<![CDATA[Introduction: Dental caries remains one of the most prevalent oral diseases worldwide. Conventional diagnostic methods rely heavily on clinician expertise and may lead to diagnostic variability, particularly in early lesion detection. Recent advances in artificial intelligence (AI), especially convolutional neural network (CNN)-based deep learning models, have shown promising performance in automated caries detection across various imaging modalities. Objective: This systematic review aimed to evaluate and summarise current evidence on CNN-based deep learning models for automated dental caries detection and segmentation, and to compare their diagnostic performance across imaging modalities. Methods: This review followed the PICO framework. Literature searches were conducted in PubMed, Cochrane, ScienceDirect, and Google Scholar for studies published between 2022 and 2025. English-language quantitative studies utilising CNN-based models for caries detection with measurable diagnostic outcomes were included. Of 262 records identified, 50 articles were screened, and 11 studies met the eligibility criteria for qualitative synthesis. Results: The included studies comprised 18,206 radiographic images and evaluated multiple CNN architectures, including YOLO, R-CNN, UNet, ResNet, and MobileNetV2. YOLO demonstrated the highest diagnostic performance, achieving a mean accuracy of 0.957 and a sensitivity of 0.976, while MobileNetV2 showed the highest specificity (0.954). Among imaging modalities, periapical radiographs exhibited the highest mean diagnostic performance (0.884), followed by intraoral photographs (0.821) and CBCT images (0.810), whereas bitewing radiographs showed the lowest value (0.791). Conclusion: CNN-based deep learning models demonstrate strong potential for automated detection of dental caries. Detection-based architectures, particularly YOLO, showed superior performance, while imaging modality influenced diagnostic outcomes.]]>
Copyrights © 2026
