This article is more than 5 years old. Information may no longer be current.

Deep learning system identifies diabetic retinopathy, related eye diseases

A machine-learning technology known as a deep learning system showed high sensitivity and specificity for recognizing diabetic retinopathy and related eye diseases among patients with diabetes, according to data published in JAMA.

“Screening for diabetic retinopathy, coupled with timely referral and treatment, is a universally accepted strategy for blindness prevention,” Daniel Shu Wei Ting, MD, PhD, from Singapore Eye Research Institute at Singapore National Eye Center and Duke-NUS Medical School at National University of Singapore, and colleagues wrote. “However, programs for screening diabetic retinopathy are challenged by issues related to implementation, availability of human assessors, and long-term financial sustainability.”

Researchers evaluated the performance of a deep learning system (DLS) in identifying diabetic retinopathy, vision-threating diabetic retinopathy, possible glaucoma and age-related macular degeneration (AMD) using 494,661 retinal images from patients of different races and ethnicities. They trained and validated a DLS for detection of referable diabetic retinopathy and vision-threatening diabetic retinopathy using a primary validation data set in the Singapore National Diabetic Retinopathy Screening Program and 10 multiethnic cohorts with diabetes.

Out of 14,880 patients and 71,896 images in the primary validation dataset, the prevalence of referable diabetic retinopathy was 3%, vision-threatening diabetic retinopathy was 0.6%, possible glaucoma was 0.1% and AMD was 2.5%. Analysis showed that the area under the receiver operating characteristic curve of the DLS for diabetic retinopathy was 0.936 (95% CI, 0.93-0.94), sensitivity was 90.5% (95% CI, 87.3-93) and specificity was 91.6% (95% CI, 91-92.2). Sensitivity was 100% and specificity was 91.1% for vision-threatening diabetic retinopathy; 96.4% and 87.2% for possible glaucoma; and 93.2% and 88.7% for AMD. In the 10 additional datasets with 40,752 images, AUC range was 0.889 to 0.983 for referable diabetic retinopathy.

“In this evaluation of nearly half a million of images from multiethnic community, population-based and clinical datasets, the DLS had high sensitivity and specificity for identifying referable diabetic retinopathy and vision-threatening diabetic retinopathy, as well as for identifying related eye diseases, including referable possible glaucoma and referable age-related macular degeneration,” Ting and colleagues wrote. “Further research is necessary to evaluate the applicability of the DLS in health care settings and the utility of the DLS to improve vision outcomes.” – by Savannah Demko

Disclosures: Ting is a coinventor on a patent for the deep learning system used in this study; potential conflicts of interests are managed according to institutional policies of the Singapore Health System and the National University of Singapore. Please see the study for a list of all other authors’ relevant financial disclosures.