New Publication in Scientific Reports: Deep Learning System for Detecting Permanent First Molars and Predicting Root Development Stages from Panoramic Radiographs

We are proud to announce that our latest research has been published in Scientific Reports (Nature Publishing Group).
The study, titled “End-to-end CNN-based Detection of Permanent First Molars and Prediction of Root Development Stages from Panoramic Radiographs,” introduces a fully automated deep learning framework combining YOLOv7 for molar localization and VGG-19 for root stage classification according to the Cvek system.

Using a dataset of 1,629 pediatric panoramic radiographs, the proposed model achieved 98.4% precision in molar detection and up to 94.9% accuracy in differentiating open and closed apical foramina—critical for treatment planning in pediatric dentistry and regenerative endodontics.
The work highlights the potential of convolutional neural networks to support dentists in assessing root development and improving treatment decisions.

Authors:
S. Türkoglu Kayaçı (University of Health Sciences),
H.O. İlhan & G. Serbes (Yıldız Technical University),
H. Arslan (Istanbul Medeniyet University)

Türkoglu Kayacı, S., İlhan, H.O., Serbes, G., & Arslan, H. (2025). End-to-end CNN-based detection of permanent first molars and prediction of root development stages from panoramic radiographs. Scientific Reports, 15, 38814.
DOI: 10.1038/s41598-025-22707-7