Panoramic Revolutionizing Orthodontic Analysis with Deep Learning and Panoramic Dental X-rays

Lately, the utilization of deep learning for medical image segmentation has gained significant attention due to its potential to assist in diagnostic accuracy. Here, we present an Attention U-Net model that is intended to highlight important areas in panoramic orthodontic radiographs. The model suppresses unnecessary background information by integrating attention gates, which improves the segmentation of intricate dental features. To enhance the clarity of input photos and masks, data preprocessing techniques such as adaptive thresholding and Contrast Limited Adaptive Histogram Equalization were used. Furthermore, histogram equalization and patch-wise training were used to highlight the radiographs’ fine structural details. SegNet, Simple U-Net, and GCN were among the baselines used to assess this model’s efficacy. Our tests showed that the Attention U-Net continuously performed better than these models, with a 99.53% segmentation accuracy on the Dental Panoramic Caries Dataset.