The Role of Artificial Intelligence in the Diagnosis of Ocular Surface Squamous Neoplasia: a systematic review and meta-analysis

Relevance: Ocular Surface Squamous Neoplasia (OSSN) is the most common malignancy of the ocular surface; timely non-invasive detection can preserve vision. This systematic review and diagnostic meta-analysis revealed that AI-based models performed with high accuracy in diagnosing OSSN. Future research requires comprehensive prospective studies, along with improvements in methodology to strengthen AI-based diagnosis. Abstract: This review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses(PRISMA) guidelines. We systematically searched four databases, PubMed, Web of Science, Embase, and Scopus, to explore the use of AI in diagnosing OSSN. The QUADAS-2 tool was used to find the risk of bias across the included studies. A diagnostic meta-analysis was conducted to determine the pooled sensitivity and specificity of AI models used in various studies to diagnose the OSSN. The summary receiver operating curve (SROC) was drawn to calculate the area under the curve (AUC).A total of 51 studies were found, and 6 studies were included in this systematic review, and four studies were eligible to be included in the diagnostic meta-analysis. All the included studies had at least one significant bias in methodology, leading to an overall high risk of bias. The diagnostic meta-analysis showed a pooled sensitivity of 95.5 % (95% CI: 67.5 to 99.5) and a specificity of 96.4 % (95% CI: 87.7 to 99.0). The summary receiver operating curve (SROC) revealed high accuracy (0.95) of AI models in diagnosing OSSN. AI-based models demonstrated high sensitivity and specificity in the diagnosis of ocular surface squamous neoplasia. AI can reliably diagnose OSSN with both deep learning and machine learning models.