The Application of Artificial Intelligence in the Diagnosis of Pulp Diseases and Periapical Periodontitis

Periapical inflammation and pulp diseases are highly prevalent and potentially destructive oral conditions whose early detection remains challenging with conventional modalities such as periapical radiography and CBCT. Recent advances in artificial intelligence (AI), primarily deep convolutional neural networks (CNNs), have enabled image-driven diagnosis and quantitative assessment of periapical and pulpal pathology. Models trained on expert-annotated two-dimensional periapical and panoramic images and 3D CBCT images demonstrate improved sensitivity for subtle bone-density and morphological changes, enabling automated segmentation, volumetric quantification of radiolucent lesions, and delineation of pulp-cavity morphology, occult cracks and early inflammatory signs. Typical 3D pipelines adopt a coarse-to-fine strategy, combining global localization with local 3D-CNN refinement to produce voxel-level outputs suitable for clinical decision support. Progress is nevertheless constrained by the high cost of expert annotation, heterogeneous multi-center data, and a training plateau that yields diminishing returns from simply enlarging datasets. Methodological remedies include semi- and self-supervised pretraining, active learning, synthetic-data augmentation and federated multi-center training, together with uncertainty quantification and explainable outputs to facilitate clinical adoption. This review aims to systematically evaluate recent advances, methodological challenges and translational pathways for AI-enabled imaging diagnosis of periapical inflammation and pulp disease.