Cornea experts from Spain created a simple network classifier system they say can detect keratoconus with significant accuracy, and it can be easily used with any placido-based topographic system, as it is platform-independent.

Their study, published in Contact Lens & Anterior Eye, describes a new Bayes classification model for keratoconus detection that used primary placido-based corneal indices noted in the literature and computed directly from the image of the discs reflected on the cornea.

The comparative study included 60 eyes of 60 patients between the ages of 20 and 60. Patients were divided into two groups: a control group with normal corneas (30 eyes) and a keratoconus group (30 eyes). The keratoconus group included all grades except grade IV with excessively distorted corneal topography.

The researchers examined all cases using topography and computed their primary corneal placido indices. The then built a classifier by fitting a conditional linear Gaussian Bayesian network to the data, supported by cross-validation. For comparison, researchers added white noise of different magnitude to the original data.

Researchers said the Bayes classifier showed perfect discrimination ability among normal and keratoconic corneas, with 100% of sensitivity and specificity, even in the presence of significant noise or incomplete data.