Neural Network Analysis for Prediction of Material Property in Human Eye with Keratoconic Cornea

Characterizing the material property of human eye is necessary to provide any biomechanical solution of eye diseases. In this study, a new approach for prediction of material property in a human eye with keratoconic cornea has been presented by using Finite Element and Neural Network analysis. In the Finite Element model, a hyper-elastic cornea was used to obtain the deformation of cornea due to intraocular pressure. The amount of keratoconus extension was simulated by sectioning the cornea in to weakened angular section tied to a healthy section. The Finite element model was used to obtain cornea curvature for different material properties. Obtained results were then used as data points in the Neural Network program. The regression and performance plots of implemented Neural Network analysis suggest that predictions are reasonable and reliable. Trained network can be used for accurately characterizing the material properties of keratoconic cornea from curvature maps of keratometers.