Comprehensive Summary
This systematic review by Sahoo et al. broadly investigates the accuracy of AI-driven imaging techniques to diagnose and detect oral potentially malignant disorders (OPMDs) and oral cancer. Abiding by the PRISMA guidelines, the authors searched PubMed, Scopus, and IEEE databases, screening 296 articles and including 55 in qualitative synthesis, with 18 selected for meta-analysis. According to the results, the meta-analysis revealed AI models achieved very high diagnostic accuracy. With a pooled sensitivity of 0.87, specificity of 0.81, a diagnostic odds ratio of 131.63, and an area under the curve (AUC) value of 0.9758, these AI models showed excellent accuracy overall. Deep learning architectures, mainly convolutional neural networks (CNNs), demonstrated the strongest detection capability, with histopathological imaging showing the highest sensitivity and specificity.
Outcomes and Implications
These findings highlight the potential of AI algorithms to work as reliable and efficient tools for early diagnosis of OPMDs and oral cancer, especially in settings where conventional diagnostic frameworks are extremely limited. AI-driven technologies’ high specificity suggests they can indeed improve early detection, which is drastically better for survival outcomes. Some other implications include how the meta-analysis shows AI performs very well, but the results may be biased toward English-speaking research settings. Therefore, the reported accuracy of AI could be slightly inflated, as insights from non-English speaking populations are missing. Finally, in low-resource countries where specialists are scarce, AI models incorporated in imaging devices could be incredibly useful for communities, as they would provide referrals for suspicious lesions.