Deep Learning Model Evaluation for Predicting Eye Refraction in Adults with Myopia Using Standard Non-Dilated Eye Measurements

What this Trial Is About

Myopia is a common and irreversible vision disorder that affects quality of life. The standard method to measure refractive error in adults considering optical or surgical correction is cycloplegic refraction, but it can be time-consuming, uncomfortable, and slow to recover from. This research aims to develop machine learning models that predict cycloplegic refractive error in adults with myopia using standard non-cycloplegic eye measurements, potentially reducing the need for cycloplegic drops while identifying patients who still require them. The study uses data from adults with myopia undergoing routine eye exams at refractive surgery centers. Participants are divided into two groups based on the difference between non-cycloplegic and cycloplegic refraction values those with a difference of 0.50 diopters or more and those with less than 0.50 diopters. All participants receive routine cycloplegic refraction with tropicamide, and a machine learning model is applied to non-cycloplegic parameters to predict the cycloplegic spherical equivalent. Participants undergo paired non-cycloplegic and cycloplegic eye measurements within 7 days. Researchers evaluate the accuracy of the predicted cycloplegic refraction, how well the model identifies patients needing cycloplegic refraction, and the agreement between predicted and measured values. The study collects detailed clinical data including visual acuity, corneal health, and intraocular pressure. Participants are observed without additional interventions beyond standard eye exams and refraction tests.

DL Models Predicting Cycloplegic Refractive Error Based on Non-Cycloplegic Parameters in Myopic Adults