Leong E1

1Royal Melbourne Hospital, 2Save Sight Institute

Biography:

I am a junior doctor passionate about Ophthalmology and global health, committed to improving equitable access to eye care in regional and remote communities.

Abstract:

Purpose: Microbial keratitis (MK) is a vision-threatening corneal infection that demands rapid and accurate identification of causative organisms and is one of the leading cause of corneal blindness globally. Conventional diagnostics such as culture and microscopy are limited by time, sensitivity and accurate identification of causative microorganisms. This systematic review evaluates the diagnostic performance of artificial intelligence (AI) models for the diagnosis of MK.

Methods: This review was conducted in accordance with PRISMA guidelines. A comprehensive search of PubMed, Scopus, Cochrane Library, ScienceDirect, and Google Scholar identified peer-reviewed studies from 2005–2025 using combinations of the terms “microbial keratitis,” “infectious keratitis,” “corneal ulcer,” “artificial intelligence,” “machine learning,” “deep learning,” and “neural networks.” Data were extracted on study design, AI model type, input modality (e.g., corneal topography, slit-lamp, and anterior segment OCT images), and diagnostic performance metrics. Quality assessment was performed using the QUADAS-2 tool.

Preliminary Findings: Early evidence suggests that AI algorithms, particularly deep learning models, show strong potential in differentiating infectious from non-infectious keratitis and in supporting clinical decision-making.

Conclusion: AI shows promising diagnostic accuracy, achieving performance comparable to corneal specialists in identifying microbial keratitis. These models may enhance diagnostic efficiency, especially in resource-limited or remote settings. Further multi-centre, prospective studies are needed to confirm generalisability and enable integration into clinical practice.