Development of an AI-powered mobile app for early caries detection in children

Thesis Abstract

Dental caries remains one of the most prevalent chronic childhood diseases globally, often going undetected during its early stages due to limited access to professional dental assessments and the inherent challenges in early visual diagnosis. This study addresses the pressing need for accessible, reliable, and non-invasive diagnostic tools by developing an artificial intelligence (AI)-powered mobile application aimed at facilitating early detection of caries in children, thereby promoting prompt intervention and reducing long-term oral health complications. The primary aim of the research is to design, implement, and evaluate a mobile application integrated with AI algorithms capable of accurately identifying early carious lesions from intraoral images captured via smartphones. Specific objectives include (1) to review existing diagnostic methodologies and AI applications in pediatric dentistry; (2) to develop a robust AI model trained on a diverse dataset of intraoral images representing various stages of caries; (3) to implement this model into a user-friendly mobile app; and (4) to validate the app’s diagnostic performance against standard clinical examinations conducted by dental professionals. The research adopts a mixed-methods approach combining quantitative and qualitative elements. The quantitative component involves a cross-sectional study design where intraoral images from a sample of 250 children aged 3 to 12 years are collected from dental clinics across urban and rural settings. The dataset encompasses images explicitly labeled by expert dentists to serve as ground truth. The AI model development employs supervised machine learning techniques, specifically convolutional neural networks (CNNs), trained on a training set of 200 images and validated on a separate set of 50 images. The app’s diagnostic accuracy is assessed using metrics such as sensitivity, specificity, positive predictive value, and overall accuracy through receiver operating characteristic (ROC) curve analysis. Qualitative data are gathered through structured interviews with pediatric dentists and potential end-users to evaluate usability and acceptability. Data analysis involves statistical techniques including logistic regression analysis to examine the app’s diagnostic performance and thematic analysis of interview transcripts to explore user perceptions. The anticipated findings suggest that the AI-powered app will demonstrate high sensitivity and specificity (expected values exceeding 85%) in detecting early carious lesions, comparable to expert clinical assessments. It is also expected to show high usability scores among clinicians and parents, indicating potential for widespread adoption. The study’s contribution to knowledge lies in providing empirical evidence supporting the integration of AI-driven mobile solutions within pediatric dental practices, expanding digital health interventions to underserved populations, and establishing a scalable model for early caries screening. The main conclusion underscores the app’s potential to serve as a supplementary screening tool, enabling early intervention, reducing the burden of untreated caries, and promoting oral health literacy among caregivers. Recommendations emphasize the need for further longitudinal studies to assess the app’s impact on clinical outcomes and health-seeking behaviors, as well as integration with existing dental technologies and health systems. Additionally, the study advocates for continuous refinement of AI models using larger, more diverse datasets to improve accuracy and address potential biases. Overall, the research advocates for leveraging ICT innovations to bridge gaps in pediatric oral health diagnostics and enhance preventive care in resource-limited settings.