DIAGNOSTIC ACCURACY OF AI-BASED VERSUS CONVENTIONAL RADIOGRAPHIC CARIES DETECTION IN PEDIATRIC PATIENTS: A CROSS-SECTIONAL STUDY

Background: Dental caries remains a leading oral health concern in children, often requiring early and accurate diagnosis to prevent progression. Conventional radiographic methods, though widely used, can be limited by human interpretation variability. Artificial intelligence (AI)-based diagnostic tools have emerged as promising alternatives, offering consistency and enhanced lesion detection, yet their clinical utility in pediatric settings remains underexplored. Objective: To compare the diagnostic accuracy of AI-assisted caries detection tools with conventional radiographic evaluation in pediatric dental patients. Methods: A cross-sectional study was conducted over eight months at a tertiary pediatric dental center involving 120 children aged 6–14 years. Standardized bitewing radiographs were analyzed using two methods: independent evaluation by calibrated pediatric dentists and AI-assisted analysis via a deep learning-based diagnostic system. Diagnostic performance was measured using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy. McNemar’s test was applied to compare paired proportions, and Cohen’s kappa assessed inter-rater reliability among clinicians. Ethical clearance and informed consent procedures were completed. Results: AI-assisted detection showed significantly higher diagnostic performance, with sensitivity of 88.3%, specificity of 90.8%, PPV of 89.1%, NPV of 89.9%, and overall accuracy of 89.6%. Conventional radiography yielded lower values across all metrics, including sensitivity (72.5%) and accuracy (78.8%). Statistical analysis confirmed significant differences between the two methods (p < 0.05), favoring AI tools for consistent caries detection in children. Conclusion: AI-assisted caries detection demonstrates superior diagnostic accuracy compared to conventional radiographic interpretation in pediatric patients, supporting its integration as a reliable clinical decision aid in routine dental practice.