Introduction
Panoramic radiography is a crucial diagnostic tool that provides a comprehensive view of the facial bones and teeth in both adults and children. It has advantages such as low radiation dose and ease of use, though it also has disadvantages like low resolution and lack of fine details. In children, panoramic radiography is used for general dental evaluation, occlusion assessment, and detection of intraosseous pathologies.
Objective
The aim of this study is to evaluate the use of deep learning approaches for the automated detection and numbering of deciduous teeth in children's panoramic radiographs.
Methods and Materials
The study utilized an artificial intelligence algorithm (CranioCatch) based on Faster R-CNN Inception v2 (COCO) models. The algorithm was trained and tested on 421 panoramic images obtained from children aged 5-7 years. Images were captured using the Planmeca Promax 2D Panoramic system. The performance of the system was assessed using metrics such as accuracy, sensitivity, and F1 score.
Results
The AI system successfully detected and numbered the deciduous teeth of children with high sensitivity and precision. In the evaluation of 46 images from the test group, sensitivity was 0.9804, precision was 0.9571, and the F1 score was 0.9686.
Conclusion
Deep learning-based AI models are a promising tool for the automated charting of panoramic dental radiographs in children. This system can save time for clinicians and plays a valuable role in forensic identification.
