Deep learning for predicting uncorrected refractive error using posterior segment optical coherence images
Description
***A sample anonymized OCT data with and without high myopia is available in this source. Note that this was not the exact data used in the research, but is a cleaned-up reproduction of our study's key insight. Background/Aims: The purpose of this study is to evaluate the deep learning model for estimating uncorrected refractive error using posterior segment OCT images. Definition of refractive error Refractive error was measured by expert optometrists during manifest refraction without cycloplegia. The target metric for refractive error is spherical equivalent (SE), which is defined as the spherical power whose focal point coincides. The SE is calculated by adding the sum of the spherical refractive power with half of the cylinder power. Specifically, we defined high myopia as an SE of -6.00 D. Overview This retrospective study protocol was approved by the Institutional Review Board of Korean National Institute for Bioethics Policy, which waived the requirement for informed consent. This study adhered to the tenets of the Declaration of Helsinki. In current study, we analyzed preoperative ocular data healthy patients who intended to undergo refractive surgery at B&VIIT Eye center from January 2016 to December 2016. We retrospectively collected preoperative refractive error and OCT measurements that was used to develop the machine learning model to exclude patients who underwent refractive surgery. All patients underwent preoperatively measurements of best-corrected distance visual acuity and manifest refraction, slit-lamp examinations of the anterior segment, and dilated fundus examinations. Eyes with a history of previous ocular surgery, corneal diseases, cataract, glaucoma, uveitis, and retinal diseases were excluded from the study. The inclusion criteria for ICL implantation at the B&VIIT Eye Center were: age between 21 and 50 years old, stable refraction, +5.00 to -20.00 diopters of hyperopia or myopia with astigmatism of 5.00 D or less, and subjects who were examined for the retina and optic disc using 3-dimensional OCT (Topcon 3D OCT‐1 Maestro, Tokyo, Japan). We captured retinal cross-sections passing through the center of the retina (fovea) with horizontal and longitudinal axes during OCT examination.