Development of an Automated Dental Caries Detection System using Federated Learning in X-Ray Images with Data Security

A fusion of federated learning approach and YOLO object detection algorithm is developed and implemented to perform an early detection of dental carries using X-ray digital images. The main objective of the proposed work is to develop a highly secured automated dental carries detection system that enables collaborative training of artificial intelligence models by utilizing the decentralized datasets which overcome and solve the limitations of the existing conventional centralized approaches in healthcare sector. Because of the high speed and precision nature of the YOLO model, it is tailored to detect dental caries in complex intraoral and radiographic images with promising sensitivity. In accordance with a federated learning approach, various dental establishments can train the dental carries detection models available locally and the trained model parameters can be assigned to a centrally located server module. During the training and distribution processes, the patient data is kept with a high degree of security though it is a decentralized approach as per the regulatory directions of General Data Protection Regulation (GDPR) under medical data protection category. To implement the federated learning, the Flower framework is utilized which ensures efficient communication and aggregation with all the connected client modules and a centrally located server module. The locally trained models by the federated learning are averaged to arrive the global model by combining diverse insights to improve the accuracy and strength of the model. The proposed methodology targets and achieves data heterogeneity along with scalability which makes the easy practical implementations for various real-time problems. The combination of YOLO algorithm and the federated learning approach characterizes a significant initiative towards privacy-protective, efficient, and scalable dental carries detection among dental healthcare organizations globally.