DentAI Vision: Automated Dental Pathologies Detection and Classification in OPG Images Using Convolutional Neural Networks

Panoramic radiographs, known as orthopantomograms (OPGs), give dentists a single view of the teeth, jaws and surrounding structures, but reading them takes time and depends on the reader's experience. Convolutional neural networks (CNNs) can learn visual patterns directly from labelled images, which makes them a natural candidate for assisting this review. Many published systems, however, report a predicted label and little else, and a label alone is hard for a clinician to judge. This paper proposes DentAI Vision, a web-based research framework that classifies OPG images with a CNN and presents the result together with a confidence score, a class probability distribution and a Grad-CAM visualization of areas of interest. A Django application handles authentication, image validation, preprocessing, inference, record storage and PDF reporting. We describe the problem, the research gap, the architecture and the planned evaluation. The web application has been developed, but no trained model has yet been evaluated with it, so no measured results are reported. The output is intended as an aid for research and professional review, not as a diagnosis.