Deep-learning-based Prediction of Late Age-Related Macular Degeneration Progression

ABSTRACT

Age-related macular degeneration (AMD) is a leading cause of irreversible blindness worldwide. Previous genetic and epidemiological studies have shown that the AMD progression are affected by genetic variants and environmental factors. AMD severity is mainly diagnosed by color fundus images and recent studies have shown the success of machine learning methods in predicting AMD progression using image data. However, none of the studies have utilized both genetics and image data for predicting AMD progression. The Age-Related Eye Disease Study (AREDS), a large-scale clinical trial from the National Eye Institute, includes massive genome-wide genotyping data, longitudinal color fundus photographs, and disease severity assessment over a period of 12 years, providing an unprecedented opportunity to investigate prediction models for AMD progression. In this report, we jointly used genotypes and fundus images to dynamically predict an eye as having progressed to late AMD (e.g., whether the time to progression to late AMD for the eye exceeds 3 years from the current visit) with a modified deep convolutional neural network (CNN). In total, we used 31,262 color fundus images centered on the macula from 1,351 subjects with corresponding genotypes and phenotypes available at baseline and follow-up visits. The first part of this model was derived from a CNN to extract features. Our results showed that the color fundus photos coupled with genotypes could predict late AMD progression with an averaged area under the curve (AUC) value of 0.85 (95%CI: 0.83-0.86). The results using fundus images alone showed an AUC of 0.81 (95%CI: 0.80-0.83). We have validated our results in an independent dataset of 200 Caucasians extracted from UK Biobank and the results showed an AUC of 0.9 (95%CI: 0.85-0.94) for predicting whether the eye progresses to late AMD exceeding the 3 years. We implemented our model in a cloud-based application for individual risk assessment.