Radiographic Caries Research Articles

Enhanced Diagnostic Accuracy for Dental Caries and Anomalies in Panoramic Radiographs Using a Custom Deep Learning Model

Enhanced Diagnostic Accuracy for Dental Caries and Anomalies in Panoramic Radiographs Using a Custom Deep Learning Model

A Deep Learning Approach to Automatic Tooth Caries Segmentation in Panoramic Radiographs of Children in Primary Dentition, Mixed Dentition, and Permanent Dentition.

The purpose of this study was to evaluate the effectiveness of dental caries segmentation on the panoramic radiographs taken from children in primary dentition, mixed dentition, and permanent dentition with Artificial Intelligence (AI) models developed using the deep learning method. This study used 6075 panoramic radiographs taken from children aged between 4 and 14 to develop the AI model. The radiographs included in the study were divided into three groups: primary dentition (n: 1857), mixed dentition (n: 1406), and permanent dentition (n: 2812). The U-Net model implemented with PyTorch library was used for the segmentation of caries lesions. A confusion matrix was used to evaluate model performance. In the primary dentition group, the sensitivity, precision, and F1 scores calculated using the confusion matrix were found to be 0.8525, 0.9128, and 0.8816, respectively. In the mixed dentition group, the sensitivity, precision, and F1 scores calculated using the confusion matrix were found to be 0.7377, 0.9192, and 0.8185, respectively. In the permanent dentition group, the sensitivity, precision, and F1 scores calculated using the confusion matrix were found to be 0.8271, 0.9125, and 0.8677, respectively. In the total group including primary, mixed, and permanent dentition, the sensitivity, precision, and F1 scores calculated using the confusion matrix were 0.8269, 0.9123, and 0.8675, respectively. Deep learning-based AI models are promising tools for the detection and diagnosis of caries in panoramic radiographs taken from children with different dentition.

Artificial intelligence for caries detection: a novel diagnostic tool using deep learning algorithms.

The aim of this study was to develop an assessment tool for automatic detection of dental caries in periapical radiographs using convolutional neural network (CNN) architecture. A novel diagnostic model named ResNet + SAM was established using numerous periapical radiographs (4278 images) annotated by medical experts to automatically detect dental caries. The performance of the model was compared to the traditional CNNs (VGG19, ResNet-50), and the dentists. The Gradient-weighted Class Activation Mapping (Grad-CAM) technique shows the region of interest in the image for the CNNs. ResNet + SAM demonstrated significantly improved performance compared to the modified ResNet-50 model, with an average F1 score of 0.886 (95% CI 0.855-0.918), accuracy of 0.885 (95% CI 0.862-0.901) and AUC of 0.954 (95% CI 0.924-0.980). The comparison between the performance of the model and the dentists revealed that the model achieved higher accuracy than that of the junior dentists. With the assist of the tool, the dentists achieved superior metrics with a mean F1 score of 0.827 and the interobserver agreement for dental caries is enhanced from 0.592/0.610 to 0.706/0.723. According to the results obtained from the experiments, the automatic assessment tool using the ResNet + SAM model shows remarkable performance and has excellent possibilities in identifying dental caries. The use of the assessment tool in clinical practice can be of great benefit as a clinical decision-making support in dentistry and reduce the workload of dentists.

Assessment of the Object Detection Ability of Interproximal Caries on Primary Teeth in Periapical Radiographs Using Deep Learning Algorithms

The purpose of this study was to evaluate the performance of a model using You Only Look Once (YOLO) for object detection of proximal caries in periapical radiographs of children. A total of 2016 periapical radiographs in primary dentition were selected from the M6 database as a learning material group, of which 1143 were labeled as proximal caries by an experienced dentist using an annotation tool. After converting the annotations into a training dataset, YOLO was trained on the dataset using a single convolutional neural network (CNN) model. Accuracy, recall, specificity, precision, negative predictive value (NPV), F1-score, Precision-Recall curve, and AP (area under curve) were calculated for evaluation of the object detection model’s performance in the 187 test datasets. The results showed that the CNN-based object detection model performed well in detecting proximal caries, with a diagnostic accuracy of 0.95, a recall of 0.94, a specificity of 0.97, a precision of 0.82, a NPV of 0.96, and an F1-score of 0.81. The AP was 0.83. This model could be a valuable tool for dentists in detecting carious lesions in periapical radiographs.

