Deep learning for predicting refractive error from retinal fundus images

摘要

Refractive error, one of the leading cause of visual impairment, can becorrected by simple interventions like prescribing eyeglasses. We trained adeep learning algorithm to predict refractive error from the fundus photographsfrom participants in the UK Biobank cohort, which were 45 degree field of viewimages and the AREDS clinical trial, which contained 30 degree field of viewimages. Our model use the "attention" method to identify features that arecorrelated with refractive error. Mean absolute error (MAE) of the algorithm'sprediction compared to the refractive error obtained in the AREDS and UKBiobank. The resulting algorithm had a MAE of 0.56 diopters (95% CI: 0.55-0.56)for estimating spherical equivalent on the UK Biobank dataset and 0.91 diopters(95% CI: 0.89-0.92) for the AREDS dataset. The baseline expected MAE (obtainedby simply predicting the mean of this population) was 1.81 diopters (95% CI:1.79-1.84) for UK Biobank and 1.63 (95% CI: 1.60-1.67) for AREDS. Attentionmaps suggested that the foveal region was one of the most important areas usedby the algorithm to make this prediction, though other regions also contributeto the prediction. The ability to estimate refractive error with high accuracyfrom retinal fundus photos has not been previously known and demonstrates thatdeep learning can be applied to make novel predictions from medical images.Given that several groups have recently shown that it is feasible to obtainretinal fundus photos using mobile phones and inexpensive attachments, thiswork may be particularly relevant in regions of the world where autorefractorsmay not be readily available.