Advanced Comparative Analysis of Deep Learning and Transformer Models for Caries Detection on Dental Radiographs

Abstract

This study presents comparison between different deep learning models to handle Caries Detection on Dental Radiographs. These models including CNNs, among which the VGGNetI6, ResNet50, Efficient Net, and transformer architectures have been successfully used for the detection of dental caries in two types of dental radiographic images. The most significant improvements in this area were observed at the time of the study because we were deploying the VIT and SWIN vision transformers. We demonstrate that the Swin Transformer model achieved the highest accuracy of97.6%, which is even better than traditional CNN Models. Data augmentation and advanced preprocessing techniques were applied to support the generalizability of the model.