Artificial intelligence (AI) in restorative dentistry: Performance of AI models designed for detection of interproximal carious lesions on primary and permanent dentition.

The objective of this study was to evaluate the effectiveness of deep learning methods in detecting dental caries from radiographic images. A total of 771 bitewing radiographs were divided into two groups: adult (n = 554) and pediatric (n = 217). Two distinct semantic segmentation models were constructed for each group. They were manually labeled by general dentists for semantic segmentation. The inter-examiner reliability of the two examiners was also measured. Finally, the models were trained using transfer learning methodology along with computer science advanced tools, such as ensemble U-Nets with ResNet50, ResNext101, and Vgg19 as the encoders, which were all pretrained on ImageNet weights using a training dataset. Intersection over union (IoU) score was used to evaluate the outcomes of the deep learning model. For the adult dataset, the IoU averaged 98%, 23%, 19%, and 51% for zero, primary, moderate, and advanced carious lesions, respectively. For pediatric bitewings, the IoU averaged 97%, 8%, 17%, and 25% for zero, primary, moderate, and advanced caries, respectively. Advanced caries was more accurately detected than primary caries on adults and pediatric bitewings P < 0.05. The proposed deep learning models can accurately detect advanced caries in permanent or primary bitewing radiographs. Misclassification mostly occurs between primary and moderate caries. Although the model performed well in correctly classifying the lesions, it can misclassify one as the other or does not accurately capture the depth of the lesion at this early stage.

Artificial intelligence based on Convolutional Neural Network for detecting dental caries on bitewing and periapical radiographs

Objectives: This narrative review is written to describe the accuracy of caries detection and find out the clinical implications and future prospects of using Convolutional Neural Network (CNN) to determine radio-diagnosis of dental caries in bitewing and periapical radiographs.&#x0D; Review: The databases used for literature searching in this narrative review were PubMed, Google Scholar, and Science Direct. The inclusion criteria were original article, case report, and textbook written in English and Bahasa Indonesia, published within 2011-2021. The exclusion criteria were articles that the full text could not be accessed, research article that did not provide the methods used, and duplication articles. In this narrative review, a total of 33 literatures consisting of 30 articles and three textbooks reviewed, including four original articles on CNN for caries detection.&#x0D; Conclusion: Results of the review reveal that GoogLeNet produces the best detection compared to Fully Convolutional Network (FCN) and U-Net for caries detection in bitewing and periapical radiographs. Nonetheless, the positive predictive value (PPV), recall, negative predictive value (NPV), specificity, F1-score, and accuracy values in these architectures indicate good performance. The differences of each CNN’s performances to detect caries are determined by the number of trained datasets, the architecture’s layers, and the complexity of the CNN architectures. The conclusion of this review is CNN can be used as an alternative to detect caries, increasing the diagnostic accuracy and time efficiency as well as preventing errors due to dentist fatigue. Yet the CNN is not able to substitute the expertise of a radiologist. Therefore, it is need to be revalidated by the radiologist to avoid diagnostic errors.

A deep learning approach to automatic detection and profiling of dental caries in bitewing and periapical radiographs

