ENHANCING DIAGNOSTIC PRECISION: A DEEP LEARNING APPROACH TO AUTOMATED DENTAL CARIES DETECTION IN BITEWING RADIOGRAPHS

Abstract

Dental Cavities are one of the most common conditions pertaining to oral health, which affect all age groups across the global population. If the condition is not diagnosed in time, the consequences can cause severe complications, including but not restricted to the loss of teeth, infections, and higher costs of treatment. The currently existing technique for diagnosing dental conditions in clinics depends greatly upon the clinical inspection of the dentist and the evaluation of radiographic images, which can be rather time-consuming and impersonal. Also, in less advanced areas, there might not be easy access to competent dentists. This work introduces SmileScan: a web-based AI system developed. It is intended for the automatic detection of dental cavities from dental images using deep learning. The proposed work is interested in the binary classification problem: classifying the image into the cavity class or the non-cavity class. A MobileNetV2 model fine-tuned on a customized dental dataset is utilized. Methods of image preprocessing and augmentation are employed.The deployed model uses the Django backend and the React frontend and helps users upload dental images, obtaining an immediate result set with confidence levels. Experiments show the proposed system reliably works at an accuracy level of 85.71% and helps implement real-time cavity identification. The smile scan system represents an inexpensive and scalable resource that could aid in the early diagnosis and effective delivery of dental services.