CBCT in dental age estimation: A systematic review and meta analysis

Objectives: This study aimed to investigate the reproducibility of dental age estimation methods in cone beam computed tomography (CBCT) and the correlation between dental (DA) and chronological (CA) ages. Methods: The scientific literature was searched in six databases (PubMed, Scopus, LILACS, Web of Science, SciELO, and OATD). Only observational studies were selected. Within each study, the outcomes of interest were (I) the quantified reproducibility of the method (κ statistics and Intraclass correlation coefficient); and (II) the correlation (r) between the dental and chronological ages. A random-effect three-level meta-analysis was conducted alongside moderator analysis based on methods, arch (maxillary/mandibular), population, and number of roots. Results: From 671 studies, 39 fulfilled the inclusion criteria, with one study reporting two different methods. The methods used in the studies were divided into metric (n = 17), volumetric (n = 20), staging (n = 2), and atlas (n = 1). All studies reported high examiner reproducibility. Group 1 (metric and volumetric) provided a high inverse weighted r ([Formula: see text] = −0.71, CI [-0.79,–0.61]), and Group 2 (staging) provided a medium-weighted r ([Formula: see text] = 0.49, CI [0.44, 0.53]). Moderator analysis on Group one did not show statistically significant differences between methods, tooth position, arch, and number of roots. An exception was detected in the analysis based on population (Southeast Asia, [Formula: see text] = −0.89, CI [-0.94,–0.81]). Conclusion: There is high evidence that CBCT methods are reproducible and reliable in dental age estimation. Quantitative metric and volumetric analysis demonstrated better performance in predicting chronological age than staging. Future studies exploring population-specific variability for age estimation with metric and volumetric CBCT analysis may prove beneficial.

Revisiting dynamic range and image enhancement ability of contemporary digital radiographic systems

Objectives: To assess the dynamic range and enhancement ability of radiographs acquired with contemporary digital systems. Methods: Five repeated periapical radiographs of human mandibles with an aluminium step-wedge were acquired using two sensor-based and three phosphor plate-based (PSP plate-based) systems and an X-ray unit at ten exposure times 0.020, 0.032, 0.063, 0.080, 0.100, 0.200, 0.320, 0.400, 0.500, and 0.630 s. All images had their brightness and contrast enhanced by two experienced oral and maxillofacial radiologists in consensus and were exported as both the original and enhanced file formats. Mean grey values were obtained from the aluminium steps and tabulated with their corresponding thicknesses for each exposure time, digital radiographic system, and file format. Images with saturated steps were excluded and the mean grey values from the remaining images were averaged to assess image brightness and the angular coefficient of the linear trendlines was generated from the relationship between mean grey values and their corresponding aluminium thicknesses to assess image contrast. Brightness and contrast values were compared using two-way ANOVA with post-hoc Tukey (α = 0.05). Results: PSP plate-based digital radiographic systems had a broader dynamic range. Longer exposure times produced original images with lower brightness and variable contrast (p < 0.05). Subjective enhancement significantly increased or reduced brightness and/or contrast in some systems (p < 0.05). Conclusions: Contemporary digital radiographic systems present different dynamic ranges and exposure-related brightness and contrast. Image enhancement may be a valuable tool at slightly suboptimal exposure times.

A fully automated method of human identification based on dental panoramic radiographs using a convolutional neural network

Objectives: This study aimed to develop a fully automated human identification method based on a convolutional neural network (CNN) with a large-scale dental panoramic radiograph (DPR) dataset. Methods: In total, 2,760 DPRs from 746 subjects who had 2 to 17 DPRs with various changes in image characteristics due to various dental treatments (tooth extraction, oral surgery, prosthetics, orthodontics, or tooth development) were collected. The test dataset included the latest DPR of each subject (746 images) and the other DPRs (2,014 images) were used for model training. A modified VGG16 model with two fully connected layers was applied for human identification. The proposed model was evaluated with rank-1, –3, and −5 accuracies, running time, and gradient-weighted class activation mapping (Grad-CAM)–applied images. Results: This model had rank-1,–3, and −5 accuracies of 82.84%, 89.14%, and 92.23%, respectively. All rank-1 accuracy values of the proposed model were above 80% regardless of changes in image characteristics. The average running time to train the proposed model was 60.9 sec per epoch, and the prediction time for 746 test DPRs was short (3.2 sec/image). The Grad-CAM technique verified that the model automatically identified humans by focusing on identifiable dental information. Conclusion: The proposed model showed good performance in fully automatic human identification despite differing image characteristics of DPRs acquired from the same patients. Our model is expected to assist in the fast and accurate identification by experts by comparing large amounts of images and proposing identification candidates at high speed.

