Deep Learning–Based Detection of Missing Teeth on Panoramic Radiographs for Prosthodontic Treatment Planning

18TH INTERNATIONAL İSTANBUL SCIENTIFIC RESEARCH CONGRESS, İstanbul, Türkiye, 28 - 30 Aralık 2025, cilt.1, ss.859-865, (Tam Metin Bildiri)

Proper tooth segmentation and accuracy in detection of missing teeth are critical in proper planning of

prosthodontic treatment as they have a direct impact on the quality of the diagnosis and clinical decision

making. This paper presents an artificial intelligence-based segmentation model that was created to

detect the presence and absence of teeth in panoramic dental radiograph image automatically. A deep

learning network based on U-Net was trained with the help of the already existing masked panoramic

images, which were then converted and re-labeled using the FDI tooth numbering system to allow

accurate localization of the missing tooth areas. The proposed model demonstrated strong segmentation

performance on the test set, achieving a low loss value 0.078, a mean Intersection over Union of 0.79,

and a mean Dice score of 0.88. Detection of existing teeth showed high precision and F1-score, while

missing tooth detection presented high recall, indicating effective identification of missing teeth with

minimal false negatives. These results indicate that the new method can aid planning a prosthodontic

treatment process through the decrease in diagnostic workload and the enhancement of the consistency

of radiographic examination. In future studies, artificial intelligence–based methods are planned to focus

on the classification of tooth loss patterns and to contribute to the presentation of the most appropriate

prosthetic design in line with these classifications.