<h3>Objective</h3> Dental caries is the most prevalent noncommunicable disease among humans, but routine misdiagnosis remains an underappreciated challenge. This study aimed to answer the challenge with the latest advances in artificial intelligence (AI), machine/deep learning, and computer vision. This study was conducted by an interdisciplinary team of academic clinicians and a dental AI startup. <h3>Study Design</h3> We curated and methodically annotated an imbalanced database of bitewing and periapical radiographs using a caries ontology derived from the ICDAS framework. Using this dataset, we leveraged machine learning algorithms to train predictive models to perform segmentation and classification of dental caries on novel radiographs. Given the trained model, we designed a comparison study to measure the qualitative and quantitative performance of the trained neural networks vs a set of experts. The objective of the study was to assess the expected effectiveness of model performance, and validate model behavior in direct comparison with humans on the same tasks. The study consisted of 3 dentists interacting with a web-based platform for collecting annotations of de-identified radiographs. Fifteen radiographs were used for the pilot study. <h3>Results</h3> In the human-human comparisons, there was 4% strong disagreement, 48% minor disagreement, and 48% agreement. In the majority of cases (69%), the neural network model was either perfectly correct or close with some nuance. In no case was the model completely incorrect (0%); it either missed one or more caries or produced a false positive (the remaining 31%). In general, it matched or exceeded the expert evaluation of radiographs. <h3>Conclusion</h3> We contend that these results show how neural network-based machine learning models hold considerable promise in providing assistance to dentists. We intend to pursue larger-scale human-machine comparison studies in subsequent efforts. <b>Statement of Ethical Review</b> Ethical review was sought and study was exempted from ethical review

Teaching Radiographic Caries Detection and Treatment Planning: A Seminar Using an Audience Response System

This study presents a seminar model for teaching radiographic caries detection and treatment planning at the Faculty of Dentistry, University of Oslo. The seminar is based partly on an audience response system (ARS) and uses patient cases to focus on caries risk assessment and treatment planning. This paper describes the seminar design, implementation, learning outcomes, and observational study of variability in caries registrations and students’ attitudes to use of ARS. Dental and dental hygiene students participate in two seminar modules. Module 1 aims to develop and increase individual student skills in radiographic caries lesion detection, scoring, and differential diagnosis. Students perform trial registrations on bitewings using an ARS with anonymous live polling, and scorings are discussed in plenum. Students then perform individual registrations on 12 bitewing pairs. Using digital scoring, students detect and grade caries lesions on all approximal and occlusal tooth surfaces. After the session, students use the ARS to repeat scorings on selected tooth surfaces, and results are again discussed in plenum. Module 2 involves group exercises on 4 patient cases that are later presented with plenary discussions. In total, 1,624 caries registrations performed by 150 students attending the seminar between 2016 and 2018 were assessed for variability between students. As expected, variations in caries registrations were observed between students, mostly related to restored surfaces or tooth surfaces that were otherwise difficult to register. In 2022, 63 dental and dental hygiene students attending the seminar answered a questionnaire about use of ARS. The responses were scored using a five-point Likert scale. Overall, no significant difference in satisfaction with the ARS-based module was observed between dental and dental hygiene students (χ² test, p > 0.05). The majority of the students were positive toward the use of ARS (94%), but some disagreed on the role of ARS in usefulness for understanding the seminar content (3.2%), and in increasing their confidence in radiographic registration of caries (3.2%). The ARS-based module provides a positive learning environment that ensures student anonymity, interactivity, and engagement, and combined with the other seminar module gives students basic skills in caries detection and treatment planning.

Caries Detection from Dental Images using Novel Maximum Directional Pattern (MDP) and Deep Learning

Various machine learning technologies and artificial intelligence techniques were applied on different applications of dentistry. Caries detection in orthodontics is a very much needed process. Computer-aided diagnosis (CAD) method is used to detect caries in dental radiographs. The feature extraction and classification are involved in the process of caries detection in dental images. In the 2D images the geometric feature extraction methods are applied and the features are extracted and then applied to machine learning algorithms for classification. Different feature extraction techniques can also be combined and then the fused features can be used for classification. Different classifiers support vector machine (SVM), deep learning, decision tree classifier (DT), Naïve Bayes (NB) classifier, k-nearest neighbor classifier (KNN) and random forest (RF) classifier can be used for the classification process. The proposed MDP extracts both intensity and edge information and creates the feature vector that increases the classification accuracy during caries detection.

