Research in Medicine

Problem: Dental caries is among the most prevalent chronic diseases in Iran. Early and accurate detection of proximal caries cannot be achieved by visual-tactile examination alone and requires radiography. However, radiographic interpretation is subject to diagnostic error, and this error varies significantly across regions due to differences in dentists' experience and skill.
Options: AI-based systems using deep learning can serve as an assistive tool to improve caries detection accuracy. The underlying study shows that the EfficientNet model achieved 94.7% accuracy, 96.7% sensitivity, and 96.2% specificity.
Implementation Considerations: Successful deployment requires infrastructure (servers, cloud storage), dentist training, integration with PACS, and regulatory approval (e.g., from the Food and Drug Administration). Key challenges include data privacy, legal liability for AI errors, ongoing maintenance costs, and the role of the Medical Council in setting protocols.