Purchase this article with an account.
Maureen G Maguire, Ebenezer Daniel, Meindert Niemeijer, Maxwell Pistilli, James C Folk, Michael David Abramoff; Identifying Diabetic Eye Disease: Comparison of Clinical Examination by Ophthalmologists to Automated Detection from Retinal Color Images. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):2014.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To compare the sensitivity and specificity of the IDx-DR photographic device to clinical examination by ophthalmologists to identify diabetic eye disease (DED).
Patients (n=528) were recruited through five ophthalmologic centers at seven sites [general ophthalmology or retinal specialist]. Eligibility criteria included a diagnosis of diabetes; age ≥18 years; no history of other ocular diseases affecting the retina or optic nerve. Ophthalmologists performed a dilated fundus examination including stereoscopic biomicroscopy. Two, 45°, color photographs of each eye were taken with a Topcon NW 200 non-mydriatic camera. Images were processed by the IDx-DR v1.0 device and graded at a central reading center by readers masked to results of the clinical examination and the device. DED was defined as more than mild nonproliferative diabetic retinopathy as detected by the Reading Center (>= ETDRS level 35) or macular edema within one disc diameter from the fovea as detected by either the Reading Center or the ophthalmologist; if images were not gradeable by the Reading Center (37 [3.5%] eyes), the findings of the ophthalmologist were used for classification of DED. Patients were classified based on the worse gradings of the 2 eyes.
Prevalence of DED was 38.1% (201/528). Sensitivity was 86.1% (173/201) for the device compared to 53.7% (108/201) for clinical examination; the difference was 32.3% (95% CI: 24.9% to 39.3%). Specificity was 68.5% (224/327) for the device compared to 99.1% (324/327) for clinical examination; the difference was -30.6% (95% CI: -35.8% to -25.7%). When DED was restricted to potentially treatable disease (severe nonproliferative or proliferative retinopathy or macular edema), sensitivity was 96.8% (61/63) for the device compared to 95.2% (60/63) for clinical examination; the difference was 1.6% (95% CI: -6.8% to 10.3%). Differences were similar for different racial and ethnic groups.
The IDx-DR device detected a higher proportion of people with DED than clinical examination. The device was comparable to clinical examination in detecting potentially treatable disease. Combined with fundus photography, devices such as the the IDx-DR have the potential to improve the accessibility of screening in people with diabetes.
This PDF is available to Subscribers Only