A Hybrid Attention U-Net Model for Enhanced Radiographic Detection of Dental Caries

Abstract

Dental caries detection is crucial for effective treatment and prevention. This study presents a novel Hybrid Attention U-Net (HAU-Net) architecture, specifically designed to improve dental caries detection in panoramic X-ray images. Leveraging advanced attention mechanisms, dilation convolutions, batch normalization, and dropout layers, HAU-Net enhances feature extraction and segmentation accuracy. Evaluations on a dataset of 100 images from Narkhede Dental Clinic in Thane demonstrate significant improvements in detection capabilities, achieving superior accuracy, loss, and segmentation quality. HAU-Net shows significant potential for enhancing dental care through automated diagnostics, early lesion detection, and informed treatment planning.