Cavident: Smart Cavity Detection using Deep Learning

Mr. Abhilash L Bhat1, Keerthana.M2, Khushi M P3, M Bhoomika4, Mythri B M5
1Assistant Professor, Dept of Computer Science and Engineering, K.S Institute of Technology 2Student, Dept of Computer Science and Engineering, K.S Institute of Technology 3Student, Dept of Computer Science and Engineering, K.S Institute of Technology 4Student, Dept of Computer Science and Engineering, K.S Institute of Technology 5Student, Dept of Computer Science and Engineering, K.S Institute of Technology

Abstract - In recent years, dental caries is one of the most common oral diseases. Due to lack of awareness and limited access to dental care mainly in rural areas so the early detection is neglected. The already existing treatment methods are based on clinical expertise or some of the specialized imaging and x-rays which are not always accessible and cost-effective. The early detection is very much needed to prevent the progressive tooth decay which in return reduces the need for invasive treatment. This paper presents the usage of YOLOv8 applied to standard photographic images which leads to the automatic dental cavity detection and risk assessment system. The system performs real-time localization of cavity regions through object detection and further leading to the analysis of the detected lesion using image processing techniques to procure clinically related features such as lesion size, colour intensity, texture irregularity, and boundary characteristics. The proposed system is integrated with a rule-based clinical reasoning module to generate interpretable textual explanations which addresses the disadvantages of black-box deep learning models. In addition to this, a quantitative risk scoring mechanism is initiated to classify dental caries into three categories such as low, medium and high-risk based on the extracted features. Experimental evaluation shows that the preferred system achieves reliable detection performance and provides a very transparent and clear clinically meaningful outputs. The system is designed in order to operate on non-specialized images while making it suitable for mobile-based prior screening and tele-dentistry applications. This work come up with the development of accessible, explainable, and also a very cost-effective AI approach for early awareness and dental diagnosis, particularly in underprivileged areas.

Key Words: Dental Cavity Detection, YOLOv8, Image Processing, Lesion Analysis, Feature Extraction, Explainable AI, Risk Assessment, Oral Health Screening