Development of an AI-powered Diagnostic System for Early Caries Detection

Thesis Abstract

Early detection of dental caries remains a critical challenge in preventive dentistry, as conventional visual-tactile examination often lacks sensitivity for identifying incipient lesions, leading to delayed intervention and increased risk of tooth loss. Despite advances in digital imaging and adjunct diagnostic tools, there exists a significant gap in accessible, accurate, and rapid diagnostic systems capable of early caries identification in diverse clinical settings. This study aims to develop an artificial intelligence (AI)-based diagnostic system that leverages machine learning algorithms to enhance the precision and efficiency of early caries detection. The specific objectives include (1) to collect and annotate a comprehensive dataset of intraoral images representing various stages of caries, (2) to design and train AI models utilizing convolutional neural networks (CNNs) for lesion classification, (3) to evaluate the diagnostic performance of the developed system in comparison with expert dentists, and (4) to assess the system’s usability and integration potential in routine dental practice. The research adopts a quantitative, developmental, and evaluative design, utilizing a cross-sectional sampling approach. The study population consists of 500 patients aged 6 to 18 years attending public dental clinics across general hospitals, with diverse socio-economic backgrounds. A stratified random sampling method is employed to ensure balanced representation of different demographic groups. Data collection involves capturing high-resolution intraoral photographs using standard digital cameras, followed by expert annotation of caries stages based on visual and radiographic confirmation. The annotated images form the dataset used for training and testing the AI models. To validate the system’s diagnostic accuracy, 200 images from an independent validation set are analyzed using the developed CNN models and compared against diagnoses by a panel of three experienced dentists. Model performance metrics such as sensitivity, specificity, accuracy, and area under the receiver operating characteristic (ROC) curve are computed. The analytical framework includes regression analysis to explore the relationship between system outputs and clinical diagnoses, coupled with confusion matrix evaluation for model validation. Preliminary expected findings suggest that the AI-powered system will attain a diagnostic accuracy exceeding 85%, with sensitivity and specificity rates above 80%, demonstrating its potential to identify early demineralization lesions with high reliability. The system’s performance is anticipated to be comparable or superior to human expert diagnosis, indicating significant improvements in early caries detection efficacy. Additionally, qualitative feedback from dental practitioners indicates a high usability score, supporting the system’s integration into routine clinical workflows. This research contributes novel insights by integrating advanced deep learning techniques within a user-friendly diagnostic platform tailored for pediatric and general dental practices. It extends existing knowledge on computer-aided diagnosis (CAD) by providing empirical evidence on the effectiveness of CNNs in subclinical caries detection and establishing a scalable framework for deploying AI solutions in resource-constrained settings. The study’s findings offer a foundation for further exploration into multi-modal AI systems combining imaging with patient-specific risk factors, which could further refine diagnostic accuracy. The study concludes that AI-driven diagnostic systems have substantial potential to revolutionize early caries detection by enabling timely, accurate, and accessible diagnosis. Recommendations include the ongoing refinement of the system through longitudinal studies, wider clinical trials across diverse populations, and integration with electronic health records for comprehensive risk assessment. Policy implications involve advocating for the adoption of AI technologies in standard dental practice to improve patient outcomes and reduce the burden of untreated dental caries globally.