A self attention based deep learning framework for accurate and efficient dental disease detection in OPG radiographs

Oral diseases are increasing now-a-days and there is a high demand for the automatic diagnostic system that helps the clinician to detect these oral diseases with more accuracy and reduced human error. Utilizing the advancement of Deep Learning techniques, this study proposes a novel comparative approach for the diagnosis of teeth diseases using Orthopantomogram (OPG) images and recent transformer based architecture. Particularly, Vision Transformer (ViT) and Swin Transformer are employed for the development of the effective automatic system. Experimental results demonstrated that the Vision Transformer achieved higher performance with a test accuracy of 96%, precision of 95.8%, recall of 96.2%. Swin transformer, with a hierarchical design and shifted window, achieved an accuracy of 95.2% but with efficient inference time and scalable complexity. Based on the findings, it is inferred that ViT outperforms Swin Transformer in diagnosing oral diseases. Thus the proposed work confirms the effectiveness of transformer based architectures in dental imaging tasks, providing a promisable solution for the clinician, for the automatic diagnosis of oral diseases with high accuracy and less time.