Retrospective diagnostic evaluation of panoramic radiographs obtained from children aged 6-12 years with artificial intelligence
Project no. 31975 · Status: Kapanmış · 2024
Summary
<p><strong>Objective: </strong>Artificial intelligence (AI) enables computer systems to perform tasks requiring human-like intelligence. This study aims to investigate the use, accuracy, and clinical applicability of deep learning algorithms, specifically convolutional neural networks (CNNs), for detecting supernumerary teeth one of the most common dental anomalies in children. These extra teeth, often seen in panoramic radiographs, can cause complications such as dental crowding, delayed eruption, or cyst formation. Due to their diagnostic importance and high prevalence in pediatric populations, this research focuses on applying AI to detect supernumerary teeth more effectively and efficiently.</p><p><strong>Materials and Methods: </strong>A total of 2,000 panoramic radiographs of children aged 6-12 years, obtained during routine examinations at the Department of Pedodontics, Ege University Faculty of Dentistry, were retrospectively analyzed. The study was conducted in two stages: (1) segmentation of supernumerary teeth and (2) classification of radiographs based on the presence of these teeth. Initially, 140 images were used 71 with and 69 without supernumerary teeth. A deep learning model based on the YOLOv8 architecture was developed and trained using this subset. To validate performance, an additional 20 radiographs were later introduced. Model performance was evaluated using standard metrics including accuracy, recall, and F1 score.</p><p><strong>Results: </strong>The YOLOv8 model achieved 100% accuracy in the segmentation of supernumerary teeth, successfully identifying affected regions in all relevant radiographs, indicating high diagnostic precision.</p><p><strong>Conclusion: </strong>CNN-based deep learning models demonstrate high accuracy and reliability in detecting supernumerary teeth, offering a promising tool to enhance diagnostic workflows and support early, effective pediatric dental treatment.</p>