Intelligent Dental Disease Diagnosis and Personalized Treatment Recommendation Using Mobilenetv3

Many dental problems go unnoticed until they cause pain or visible damage, which makes easy-to-reach screening and prompt professional follow-up valuable.In this paper we describe a web-based decision-support system that pairs deep-learning image analysis with visual explanation, patient-friendly guidance, and a dentist's review.The users upload an intraoral picture or dental x-ray, which is then fed into a MobileNetV3 classifier dedicated to that particular modality.There are seven classes for each classifier.The pictures are categorized according to the following classes: calculus, caries, gingivitis, healthy, hypodontia, oral ulcer, and tooth discoloration, whereas x-rays are classified in the following order: dental caries, dental infection, endoperio lesion, impacted tooth, periapical lesion, periodontal bone loss, and root fracture.The application presents the user with a Grad-CAM heatmap showing the part of the image that made the AI give a particular diagnosis.In addition, it gives the confidence score, indicative severity, and corresponding home care recommendations.Prior to giving the final diagnosis to the patient, the dentist can approve, deny, or modify the prediction of the AI.The application is developed using React.js,FastAPI, and PostgreSQL technologies and offers a prediction history, PDF reports, and a chatbot powered by Llama 3.3 model hosted on Groq.We present our work as a prototype of clinical decision support rather than fully autonomous diagnostic software since its accuracy and practical use require further validation on independent test sets and in prospective studies with clinicians.