Diagnostic accuracy of artificial intelligence compared to histopathologic examination in assessment of oral cancer – A systematic review and meta-analysis

Abstract

Screening and early detection of oral cancer have always proved to be a diagnostic dilemma and challenging for oral physicians. Artificial intelligence (AI) has lately emerged as a promising new tool in this area. The aim of this systematic review was to explore the accuracy of AI-based technology compared to gold standard routine histopathological examination in the diagnosis of oral cancer. The study was carried out using PRISMA guidelines. Studies published between 1-1-2000 and 31-12-2022, searched using three databases (PubMed, DOAJ, and Google Scholar) were reviewed, and data extraction was conducted from selected eight studies by two independent reviewers. Meta-analysis was carried out among studies with similar outcomes. Pooled sensitivity of AI was found to be 0.83 (95% CI: 0.80-0.86). This value was statistically significant (P < 0.05). However, heterogeneity (I2) value was 92%, indicating high heterogeneity. Our review and meta-analysis indicated that AI was efficient in diagnosing oral malignant and premalignant lesions when compared to the gold standard, i.e. histopathological examination.

Keywords

• artificial intelligence
• diagnosis
• histology
• machine learning
• oral cancer
• oral precancer