Classification of Apical Openness Using Vision Transformer: A Comparative Approach with Expert Decisions

Abstract

Teeth play a key role in essential functions such as mastication and speech. Evaluating root morphology is crucial in both diagnosis and treatment planning. Apical openness is a significant radiographic indicator of incomplete root development, which can complicate endodontic and orthodontic procedures, especially in young individuals. Factors such as caries, trauma, or lesions may interrupt root development, resulting in an open apex and clinical challenges. Panoramic radiographs are commonly used in dentistry due to their low radiation dose and wide anatomical coverage. This study aimed to develop an artificial intelligence (AI)-based method to classify apical root openness in panoramic radiographs. A total of 902 single-rooted permanent teeth were manually cropped from 512 panoramic radiographs archived at XXXX. Teeth were categorized into three groups: closed apex, anatomically open, and pathologically open. Image preprocessing was performed using ImageJ, and classification was conducted using a Vision Transformer model (ViT Base Patch32). Model performance was evaluated based on accuracy, precision, recall, and F1-score. The ViT model achieved 88% in accuracy, precision, recall, and F1-score. Compared with manual classifications performed by dental specialty students, the model provided more consistent outcomes, particularly outperforming less experienced participants. The ViT model demonstrated high accuracy in detecting apical root openness on panoramic radiographs and shows promise as a reliable component of clinical decision support systems.