Ensemble Machine Learning Prediction Model for Ocular Refraction Using Partial Interferometry Measurements in Childhood

Abstract

Abstract To develop an ensemble machine learning prediction model for ocular refraction in childhood using partial interferometry measurements, Age, sex, cycloplegic refraction, and partial interferometry data collected within one month were obtained from patients aged 5 to 16 years. Four ensemble regression models were used to develop prediction models of spherical equivalents (SE) from the data. Root mean squared error (RMSE) was used to compare the accuracy among the models. That was compared with that of a previously developed multiple linear regression model. 4156 eyes from 1965 patients (50.3% female) were included. Mean age was 8.4 ± 2.3 years and mean SE was-1.01 ± 2.94 diopters(D). Mean axial length(Axl) was 23.63 ± 1.41 mm and mean keratometry reading of flat and steep axis was 43.58 ± 1.40 D. Developed models had accuracy of RMSE 0.800 to 0.829 D, which was superior to that of the conventional model (1.213 D). Simulations showed that female was associated more with myopia than that of male. Long eyes showed an increase in the myopic refraction per Axl. Refractive errors can be calculated in the childhood with ocular biometric parameters using new models. Moreover, the models could simulate hypothetical relationships between ocular parameters and SE.