Neural Network-Based Inverse Model for Non-Invasive Estimation of Corneal Mechanical Properties

This study presents a novel approach to estimate corneal mechanical properties using non-contact tonometry data and machine learning techniques. A neural network-based inverse model was developed to predict Ogden material parameters from corneal apex displacement data. The model was trained on simulated data generated via finite element analysis. Rather than employing standard evaluation metrics, the mechanical behavior of the material model was integrated into the model as the loss function, which minimizes the difference in stress fields between predicted and reference data. The method demonstrated strong performance in accurately predicting the mechanical response of the cornea. This approach offers a promising non-invasive diagnostic tool, bridging the gap between clinical measurements and complex biomechanical properties.

Keywords:

Cornea , Inverse Problem , Machine Learning , Neural Networks , Ogden Material Model