Development and Validation of a Deep Learning System for Diabetic Retinopathy and Related Eye Diseases Using Retinal Images From Multiethnic Populations With Diabetes.

Importance: A deep learning system (DLS) is a machine learning technology with potential for screening diabetic retinopathy and related eye diseases. Objective: To evaluate the performance of a DLS in detecting referable diabetic retinopathy, vision-threatening diabetic retinopathy, possible glaucoma, and age-related macular degeneration (AMD) in community and clinic-based multiethnic populations with diabetes. Design, Setting, and Participants: Diagnostic performance of a DLS for diabetic retinopathy and related eye diseases was evaluated using 494 661 retinal images. A DLS was trained for detecting diabetic retinopathy (using 76 370 images), possible glaucoma (125 189 images), and AMD (72 610 images), and performance of DLS was evaluated for detecting diabetic retinopathy (using 112 648 images), possible glaucoma (71 896 images), and AMD (35 948 images). Training of the DLS was completed in May 2016, and validation of the DLS was completed in May 2017 for detection of referable diabetic retinopathy (moderate nonproliferative diabetic retinopathy or worse) and vision-threatening diabetic retinopathy (severe nonproliferative diabetic retinopathy or worse) using a primary validation data set in the Singapore National Diabetic Retinopathy Screening Program and 10 multiethnic cohorts with diabetes. Exposures: Use of a deep learning system. Main Outcomes and Measures: Area under the receiver operating characteristic curve (AUC) and sensitivity and specificity of the DLS with professional graders (retinal specialists, general ophthalmologists, trained graders, or optometrists) as the reference standard. Results: In the primary validation dataset (n = 14 880 patients; 71 896 images; mean [SD] age, 60.2 [2.2] years; 54.6% men), the prevalence of referable diabetic retinopathy was 3.0%; vision-threatening diabetic retinopathy, 0.6%; possible glaucoma, 0.1%; and AMD, 2.5%. The AUC of the DLS for referable diabetic retinopathy was 0.936 (95% CI, 0.925-0.943), sensitivity was 90.5% (95% CI, 87.3%-93.0%), and specificity was 91.6% (95% CI, 91.0%-92.2%). For vision-threatening diabetic retinopathy, AUC was 0.958 (95% CI, 0.956-0.961), sensitivity was 100% (95% CI, 94.1%-100.0%), and specificity was 91.1% (95% CI, 90.7%-91.4%). For possible glaucoma, AUC was 0.942 (95% CI, 0.929-0.954), sensitivity was 96.4% (95% CI, 81.7%-99.9%), and specificity was 87.2% (95% CI, 86.8%-87.5%). For AMD, AUC was 0.931 (95% CI, 0.928-0.935), sensitivity was 93.2% (95% CI, 91.1%-99.8%), and specificity was 88.7% (95% CI, 88.3%-89.0%). For referable diabetic retinopathy in the 10 additional datasets, AUC range was 0.889 to 0.983 (n = 40 752 images). Conclusions and Relevance: In this evaluation of retinal images from multiethnic cohorts of patients with diabetes, the DLS had high sensitivity and specificity for identifying diabetic retinopathy and related eye diseases. 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.

Optical Coherence Tomographic Angiography in Type 2 Diabetes and Diabetic Retinopathy.

IMPORTANCE: Optical coherence tomographic angiography (OCT-A) is able to visualize retinal microvasculature without the need for injection of fluorescein contrast dye. Nevertheless, it is only able to capture a limited view of macula and does not show leakage. OBJECTIVES: To evaluate the retinal microvasculature using OCT-A in patients with type 2 diabetes as well as the association of OCT-A characteristics with diabetic retinopathy (DR) and systemic risk factors. DESIGN, SETTING, AND PARTICIPANTS: A prospective, observational study was conducted from January 1 to June 30, 2016, at medical retina clinics at the Singapore National Eye Center among 50 patients with type 2 diabetes with and without DR (n = 100 eyes). We examined the retinal microvasculature with swept-source OCT-A and a semiautomated software to measure the capillary density index (CDI) and fractal dimension (FD) at the superficial vascular plexus (SVP) and deep retinal vascular plexus (DVP). We collected data on histories of patients' glycated hemoglobin A1c, hypertension, hyperlipidemia, smoking, and renal impairment. MAIN OUTCOMES AND MEASURES: The CDI and FD at the SVP and DVP for each severity level of DR and the association of systemic risk factors vs the CDI and FD. RESULTS: The mean (SD) glycated hemoglobin A1c of the 50 patients (26 men and 24 women; 35 Chinese; mean [SD] age, 59.5 [8.9] years) was 7.9% (1.7%). The mean (SD) CDI at the SVP decreased from 0.358 (0.017) in patients with no DR to 0.338 (0.012) in patients with proliferative DR (P < .001) and at the DVP decreased in patients with no DR from 0.361 (0.019) to 0.345 (0.020) in patients with proliferative DR (P = .04). The mean (SD) FD at the SVP increased from 1.53 (0.05) in patients with no DR to 1.60 (0.05) in patients with proliferative DR (P < .01) and at the DVP increased from 1.55 (0.06) in patients with no DR to 1.61 (0.05) in patients with proliferative DR (P = .02). For systemic risk factors, hyperlipidemia (odds ratio [OR], 9.82; 95% CI, 6.92-11.23; P < .001), smoking (OR, 10.90; 95% CI, 8.23-12.34; P < .001), and renal impairment (OR, 3.72; 95% CI, 1.80-4.81; P = .05) were associated with reduced CDI, while increased glycated hemoglobin A1c (≥8%) (OR, 8.77; 95% CI, 5.23-10.81; P < .01) and renal impairment (OR, 10.30; 95% CI, 8.21-11.91; P < .001) were associated with increased FD. CONCLUSIONS AND RELEVANCE: Optical coherence tomographic angiography is a novel imaging modality to quantify the retinal capillary microvasculature in patients with diabetes. It can be potentially used in interventional trials to study the effect of systemic risk factors on the microvasculature that was previously not accessible in a noninvasive manner. The relevance of these findings relative to visual acuity, however, remains largely unknown at this time.