Automatic analysis and diagnosis of dental caries in panoramic images using deep learning

Panoramic dental imaging plays an essential role in detecting oral diseases, especially globally prevalent dental caries. Conventional methods depend on clinicians' subjective assessment, yet existing computer-aided systems are deficient in certain aspects.The study proposes a deep learning model named MengNet, developed on the basis of ResNet 50. It extracts hierarchical residual features and augments geometric data, adapting to complex features in clinical images. Tested on 2,000 panoramic images, MengNet outperformed other top models in precision, recall, and F1 score, achieving a 90.5% test accuracy. The findings demonstrate the efficacy of MengNet in automated caries screening, particularly in scenarios where resources are constrained.