Detection of Proximal Caries Lesions with Deep Learning Algorithm

This study aimed to evaluate the effectiveness of deep convolutional neural networks (CNNs) for diagnosis of interproximal caries in pediatric intraoral radiographs. A total of 500 intraoral radiographic images of first and second primary molars were used for the study. A CNN model (Resnet 50) was applied for the detection of proximal caries. The diagnostic accuracy, sensitivity, specificity, receiver operating characteristic (ROC) curve, and area under ROC curve (AUC) were calculated on the test dataset. The diagnostic accuracy was 0.84, sensitivity was 0.74, and specificity was 0.94. The trained CNN algorithm achieved AUC of 0.86. The diagnostic CNN model for pediatric intraoral radiographs showed good performance with high accuracy. Deep learning can assist dentists in diagnosis of proximal caries lesions in pediatric intraoral radiographs.

Patients' Perspectives on Artificial Intelligence in Dentistry: A Controlled Study.

Background: As artificial intelligence (AI) becomes increasingly important in modern dentistry, we aimed to assess patients’ perspectives on AI in dentistry specifically for radiographic caries detection and the impact of AI-based diagnosis on patients’ trust. Methods: Validated questionnaires with Likert-scale batteries (1: “strongly disagree” to 5: “strongly agree”) were used to query participants’ experiences with dental radiographs and their knowledge/attitudes towards AI as well as to assess how AI-based communication of a diagnosis impacted their trust, belief, and understanding. Analyses of variance and ordinal logistic regression (OLR) were used (p < 0.05). Results: Patients were convinced that “AI is useful” (mean Likert ± standard deviation 4.2 ± 0.8) and did not fear AI in general (2.2 ± 1.0) nor in dentistry (1.6 ± 0.8). Age, education, and employment status were significantly associated with patients’ attitudes towards AI for dental diagnostics. When shown a radiograph with a caries lesion highlighted by an arrow, patients recognized the lesion significantly less often than when using AI-generated coloured overlays highlighting the lesion (p < 0.0005). AI-based communication did not significantly affect patients’ trust in dentists’ diagnosis (p = 0.44; OLR). Conclusions: Patients showed a positive attitude towards AI in dentistry. AI-supported diagnostics may assist communicating radiographic findings by increasing patients’ ability to recognize caries lesions on dental radiographs.

Training in radiographic caries detection and staging using an interactive tool.

The objective of the study was to compare a dental student's practical ability to detect and stage radiographic caries per International Caries Detection and Assessment System (ICDAS), following a traditional lecture and a lecture containing an interactive session using an audience response system (ARS). Associations between the order of instructions and student performance were also evaluated. Eighty-three dental students were randomly assigned to groups A and B. On the first day, group A received a traditional lecture and group B received content using the ARS. All students then took an electronic quiz (T1) identifying and staging caries on radiographs per ICDAS. For the second day, group A received the content using the ARS system and group B received a traditional lecture. All students subsequently took a second electronic quiz (T2). Two survey questions about the learning experience were also included. Wilcoxon rank-sum analysis of scores from consenting students (81) showed no difference between the quiz 1scores of two groups (p=.61). Whilst not statistically significant (p=.07), the group that had the ARS initially scored marginally higher on quiz 2. Survey results showed that most participants preferred either the ARS alone (49.38%) or a combination of the ARS and a traditional lecture (40.74%). A majority of them (80%) found the ARS helpful. When training students in practical skills of detection and staging radiographic presence of dental caries per ICDAS, hands-on learning tools, such as an ARS, complement traditional lectures.

Radiographic Proximal Caries Detection and Multiple Intelligences

Introduction: The accuracy of radiographic caries detection is a multifactorial diagnostic task. One the important factors are the Inter-observer variability. One of the suggested procedures to estimate the variations is the Multiple Intelligence (MI) theory of Howard Gardner.&#x0D; Aim: The current research aims to study the relation among radiographic proximal caries detection accuracy and MI for dental undergraduate and new graduates (trainees)&#x0D; Patients and Methods: Twenty dental students who have just terminated caries detection training and 19 interns were employed. The sample composed of seventeen men and twenty two women. Caries detection precision was estimated as the region under Receiver Operating Characteristic (ROC) curve (Az) scores, and MI scores were estimated by sing a modified, bilingual MIS self-recorded questionnaire. MS Excel software was applied to calculate means, percentages and correlation.&#x0D; Results: There was little and mild positive correlation among Az and Spatial/Visual (0.29) and Intrapersonal (0.31) domains of MI respectively. The current correlation was mild for students (0.38) and (0.40), and it was lowered for interns (0.28) and (0.26) respectively.&#x0D; Conclusions: There is positive weak relationship between Az and the domains of Spatial/Visual and Intrapersonal of MI. This relationship might be diminished by education and training.