Automatic detection of anteriorly displaced temporomandibular joint discs on magnetic resonance images using a deep learning algorithm

Objectives: This study aimed to develop models that can automatically detect anterior disc displacement (ADD) of the temporomandibular joint (TMJ) on magnetic resonance images (MRI) before orthodontic treatment to reduce the risk of developing serious complications after treatment. Methods: We used 9009 sagittal MRI of the TMJ as input and constructed three sets of deep learning models to detect ADD automatically. Deep learning models were developed using a convolutional neural network (CNN) based on the ResNet architecture and the “Imagenet” database. Five-fold cross-validation, over sampling, and data augmentation techniques were applied to reduce the risk of overfitting the model. The accuracy and area under the curve (AUC) of the three models were compared. Results: The performance of the maximum open mouth position model was excellent with accuracy and AUC of 0.970 (±0.007) and 0.990 (±0.005), respectively. For closed mouth position models the accuracy and AUC of diagnostic criteria One were 0.863 (±0.008) and 0.922 (±0.009), respectively significantly higher than that of diagnostic criteria two with an 0.839 (±0.013) (p = 0.009) and AUC of 0.885 (±0.018) (p = 0.003). The classification activation heat map also improved our understanding of the models and visually displayed the areas that play a key role in the model recognition process. Conclusion: Our CNN model resulted in high accuracy and AUC in detecting ADD and can therefore potentially be used by clinicians to assess ADD before orthodontic treatment and hence improve treatment outcomes.

Is jaw fractal dimension a reliable biomarker for osteoporosis screening? A systematic review and meta-analysis of diagnostic test accuracy studies

Objective: To summarize the evidence on the feasibility of maxillomandibular imaging exams-related fractal dimension (FD) in screening patients with osteoporosis. Methods: This registered systematic review followed the PRISMA-DTA statement. High sensitivity search strategies were developed for six primary databases and grey literature. QUADAS-2 items evaluated the risk of bias, and the GRADE approach assessed the evidence certainty. Results: From 1,034 records initially identified through database searching, four studies were included (total sample of 747 patients [osteoporosis, 136; control group, 611]). The meta-analysis showed that the overall sensitivity and specificity of the FD were 86.17 and 72.68%, respectively. In general, all studies showed low RoB and applicability concern. The certainty of the evidence was very low to moderate. Conclusions: This systematic review showed that the jaw-related FD presented sensitivity and specificity values higher than 70%, and its sensitivity in osteoporosis screening was a better parameter than specificity.

Exploring digital filters for internal root resorption: how can we improve the diagnosis of small lesions?

Objectives: This study aimed to evaluate the impact of enhancement filters in detecting small simulated internal root resorptions (IRR). Methods: : forty-two extracted human teeth were sectioned, connected, and stored in a dry human jaw and x-rayed with Photostimulable Phosphor Plates (PSPs), composing the control group (CG). In the middle third of the root canals, IRR lesions were simulated using Da Silveira protocol. After, the specimens were x-rayed to create the test group (TG). All images acquired were exported with seven enhancement filters plus the original image. Three examiners used a five-point Likert scale to evaluate the images regarding the presence/absence of IRR. Diagnostic efficacy was assessed from sensitivity and specificity results. Comparison among filters was performed by using receiver operating characteristic (ROC) curve analysis. Results: : moderate values of Kappainterexaminer (0.403–0.620) and high values of Kappa-intraexaminer (0.757–0.915) were observed. The best performance occurred in the CG (p < 0.05). Original images presented the greatest sensitivity and area under the ROC curve (0.595–0.750), while the Endo filter presented the greatest specificity (0.952). Inversion and Pseudo3D images produced the greatest doubt in the diagnosis, significant for CG with the Pseudo-3D filter (p < 0.05). Conclusions: : the Original and ‘Endo’ filters should be chosen as it offers greater diagnostic ability and allows more confidence during the evaluation.