Detection of apical openness using artificial intelligence methods

Introduction: Teeth play a key role in basic functions like chewing and speaking. Evaluating tooth morphology is essential in diagnosis and treatment planning. Apical root opening is a significant radiological finding indicating incomplete root development, which can complicate endodontic and orthodontic planning, particularly in young individuals. Factors like caries, trauma, or lesions can halt root development, leaving the apex open and posing clinical challenges. Panoramic radiographs are widely used in dentistry due to their low radiation dose and wide field of view. This study developed an artificial intelligence (AI)-based approach to classify apical root openings in panoramic radiographs. Materials and Methods: A total of 902 single-rooted permanent tooth images were manually extracted from 512 panoramic radiographs obtained from the archive of Fırat University. The images were categorized into three classes: closed apex, anatomically open, and pathologically open. Image processing was performed using ImageJ, and classification was conducted with the Vision Transformer (ViT Base Patch32) model. Model performance was evaluated using accuracy, precision, recall, and F1-score. Results: The ViT model achieved 88% in accuracy, precision, recall, and F1-score. When compared with classifications made by dental specialty students, the model showed more consistent results, especially compared to less experienced users. Conclusion: The ViT model demonstrated high accuracy in detecting apical root openings on panoramic radiographs and has the potential to serve as a reliable tool in clinical decision support systems.