Evaluation of severity of HSV-1 corneal infection by statistical pattern classification technique.

Abstract

PURPOSE: Slit lamp photography is a daily practice in ophthalmic examinations in clinic settings. Images of the ocular integrity enable the clinicians to evaluate the progress of disease. However, the lack of a consistent objective measuring scheme to evaluate the corneal epithelial defects is a concern. METHODS: Rabbits were ocularly infected with HSV-1. Fluorescein solution was applied to the rabbit eyes to reveal lesions. Images were taken via a Nikon digital camera equipped with cobalt blue excitation filter. A modified Bayes classifier (MBC) based on two carefully chosen features was applied to analyze images of infected rabbit corneas. RESULTS: MBC utilizing a feature set comprising green-to-red ratio and green-to-blue ratio could differentiate infected from normal pixels with an error rate as low as 2.2%. CONCLUSION: The ability of MBC to measure the percentage of corneal surface suffering from infection makes this approach a valuable tool for investigating the dynamics of corneal-related diseases.