONS

Fluoride varnish produced significantly greater increase in enamel fluorescence compared to xylitol and placebo varnishes. In short term, both fluoride and xylitol varnishes produced remineralization of NCLs in orthodontic patients.

CLINICAL RELEVANCE

Non-cavitated lesions can be effectively controlled in high-risk orthodontic patients by means of fluoride varnishes.

CLINICAL TRIAL REGISTRATION

ReBEC Identifier: RBR-6mdxfq; Date of Register: March 19th, 2020. Retrospectively Registered.

Diagnosis of Occlusal Caries with Dynamic Slicing of 3D Optical Coherence Tomography Images

Detecting the extent of occlusal caries is a clinically important but challenging task required for treatment decision making. The aim of this study was to assess the diagnostic power of 3D swept-source optical coherence tomography (OCT) for evaluation of occlusal caries in comparison with X-ray radiography. Extracted human molars not exhibiting American Dental Association (ADA) criteria advanced caries were mounted in a silicone block and digital dental radiographs were captured from the buccal side. Subsequently, occlusal surfaces were scanned with a prototype Yoshida Dental OCT. Thirteen examiners evaluated the presence and extent of caries on radiographs and dynamically sliced 3D OCT video images, using a 4 level scale—0: intact; 1: enamel demineralization without cavitation; 2: enamel caries with cavitation; 3: dentin caries with or without cavitation. Sensitivity, specificity and area under operating characteristic curves (Az) were statistically analyzed (α = 0.05). Reliability analysis showed an excellent agreement among the 13 examiners for both methods. The OCT presented a significantly higher sensitivity and Az value for the detection of caries compared to radiographs (p < 0.05). Radiography showed especially low sensitivity for dentin caries (0–2 versus 3). Dynamic slicing of 3D OCT volumes is a powerful adjunct tool to visual inspection to diagnose the dentin occlusal caries in vitro.