Project information
- Category: Federated Orthodontic Skeletal Diagnosis
- Data: ISBI 2015 and Dicle University Datasets
- Project date: April, 2025
- Project URL: https://doi.org/10.3390/diagnostics15070920
Federated Learning-Based CNN Models for Orthodontic Skeletal Classification and Diagnosis
This study investigates the use of federated convolutional neural networks (CNNs) for orthodontic skeletal classification using cephalometric images from the ISBI and Dicle datasets, comparing federated learning (FL) with centralized (CL) and local learning (LL) methods. The adapted DenseNet121 models, enhanced with attention mechanisms and pooling blocks, showed significant performance improvements, with FL models achieving over 26% accuracy gains and maintaining competitive performance while preserving data privacy. The results highlight the potential of FL to enable secure, collaborative AI development across multiple orthodontic institutions, advancing precision in orthodontic diagnostics.