Application of Deep Learning on Large-Scale Dental Imaging for Automated Diagnosis and Treatment Planning in Orthodontics and General Dentistry

Abstract

Dental imaging plays an important role in automated diagnosis, treatment planning, and clinical decision-making in orthodontics and general dentistry. Early and accurate identification of dental abnormalities from radiographic images can assist dentists in detecting pathological conditions, evaluating anatomical structures, and supporting effective treatment strategies. This study aims to apply deep learning on large-scale dental imaging for automated diagnosis and treatment planning support in orthodontics and general dentistry. For this research, a large dental radiographic image dataset was used, containing training, validation, and testing image samples with multiple dental conditions and anatomical categories. The proposed framework applied deep learning-based feature extraction, K-means pseudo-label clustering, and supervised classification models to classify dental image patterns into four cluster-based diagnostic groups. Model performance was evaluated using accuracy, precision, recall, F1-score, confusion matrix, ROC curve, precision–recall curve, training-validation accuracy and loss curves, and LIME-based explainable artificial intelligence analysis. The results show that the proposed model achieved strong classification performance with an overall accuracy of 99.62% on 1,580 test images. The confusion matrix showed that most predictions were correctly classified across all four clusters, while ROC and precision–recall curves demonstrated excellent discriminative capability. The LIME visualization further indicated that the model focused on clinically meaningful dental and maxillofacial regions during prediction. The findings suggest that deep learning-based dental image analysis can support automated diagnosis, improve diagnostic efficiency, and contribute to preliminary treatment planning in orthodontics and general dentistry.