Radiographic modalities for diagnosis of caries in a historical perspective: from film to machine-intelligence supported systems.

Radiographic imaging for the diagnosis of caries lesions has been a supplement to clinical examination for approximately a century. Various methods, and particularly X-ray receptors, have been developed over the years, and computer systems have focused on aiding the dentist in the detection of lesions and in estimating lesion depth. The present historical review has sampled accuracy ex vivo studies and clinical studies on radiographic caries diagnosis that have compared two or more receptors for capturing the image. The epochs of film radiography, xeroradiography, digital intraoral radiography, panoramic radiography and other extraoral methods, TACT analysis, cone-beam CT and artificial intelligence systems aiding in decision-making are reviewed. The author of this review (43 years in academia) has been involved in caries research and contributed to the literature in all the mentioned epochs.

Restorative thresholds for carious lesions in primary molars: French dentist's decisions.

Questionnaire surveys have been undertaken worldwide to investigate practices and knowledge related to carious lesion management, particularly in adults. The primary objective of this cross-sectional survey was to investigate restorative thresholds (RTs) used for carious lesions in primary molars by dentists practising paediatric dentistry in France. Dentists were surveyed by a specifically developed questionnaire based on clinical and radiographic caries classifications. The secondary objective was to explore restorative management strategies in primary molars. A structured questionnaire assessing RTs and management strategies for occlusal and approximal carious lesions of primary molars was anonymously and electronically administered via SurveyMonkey® to dentists who were members of the Société Française d'Odontologie Pédiatrique. Descriptive analyses, Chi-square test, McNemar test, and logistic regression analyses considering dependent RT variables for occlusal and approximal carious lesions were performed. Among 250 dentists surveyed, 201 responded (response rate 80.4%). Overall, 43% (n = 87) and 75% (n = 151) of respondents would place their RTs in enamel for occlusal and approximal lesions, respectively. Dentists with an exclusive practice of paediatric dentistry more frequently would choose a RT in dentine for approximal lesions than did other dentists (p = 0.010). A preparation technique including sound dental tissues was less frequent for occlusal than approximal lesions (n = 31; 15% vs n = 60; 30%). Overall, 75% (n = 151) of respondents used the same restorative material for occlusal and approximal lesions. In general, dentists practising paediatric dentistry in France overtreated lesions on primary molars, which contradicts minimal intervention recommendations. RTs are too often indicated for enamel-confined carious lesions.

Analysis of Image Enhancement Techniques for Dental Caries Detection Using Texture Analysis and Support Vector Machine

ABSTRACT Different enhancement techniques are used to diagnose dental caries in dental X-ray images. The diagnostic performances of the enhancement methods for caries detection in digital radiographs are evaluated. The proposed method consists of pre-processing of dental X-ray images using Gaussian low pass filter in frequency domain, extraction of statistical features from enhanced image using Support Vector Machine (SVM) classifier. 105 images of normal and dental caries derived using digital radiography are used to evaluate the performance of the SVM classifier with 10-fold cross validation. The images are annotated by a dentist. The quantitative analysis is done after evaluating the performance parameters of SVM classifier. The findings show that the proposed method gives TP rate = 0.982, FP rate = 0.018, ROC area = 0.982, and PRC area = 0.973. By using 2-way ANOVA, the results are tested at significant level of 5%, show that the interaction of enhancement methods with dental images on performance parameter values are significant. The results suggest that the proposed framework is a promising approach for the automatic detection of dental caries in dental radiographs, can also be used for other dental applications. The performance of the system can be further improved by high quality and high quantity dataset and suitable segmentation technique.

