Dental age estimation is considered to be an accurate method for identifying the age of an unknown person. This study aims to apply a deep-learning algorithm to assess mandibular third molar development following Demirjian’s method. Demirjian’s classification of tooth development consists of eight stages (Stages A to H). This study has focussed on only the final five stages (Stages D to H), which are found in the 15–23 age range. Eight-hundred mandibular third molar images per stage were cropped manually. In each stage, a total of 720 images were assigned to a training group and the remaining images were assigned to a test group. Automatic developmental stage assessment was performed using the GoogLeNet, a deep-learning algorithm. The overall accuracy of this method was 82.50%, whereas the accuracy in each developmental stage ranged from 87.50% to 97.50%. All of the misinterpreted results of this automatic method revealed only one-stage deviation from the developmental stages assessed by the observer. This developed method revealed a high degree of accuracy. The method can be used in clinical practice as an assistive tool to help clinicians to easily assess the development of mandibular third molars for dental age estimation, reduce time and decrease subjectivity.

Objective: This study aimed to develop a fully automated and explainable framework for dental age estimation from panoramic radiographs in young individuals. Methods: A dataset of 1639 radiographs from individuals aged 8 to 23 years was used. The proposed 2-stage pipeline involved: (1) oriented tooth detection using the YOLO11-OBB model and (2) age estimation using deep learning-based regression models with an attention-weighting module to aggregate predictions from individual teeth. Auxiliary features, including the presence of deciduous teeth and sex, were also evaluated for their impact on model performance. Results: For the first stage, the tooth detection model achieved an F1 score of 0.981, demonstrating accurate tooth localization and identification. In the later stage, the best-performing model, DenseNet-121 with the deciduous teeth feature, achieved a mean absolute error (MAE) of 1.05 ± 0.95 years. Compared to traditional methods, the proposed framework significantly reduced the MAE. Conclusions: This study developed an explainable, high-performing deep learning framework that offers a promising solution for real-world age estimation in the forensic domain.

Background/Objectives: Accurate age estimation is a critical task in forensic and legal contexts, where reliable and objective assessment is essential for age-dependent decision-making. This study aims to explore the application of convolutional neural network-based regression models for age estimation from panoramic radiographs, incorporating conformal prediction to generate prediction intervals representing uncertainty. The estimated ages were also evaluated for their application in legal age classification at thresholds of 10, 15, 18, and 20 years. Methods: A total of 1734 panoramic radiographs of individuals aged 8 to 23 years were used in this study. We employed six models, including ResNet-50, EfficientNet-B0, DenseNet-121, MobileNetV3-Small, VGG-16, and ResNeXt-50, to perform regression and provide point estimates of chronological age. Conformal prediction using a split conformal framework with absolute residual scores was implemented to construct statistically valid prediction intervals with finite-sample coverage guarantees. Results: The results revealed that ResNeXt-50 achieved the lowest mean absolute error for point estimates (0.92 years) and produced among the narrowest conformal margins, ranging from 1.71 to 3.72 years across non-coverage probabilities from 0.20 to 0.01. In legal age classification, the ResNeXt-50 model achieved its highest classification accuracy using point estimates, whereas the application of the conformal lower bound improved specificity and positive predictive value. This shift facilitates conservative decision-making, aligning with the forensic minimum-age principle. Conclusions: The proposed framework offers a practical solution for automated age estimation by quantifying uncertainty through prediction intervals, allowing for more informed assessments aligned with forensic protocols for legal age determination.

This study aimed to determine the agreement of skeletal maturation (SM) assessed by hand-wrist radiographs and cervical vertebral maturation (CVM) assessed by lateral cephalometric radiographs. It also explored the feasibility of using CVM as an alternative to hand-wrist radiographs among Thai children and adolescents. The samples comprised 95 males and 67 females from Thailand, aged 7–17 years. A total of 162 hand-wrist radiographs and lateral cephalometric radiographs were included. Two experienced oral radiologists independently assessed SM (using Fishman’s method) and CVM (using Baccetti’s method); a third referee observer resolved any discrepancies. The correlation and agreement between the methods were evaluated. The results found a substantial agreement (weighted kappa = 0.594) between SM and CVM. The interobserver discrepancy in categorizing SM stages was mostly found at stage 5, while the most common discrepancy for CVM was at stage 6. A significant positive correlation (Spearman’s correlation coefficient = 0.677, P < 0.001) indicated a moderate to strong association between SM and CVM. The substantial agreement and moderate to strong correlation between CVM and SM highlighted the feasibility of using CVM to assess skeletal growth. However, to ensure reliability, it was necessary to precisely evaluate the appearance of the capping of the epiphyseal plate for SM and the distinction between a rectangular and square cervical spine shape at C3 and C4 for CVM differentiation. Keywords: Cervical spine, Hand, Radiographic evaluation, Orthodontics