Diagnostic Accuracy and Efficiency of AI-Assisted Radiographic Interpretation Compared to Conventional Methods in Early Detection of Dental Caries.

BACKGROUND

Early detection of dental caries is critical for implementing preventive measures and avoiding invasive treatments. Traditional radiographic interpretation by clinicians, though effective, is prone to variability. Artificial intelligence (AI)-assisted interpretation has emerged as a promising adjunct.

OBJECTIVE

To assess and compare the diagnostic accuracy and efficiency of AI-assisted radiographic interpretation with conventional clinician-based evaluation in the early detection of dental caries.

METHODS

A cross-sectional analytical study was conducted on 400 bitewing radiographs from 200 patients aged 18-40 years. Two groups were compared: Group A (conventional method) with three experienced clinicians, and Group B (AI-assisted method) using a convolutional neural network (CNN)-based software. The reference standard was consensus diagnosis after clinical validation. Diagnostic performance was evaluated using sensitivity, specificity, accuracy, and time-efficiency metrics.

RESULTS

AI-assisted interpretation showed significantly higher sensitivity (91.2% ± 3.5) than conventional interpretation (84.6% ±4.2; P = 0.01). Specificity was marginally better in the conventional group (89.5% ±3.1) compared to AI-assisted (87.1% ± 2.9; P = 0.14). Overall diagnostic accuracy was higher in the AI group (89.4% ± 3.2 vs. 87.0% ±3.6; P = 0.04). Mean interpretation time was significantly lower in the AI group (21.5 ± 5.3 seconds) than in the clinician group (48.2 ± 6.7 seconds; P < 0.001).

CONCLUSION

AI-assisted radiographic interpretation demonstrated superior diagnostic accuracy and markedly improved efficiency compared to conventional methods in early caries detection. AI tools can serve as reliable diagnostic adjuncts, enhancing clinical workflow and reducing diagnostic variability.