Predicting the risk of developing diabetic retinopathy using deep learning

Abstract

Background Diabetic retinopathy screening is instrumental to preventing blindness, but scaling up screening is challenging because of the increasing number of patients with all forms of diabetes. We aimed to create a deep-learning system to predict the risk of patients with diabetes developing diabetic retinopathy within 2 years. MethodsWe created and validated two versions of a deep-learning system to predict the development of diabetic retinopathy in patients with diabetes who had had teleretinal diabetic retinopathy screening in a primary care setting. The input for the two versions was either a set of three-field or one-field colour fundus photographs. Of the 575 431 eyes in the development set 28 899 had known outcomes, with the remaining 546 532 eyes used to augment the training process via multitask learning. Validation was done on one eye (selected at random) per patient from two datasets: an internal validation (from EyePACS, a teleretinal screening service in the USA) set of 3678 eyes with known outcomes and an external validation (from Thailand) set of 2345 eyes with known outcomes. Findings The three-field deep-learning system had an area under the receiver operating characteristic curve (AUC) of 0•79 (95% CI 0•77-0•81) in the internal validation set. Assessment of the external validation set-which contained only one-field colour fundus photographs-with the one-field deep-learning system gave an AUC of 0•70 (0•67-0•74). In the internal validation set, the AUC of available risk factors was 0•72 (0•68-0•76), which improved to 0•81 (0•77-0•84) after combining the deep-learning system with these risk factors (p<0•0001). In the external validation set, the corresponding AUC improved from 0•62 (0•58-0•66) to 0•71 (0•68-0•75; p<0•0001) following the addition of the deep-learning system to available risk factors.Interpretation The deep-learning systems predicted diabetic retinopathy development using colour fundus photographs, and the systems were independent of and more informative than available risk factors. Such a risk stratification tool might help to optimise screening intervals to reduce costs while improving vision-related outcomes.Funding Google.

Assessing the Performance of Artificial Intelligence Systems for the Screening of Diabetic Retinopathy: A Systematic Review and Meta-Analysis

“…The lack of sustainable diabetic retinopathy screening programs has formed a serious clinical gap for managing the condition (23,29). In order to close this gap, sustainable screening programs must be developed globally so that diabetes mellitus patients can be cared for in accessible primary care settings (30). Additionally, the advancement of risk prediction methods is necessary to ascertain which patients will likely develop visionthreatening diabetic retinopathy (28,30).…”

“…In order to close this gap, sustainable screening programs must be developed globally so that diabetes mellitus patients can be cared for in accessible primary care settings (30). Additionally, the advancement of risk prediction methods is necessary to ascertain which patients will likely develop visionthreatening diabetic retinopathy (28,30). At present, not even ophthalmologists can predict which group of diabetic retinopathy patients are at increased risk of vision loss.…”

“…At present, not even ophthalmologists can predict which group of diabetic retinopathy patients are at increased risk of vision loss. Lastly, improvements in accessibility and treatments for those experiencing vision-loss from diabetic retinopathy is essential (30). Such treatments include acellular capillary laser treatments, anti-VEGF injections, and vitreoretinal surgeries (26).…”

“…The accuracy of such technological developments, as assessed by prior investigations, has surpassed clinically acceptable thresholds of sensitivity (number of true positive assessments over the number of all positive assessments reported) and specificity (number of true negative assessments over the number of all negative assessments reported). With recent advancements in ML and DL, countries across the globe are revisiting the incorporation of AI systems for diabetic retinopathy screening (30,35).…”

Efficient, high-performance semantic segmentation using multi-scale feature extraction

“…Common machine learning datasets, such as ImageNet [ 1 ] with >1 million images, are much larger than their counterparts used in medical studies. Even large recent studies [ 2 , 3 ] use datasets significantly smaller than ImageNet and orders of magnitude smaller than the datasets used to train state-of-the-art language models [ 4 ]. Furthermore, current medical studies often source data from only few institutions, thus preventing the training of representative and unbiased models, suitable for application in a broad variety of patient collectives [ 5 ].…”