Abstract
BACKGROUND: Pemphigus and pemphigoid disease pose diagnostic challenges to pathologists and clinician due to overlapping clinical and histopathological features. The systematic review evaluated the diagnostic accuracy of artificial intelligence (AI) in identifying pemphigus and pemphigoid disorders based on clinical images, compared to conventional diagnostic techniques. METHODS: A search was carried out in PubMed, Scopus, and Web of Science databases since inception using search strategy comprising of "Artificial Intelligence", "Pemphigus", "Pemphigoid", "Diagnosis", "sensitivity", "specificity", and "accuracy". Studies examined AI models' accuracy for diagnosing pemphigus and pemphigoid were included in the review. A meta-analysis was conducted using Meta-Disc software, with pooled sensitivity, specificity, and summary receiver operating characteristic (SROC) curve analysis. RESULTS: Out of 4094 search results, five studies met the eligibility criteria after screening and selection steps. The pooled diagnostic accuracy of AI-based systems was 0.83 (CI 0.73-0.90) with a of 0.86 % (CI: 0.76-0.91) pooled sensitivity and 0.84 % (CI: 0.79-0.87) pooled specificity, reflecting its moderate effectiveness in ruling out pemphigus and pemphigoid disorders. Further, the positive likelihood ratio (PLR) was 5.34 (95 % CI: 4.05-7.05), and the negative likelihood ratio (NLR) was 0.16 (95 % CI: 0.09-0.28), with a DOR of 32.53 (95 % CI: 14.70-71.94) indicates moderate to strong accuracy. CONCLUSION: AI have moderate diagnostic efficacy in diagnosis of pemphigus and pemphigoid diseases. However, additional research is needed to develop standardized methodologies and ensure generalizability across different populations before integrating into clinical practice.