Detection of errors in bitewing x-ray images using deep learning

Quality assurance (QA) is a process put in place in the hospital to guarantee ideal diagnostic image quality with minimum danger to patients. It entails frequent quality control checks, preventive support procedures, authoritative approaches, and planning. The process of acquiring quality images, especially for radiography students and trainees, requires a steep learning curve. This study proposes deep learning models that may serve as a guide to ensure proper images are captured and help improve the quality assurance process. The models are intended to determine that the images captured are optimal by ensuring adequate precautions in the capturing process, thereby automatically identifying and correcting any mistakes or issues in the quality or interpretation of the image. This study acquired 4955 radiographs that have been labeled by dental experts. Four deep learning models, specifically CNN, AlexNet, RestNet-50, and ViTs have developed with respective accuracies of 78.98%, 24.84%, 78.03%, and 81.34%. The performance results show that deep learning models have the potential to be utilized to assist dental practitioners in error detection and quality assurance