Automatized Detection and Categorization of Fissure Sealants from Intraoral Digital Photographs Using Artificial Intelligence

The aim of the present study was to investigate the diagnostic performance of a trained convolutional neural network (CNN) for detecting and categorizing fissure sealants from intraoral photographs using the expert standard as reference. An image set consisting of 2352 digital photographs from permanent posterior teeth (461 unsealed tooth surfaces/1891 sealed surfaces) was divided into a training set (n = 1881/364/1517) and a test set (n = 471/97/374). All the images were scored according to the following categories: unsealed molar, intact, sufficient and insufficient sealant. Expert diagnoses served as the reference standard for cyclic training and repeated evaluation of the CNN (ResNeXt-101-32x8d), which was trained by using image augmentation and transfer learning. A statistical analysis was performed, including the calculation of contingency tables and areas under the receiver operating characteristic curve (AUC). The results showed that the CNN accurately detected sealants in 98.7% of all the test images, corresponding to an AUC of 0.996. The diagnostic accuracy and AUC were 89.6% and 0.951, respectively, for intact sealant; 83.2% and 0.888, respectively, for sufficient sealant; 92.4 and 0.942, respectively, for insufficient sealant. On the basis of the documented results, it was concluded that good agreement with the reference standard could be achieved for automatized sealant detection by using artificial intelligence methods. Nevertheless, further research is necessary to improve the model performance.

Caries Detection on Intraoral Images Using Artificial Intelligence

Although visual examination (VE) is the preferred method for caries detection, the analysis of intraoral digital photographs in machine-readable form can be considered equivalent to VE. While photographic images are rarely used in clinical practice for diagnostic purposes, they are the fundamental requirement for automated image analysis when using artificial intelligence (AI) methods. Considering that AI has not been used for automatic caries detection on intraoral images so far, this diagnostic study aimed to develop a deep learning approach with convolutional neural networks (CNNs) for caries detection and categorization (test method) and to compare the diagnostic performance with respect to expert standards. The study material consisted of 2,417 anonymized photographs from permanent teeth with 1,317 occlusal and 1,100 smooth surfaces. All the images were evaluated into the following categories: caries free, noncavitated caries lesion, or caries-related cavitation. Each expert diagnosis served as a reference standard for cyclic training and repeated evaluation of the AI methods. The CNN was trained using image augmentation and transfer learning. Before training, the entire image set was divided into a training and test set. Validation was conducted by selecting 25%, 50%, 75%, and 100% of the available images from the training set. The statistical analysis included calculations of the sensitivity (SE), specificity (SP), and area under the receiver operating characteristic (ROC) curve (AUC). The CNN was able to correctly detect caries in 92.5% of cases when all test images were considered (SE, 89.6; SP, 94.3; AUC, 0.964). If the threshold of caries-related cavitation was chosen, 93.3% of all tooth surfaces were correctly classified (SE, 95.7; SP, 81.5; AUC, 0.955). It can be concluded that it was possible to achieve more than 90% agreement in caries detection using the AI method with standardized, single-tooth photographs. Nevertheless, the current approach needs further improvement.

Identifying and Avoiding Risk of Bias in Caries Diagnostic Studies

Caries diagnostic studies differ with respect to their design, included patients/tooth samples, use of diagnostic and reference methods, calibration, blinding and data reporting. Such heterogeneity makes comparisons between studies difficult and could represent a substantial risk of bias (RoB) when it is not identified. Therefore, the present report aims to describe the development and background of a RoB assessment tool for caries diagnostic studies. The expert group developed and agreed to use a RoB assessment tool during three workshops. Here, existing instruments (e.g., QUADAS 2 and the Joanna Briggs Institute Reviewers' Manual) influenced the hierarchy and phrasing of the signalling questions that were adapted to the specific dental purpose. The tailored RoB assessment tool that was created consists of 16 signalling questions that are organized in four domains. This tool considers the selection/spectrum bias (1), the bias of the index (2) and reference tests (3), and the bias of the study flow and data analysis (4) and can be downloaded from the journal website. This paper explores possible sources of heterogeneity and bias in caries diagnostic studies and summarizes the relevant methodological aspects.

Examiner calibration in caries detection for populations and settings where in vivo calibration is not practical

AIM

to compare two methods of in vitro examiner calibration in populations and settings where clinical (in vivo) calibration is not practical.

METHODS

Study design was cross-sectional and fully-crossed. The units of analysis were 880 tooth surfaces, from ten children ages 3 to 4 years. The study had three data components: (1) Examiner training and calibration using the International Caries Detection and Assessment System (ICDAS) e-Learning programme (2) In vivo community-based visual examination and (3) Intra-oral digital photographs of the same tooth surfaces from the in vivo visual examination. Kappa and weighted kappa scores were used to study reliability estimates. Systematic differences in caries assessments were determined using the Stuart Maxwell test. Data were analysed using STATA 13.1 and SAS 9.2.

RESULTS

Weighted kappa scores for the in vivo component ranged from 0.50 to 0.66 and from 0.64-0.74, for inter- and intraexaminer reliability, respectively. Caries lesions detected in vivo were also detected on photographs, albeit with more false positives when using photographs. For example, of 46 tooth surfaces assessed as being sound in the in vivo examination, 22 (48%) of these were assessed as having caries when photographs were used as the diagnostic method.

CONCLUSIONS

From this research it appears that good quality photographs alone may be used for training and calibration among challenging populations or settings without adversely affecting data quality.

White spot lesions after orthodontic treatment assessed by clinical photographs and by quantitative light-induced fluorescence imaging; a retrospective study

OBJECTIVE

White spot lesions (WSL) are an important side-effect of orthodontic multi-bracket (MB) treatment. Standardized monitoring of such WSL may help in caries management.

MATERIALS AND METHODS

In this retrospective study the discriminatory power of caries assessment on routine digital oral photographs was compared to quantitative light-induced fluorescence (QLF) imaging in monitoring WSL development after debonding of orthodontic appliances. Oral and QLF photographs captured directly after debond (T1) and 1 year thereafter (T2) of 51 subjects, treated with full MB orthodontic appliances were used. Oral photographs were assessed by use of The International Caries Detection and Assessment System (ICDAS) at both time points independently and by side-by-side comparison to assess visual transition (VT). QLF photographs were categorized based on integrated fluorescence loss at T1 and T2.

RESULTS

At T1 433 and 384 lesions on 918 buccal surfaces were detected using ICDAS and QLF, respectively. For both methods these numbers were reduced at T2. Changes within ICDAS scores were recorded by VT and showed mainly lesion improvement within ICDAS score 2.

CONCLUSION

The oral and QLF photographs both showed regression of WSL after debonding of MB orthodontic appliances. The VT evaluation was found to have higher discriminatory power in comparison to ICDAS.