MaxViT-T deep learning model estimates dental age from panoramic radiographs with high tolerance accuracy

In a study of 2,761 panoramic radiographs, a hybrid deep learning model (MaxViT-T) estimated dental age with 96.1% accuracy within ±2 years and classified age groups (6–9, 10–13, 14–18y) with 91.13% accuracy (AUC 0.97), though not significantly better than other architectures, a…

• Retrospective, 2,761 panoramic radiographs, 5-fold cross-validation, comparing 11 models (CNNs and Transformers).
• Explainable AI (Grad-CAM/SHAP) revealed attention shifted from deciduous/erupting teeth to third molars with age, focusing on the mandible.
• Performance advantage over other models not statistically significant; no external validation reported.

RadPigeon summaries are original and for information only. They are not clinical advice.