What question did this study set out to answer?

This pilot study aims to evaluate various convolutional neural network architectures for classifying dental age thresholds using orthopantomograms.

May 16, 2026Open Access

Assessing Convolutional Neural Networks Architectures for Orthopantomography Image Analysis: Judicial Dental Age Threshold Classification

Key Points

This pilot study aims to evaluate various convolutional neural network architectures for classifying dental age thresholds using orthopantomograms.
Selected orthopantomograms from individuals aged 0 to 25 years, totaling 3,551 images.
Used an 80-20 split for training and validation sets while maintaining class distributions.
Applied data augmentation, preprocessing with contrast enhancement, and trained CNN architectures in PyTorch with binary cross entropy loss.
EfficientNetV2 M achieved 93.5% accuracy at the 10-year threshold.
DenseNet 169 reached 94.5% accuracy at the 12-year threshold.
Xception showed varying accuracy across other thresholds: 92.4% (14 years), 94.7% (16 years), 90.6% (18 years), and 89.0% (21 years).

Abstract

Forensic dental age assessment is required when documentary evidence is absent or unreliable, and judicial decisions often depend on whether an individual falls below or above legally defined age thresholds. This pilot study evaluates ImageNet pretrained convolutional neural network architectures for binary age threshold classification. Orthopantomograms from individuals aged 0 to 25 years (1,887 males and 1,664 females) was selected using predefined inclusion criteria and labelled by chronological age. For each legal threshold (10, 12, 14, 16, 18 and 21 years), images were stratified into training and validation sets using an 80 20 split while preserving class distributions. Preprocessing included contrast enhancement with contrast limited adaptive histogram equalization, automated cropping, resizing to model specific input dimensions and ImageNet normalisation. Data augmentation was applied only to training images. Seven convolutional neural network architectures (ResNet 50, ResNet 152, VGG19, DenseNet 121, DenseNet 169, EfficientNetV2 M and Xception) were fine tuned in PyTorch using binary cross entropy loss with early stopping. Hyperparameters were optimised through Bayesian search targeting macro F1 score. Model interpretability was assessed using Grad CAM heatmaps reviewed by dental experts. EfficientNetV2 M showed the best performance for the 10-year threshold (accuracy: 0.935), DenseNet 169 for the 12-year threshold (accuracy: 0.945) and Xception for the remaining thresholds (accuracy: 14-year: 0.924; 16-year: 0.947; 18-year: 0.906; 21-year: 0.890). Errors clustered near age cut offs and increased at higher thresholds. Grad CAM highlighted posterior dento alveolar regions associated with root development. These results support convolutional neural network based orthopantomogram analysis as a judicial decision support approach and guided model selection for large scale evaluation. • CNNs enable binary age-threshold decision support from OPGs. • Different CNN backbones perform best at different legal age thresholds. • EfficientNetV2-M, DenseNet-169 and Xception show highest internal validity. • Errors cluster near legal cut-offs, underscoring medico-legal relevance. • Grad-CAM highlights dento-alveolar regions used in forensic age assessment.

Assessing Convolutional Neural Networks Architectures for Orthopantomography Image Analysis: Judicial Dental Age Threshold Classification

Key Points

Abstract

Cite This Study