Dental arch classification using deep learning and federated learning on DenPAR radiographs
6 affiliations
- 1Department of Computing & Technology, Iqra University
- 2Faculty of Computing and Technology, Iqra University
- 3Department of Electric Stations, Grids and Systems of Power Supply
- 4Department Head and Professor of Computer Science and Technology Department, Wor-wic Community College
- 5College of Electrical and Mechanical Engineering, National University of Science and Technology
- 6Department of Electrical and Computer Engineering, King Abdulaziz University
Abstract
Automated analysis of dental radiographs can improve the efficiency and consistency of diagnostic workflows. This study proposes a deep learning and federated learning framework for dental arch classification using panoramic radiographs from the DenPAR dataset. A pre-trained ResNet50 model was employed for feature extraction, while CatBoost, Support Vector Machine (SVM), and Recurrent Neural Network (RNN) classifiers were evaluated for classification. A privacy-preserving Federated Learning approach based on FedBM was also implemented to enable collaborative model training without sharing raw data. Experimental results demonstrated strong classification performance, with CatBoost and SVM achieving 98% accuracy, while the federated model achieved 99% accuracy on the test set. Evaluation using precision, recall, specificity, F1-score, ROC-AUC, and confusion matrices further confirmed the robustness of the proposed framework. The results indicate that deep feature extraction combined with federated learning provides an effective and privacy-preserving solution for automated dental arch classification in panoramic radiographs.
Related
- Federated learning for privacy-preserving opportunistic screening of lumbar-spine DXA-defined osteoporotic-level bone mineral density using chest radiographs: A multi-center independent validation study
- Montage-agnostic federated learning for privacy-preserving Alzheimer’s disease classification from heterogeneous multi-site EEG
- Multistage CBAM-enhanced deep learning for three-class classification of radiographically derived bone mass status labels on knee radiographs
- SBRF-Det: Scale-Boundary Representative Fusion for Multi-Finding Detection in Panoramic Dental Radiographs.
- The accuracy of large language models in assessing tooth detection and identification, dental caries, periapical status, and tooth development from mixed dentition radiographs
BibTeX
@article{qureshi2026dental,
title = {{Dental arch classification using deep learning and federated learning on DenPAR radiographs}},
author = {Abdullah Qureshi and Qurat Ul Ain Aini and Evgeny Solomin and Nadeem Ahmad and Zaki Uddin and Ahmad H. Milyani},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-62329-1},
url = {https://doi.org/10.1038/s41598-026-62329-1},
note = {CC BY},
}
RIS
TY - JOUR
TI - Dental arch classification using deep learning and federated learning on DenPAR radiographs
AU - Abdullah Qureshi
AU - Qurat Ul Ain Aini
AU - Evgeny Solomin
AU - Nadeem Ahmad
AU - Zaki Uddin
AU - Ahmad H. Milyani
PY - 2026
DA - 2026/07/01
JO - Scientific Reports
DO - 10.1038/s41598-026-62329-1
UR - https://doi.org/10.1038/s41598-026-62329-1
AB - Automated analysis of dental radiographs can improve the efficiency and consistency of diagnostic workflows. This study proposes a deep learning and federated learning framework for dental arch classification using panoramic radiographs from the DenPAR dataset. A pre-trained ResNet50 model was employed for feature extraction, while CatBoost, Support Vector Machine (SVM), and Recurrent Neural Network (RNN) classifiers were evaluated for classification. A privacy-preserving Federated Learning approach based on FedBM was also implemented to enable collaborative model training without sharing raw data. Experimental results demonstrated strong classification performance, with CatBoost and SVM achieving 98% accuracy, while the federated model achieved 99% accuracy on the test set. Evaluation using precision, recall, specificity, F1-score, ROC-AUC, and confusion matrices further confirmed the robustness of the proposed framework. The results indicate that deep feature extraction combined with federated learning provides an effective and privacy-preserving solution for automated dental arch classification in panoramic radiographs.
SN - 2045-2322
ER -