Role of Artificial Intelligence in Early Detection of Dental Caries using Radiographic Images

Background: Early detection of dental caries is essential to permit minimally invasive treatment. Artificial intelligence (AI) applied to radiographic images may improve diagnostic sensitivity and throughput. Objective: To evaluate the diagnostic performance of an AI model for radiographic detection of dental caries and to identify clinical and socioeconomic predictors of radiographic caries. Study Design: A Cross-sectional study. Place and Duration of the Study: Department of Operative Dentistry and Endodontics, School of Dentistry, Shaheed Zulfiqar Ali Bhutto Medical University, Islamabad, Pakistan, from July to December 2024. Materials and Methods: Total of 100 anonymous intraoral radiographs were preprocessed and annotated by dental radiologists. The image dataset was divided (70:15:15) for model training/validation/testing. Performance vs radiologist reference was evaluated with confusion matrix–derived metrics (sensitivity, specificity, PPV, NPV, F1, accuracy), and logistic regression was used to identify predictors of radiographic caries. Results: The AI model achieved overall accuracy of 90.0% (95% CI 82.6–94.5). Sensitivity was 94.0% (95% CI 86.7–97.4) and specificity 70.6% (95% CI 46.9–86.7). PPV and NPV were 94.0% (95% CI 86.7–97.4) and 70.6% (95% CI 46.9–86.7), respectively; F1-score = 0.94. Logistic regression found high DMFT (>10) (adjusted OR = 2.85; 95% CI 1.20–6.78; p = 0.018) and low socioeconomic status (adjusted OR = 2.10; 95% CI 1.00–4.40; p = 0.049) as independent predictors. Conclusion: AI shows high sensitivity and overall accuracy as an adjunct for radiographic caries detection. Preventive programmes should consider AI-based screening integrated with targeted interventions for high-risk groups.