Diagnosis of Keratoconus Using Machine Learning

Abstract

Keratoconus is a progressive, non-inflammatory corneal disorder that can significantly impair vision if not detected and treated early. Accurate diagnosis of keratoconus, especially in its early stages, is crucial to prevent severe visual deterioration and reduce the need for invasive treatments such as corneal transplantation. This study proposes a machine learning-based approach for the diagnosis of keratoconus using topographic and tomographic features of the cornea. A large dataset containing 423 features was analyzed, and univariate feature selection was applied to identify the most discriminative attributes. Several supervised learning algorithms—including Random Forest, Support Vector Machines, k-Nearest Neighbors, and Logistic Regression—were trained and evaluated. The Random Forest classifier achieved the highest diagnostic accuracy of 95.8%, showcasing the potential of machine learning in aiding clinicians with accurate and early detection of keratoconus.

Keywords: Microbial keratitis and inflammation, Probable allograft rejection, Transplant failure, Contact lenses, astigmatism