Construction method of early keratoconus classification model

Abstract

The invention provides a cornea conus classification model construction method based on a Zernike polynomial and deep learning. The cornea conus classification model construction method comprises the following steps: acquiring cornea data and preprocessing the cornea data into a Zernike coefficient data set; wherein the cornea data comprises anterior cornea curvature data, posterior cornea curvature data, anterior cornea height data, posterior cornea height data and cornea thickness data; constructing a keratoconus classification model based on a BP neural network; wherein the output categories of the keratoconus classification model comprise normal keratoconus, subclinical keratoconus and keratoconus; and classifying the Zernike coefficient data set to form a training set and a test set, and training and testing the keratoconus classification model to obtain a finally trained keratoconus classification model. By implementing the method, the problem of low detection precision of the existing machine learning method can be solved.

Description

technical field [0001] The invention relates to the technical fields of computer data processing and medical technology, in particular to a method for constructing a keratoconus classification model based on Zernike polynomials and deep learning. Background technique [0002] The cornea provides approximately 70% of the eye's focusing power and is the most efficient refractive element of the eye. The optical action of the cornea is mainly determined by its sphericity and the difference in refractive index at the air-tissue interface. Keratoconus (KC) is a progressive, noninflammatory thinning and cone-shaped protrusion of the cornea. It is usually detected in teens to early adulthood and causes increasing myopia and astigmatism. Early detection of KC is crucial in the preoperative screening of refractive surgery candidates, as surgery on undiagnosed keratoconus may lead to unsatisfactory postoperative outcomes and iatrogenic corneal ectasia. [0003] Machine learning has ...

Images