Dental caries diagnosis in digital radiographs using back-propagation neural network

An algorithm for diagnostic system with neural network is developed for diagnosis of dental caries in digital radiographs. The diagnostic performance of the designed system is evaluated. The diagnostic system comprises of Laplacian filtering, window based adaptive threshold, morphological operations, statistical feature extraction and back-propagation neural network. The back propagation neural network used to classify a tooth surface as normal or having dental caries. The 105 images derived from intra-oral digital radiography, are used to train an artificial neural network with 10-fold cross validation. The caries in these dental radiographs are annotated by a dentist. The performance of the diagnostic algorithm is evaluated and compared with baseline methods. The system gives an accuracy of 97.1%, false positive (FP) rate of 2.8%, receiver operating characteristic (ROC) area of 0.987 and precision recall curve (PRC) area of 0.987 with learning rate of 0.4, momentum of 0.2 and 500 iterations with single hidden layer with 9 nodes. This study suggests that dental caries can be predicted more accurately with back-propagation neural network. There is a need for improving the system for classification of caries depth. More improved algorithms and high quantity and high quality datasets may give still better tooth decay detection in clinical dental practice.

Radiographic Caries Detection Behavior in a Sample of Dentists and Dental Specialists

Radiographic caries detection depends not only on correctly acquired radiographs but also on the way they are displayed and interpreted. Objective: This research aims to measure behavioral factors important for caries detection among dentists. Materials & Methods: A sample of 206 dentists was surveyed using a self-reported questionnaire after a focus group and a pilot study. Questions asked were related to the type of image used, the way the image is displayed, the use of magnification, and the use of vision correction aids if needed. Responses were analyzed using descriptive statistics. Results: The use of the following was BW (>95%), viewing box (68%), masking extra-light (<15%), dimming room’s light (<15%), magnification (<13%), and vision correction aid (85% from those who need it). The most frequently reported reasons were “availability” and “lack of knowledge”. Conclusions: There is a significant undesirable behavioral practice during radiographic caries detection among the studied sample. The main reasons seem to be “convenience” and lack of knowledge.

Correlation of Visual and Radiographic Caries Diagnosis using the ICDAS in Permanent Molars of a Group of Egyptian Children

ICDAS (International Caries Detection and Assessment System) is an evidence-based detection and assessment system classifying stages of the caries process on the basis of histological extent and activity. The ICDAS measures the visual surface changes which corresponds to the potential histological depth of carious lesions. In addition, it proposes the ICDAS radiographic scoring system, which categorizes lesions according to the depth of caries demineralization into enamel and dentin represented by a radiolucency on the radiographs. In this study, the aim was to assess the relationship between clinical and radiographic caries diagnoses among 12 to 14-year-old Egyptians using the ICDAS scoring system which differentiates between cavitated and non-cavitated caries lesions.

Dental Caries Diagnosis in X-ray Images Using KNN Classifier

Objectives: Early diagnosis of dental caries helps in maintaining good oral health. The current study focuses on diagnosis of dental caries in dental radiographs through machine learning. Methods: Dental radiographic images in bmp format are considered for study. The images are trained, validated and tested with 10-fold cross validation. Diagnosis method involves Laplacian filter, adaptive thresholding, morphological transformation, Grey Level Co-occurrence Matrix (GLCM) based texture analysis and K Nearest Neighbors (KNN) Classifier. The diagnostic performance measures accuracy, False Positive Ratio (FPR), precision, recall, Mathews Correlation Coefficient (MCC), Receiver Operating Characteristic (ROC) area are calculated for detection and diagnosis of dental caries. Findings: Proposed method is giving better performance of 98.5% accuracy, 98.5% precision, 4.7% False Positive Rate (FPR) and 0.953 Receiver Operating Characteristic (ROC) curve area with 10-fold cross validation. The validity of the results tested using two-way ANOVA, at significant level of 5%, shows that the interaction of proposed method on performance parameter measures is significant. Applications/Improvement: The study highlighted the potential utility of machine learning for detection of dental caries in automated computer assisted diagnosis system. The proposed method provided good performance in detecting caries in dental radiographs. The results suggest that the proposed framework is a promising approach for the automatic detection of dental caries in dental radiographs. The performance of the system can be further improved by high quality and quantity dataset. Keywords: Computer Assisted Diagnosis, Dental Caries, Dental Radiography, Machine Learning