Search

To assess the performance of machine learning (ML) ensemble models for predicting patient subjective refraction (SR) using demographic factors, wavefront aberrometry data, and measurement quality related metrics taken with a low-cost portable autorefractor.Four ensemble models were evaluated for predicting individual power vectors (M, JAll models considerably outperformed the predictions from the autorefractor, while ASB obtained the best results. The accuracy of the predictions for each individual power vector component was substantially improved resulting in a ± 0.63 D, ±0.14D, and ±0.08 D reduction in the 95% limits of agreement of the error distribution for M, JThese results suggest that ML is effective for improving precision in predicting patient's SR from objective measurements taken with a low-cost portable device.

This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited, Purpose To assess the performance of an open-view binocular handheld aberrometer (QuickSee) for diagnosing refractive errors in children. Methods 123 school-age children (9.9 ± 3.3 years) with moderate refractive error underwent autorefraction (AR) with a standard desktop device and subjective refraction (SR), with or without cycloplegia to determine their eyeglass prescription. Measurements with QuickSee (QS) were taken in 62 of these patients without cycloplegia (NC), and in 61 under cycloplegia (C). Differences in refraction values (AR vs SR vs QS) as well as the visual acuity (VA) achieved by the patients with each method (QS vs SR) were used to evaluate the performance of the device in measuring refractive error. Results The spherical equivalent refraction obtained by QS agreed within 0.5 D of the SR in 71% (NC) and 70% (C) of the cases. Agreement between the desktop autorefractor and SR for the same threshold was of 61% (NC) and 77% (C). VA resulting from QS refractions was equal to or better than that achieved by SR procedure in 77% (NC) and 74% (C) of the patients. Average improvement in VA with the QS refractions was of 8.6 and 13.4 optotypes for the NC and C groups respectively, while the SR procedure provided average improvements of 8.9 (NC) and 14.8 (C) optotypes. Conclusions The high level of agreement between QuickSee and subjective refraction together with the VA improvement achieved in both study groups using QuickSee refractions suggest that the device is a useful autorefraction tool for school-age children, Eduardo Lage is funded by the Ramon y Cajal program from the Spanish Ministry of Economy, Industry and Competitivity (RYC-2016-21125). Pablo Perez-Merino was funded by the Juan de la Cierva program from the same Ministry (FJCI-2015-27101). Carlos S. Hernandez and Andrea Gil are funded by the Madrid Regional Government through IND2019/TIC-17116 grant

Significance There is a critical need for tools that increase the accessibility of eye care to address the most common cause of vision impairment: uncorrected refractive errors. This work assesses the performance of an affordable autorefractor, which could help reduce the burden of this health care problem in low-resource communities. Purpose The purpose of this study was to validate the commercial version of a portable wavefront autorefractor for measuring refractive errors. Methods Refraction was performed without cycloplegia using (1) a standard clinical procedure consisting of an objective measurement with a desktop autorefractor followed by subjective refraction (SR) and (2) with the handheld autorefractor. Agreement between both methods was evaluated using Bland-Altman analysis and by comparing the visual acuity (VA) with trial frames set to the resulting measurements. Results The study was conducted on 54 patients (33.9 ± 14.1 years of age) with a spherical equivalent (M) refraction determined by SR ranging from -7.25 to 4.25 D (mean ± SD, -0.93 ± 1.95 D). Mean differences between the portable autorefractor and SR were 0.09 ± 0.39, -0.06 ± 0.13, and 0.02 ± 0.12 D for M, J0, and J45, respectively. The device agreed within 0.5 D of SR in 87% of the eyes for spherical equivalent power. The average VAs achieved from trial lenses set to the wavefront autorefractor and SR results were 0.02 ± 0.015 and 0.015 ± 0.042 logMAR units, respectively. Visual acuity resulting from correction based on the device was the same as or better than that achieved by SR in 87% of the eyes. Conclusions This study found excellent agreement between the measurements obtained with the portable autorefractor and the prescriptions based on SR and only small differences between the VA achieved by either method.