Validity and reliability of an artificial intelligence-based application for detecting dental caries in the pediatric...

Dental caries remains one of the most prevalent diseases in children, requiring early and accurate detection to prevent progression. Conventional visual examination, although widely used, is subjective and operator-dependent. With advancements in artificial intelligence (AI), image-based AI tools offer potential for more consistent caries detection. This study aimed to compare the validity and reliability of an AI tool (ScanO) with conventional visual examination in identifying dental caries in children.

This observational study included 200 children attending the Department of Pediatric and Preventive Dentistry. Visual examination was performed independently by two trained examiners, and decayed, missing, and filled teeth (DMFT)/deft scores were recorded. Standardized intraoral photographs were captured and analyzed using the AI tool ScanO, which generated corresponding DMFT/deft scores. Data were analyzed using SPSS version 21.

The AI tool successfully detected dental caries but consistently underestimated DMFT/deft scores compared to clinical examination (3.48 ± 2.94 vs. 4.70 ± 3.31; P < 0.001). Despite this, the intraclass correlation coefficient (ICC) analysis demonstrated excellent reliability for the AI tool (ICC = 0.968 single, 0.984 average) and near-perfect reliability for clinical examination (ICC = 0.999).

The AI tool demonstrated good reliability in detecting dental caries but consistently underestimated DMFT/deft scores compared to clinical examination. The clinical method identified more carious lesions, reaffirming its role as the gold standard. Within the limitations of this study, the AI tool can be considered a supplementary screening aid in pediatric dentistry but cannot replace clinical judgment.