Abstract
Abstract: :
Purpose: To compare the performance of film with digital photography for determining the presence and severity of diabetic retinopathy (DR) and its component lesions. Methods: Seventy–two eyes of 36 subjects with a history of diabetes from a clinic population were photographed. Photography included two 45° nonstereoscopic color digital images centered on the macula and optic disc taken through aphotically (DA) and pharmacologically (DP) dilated pupils with a 6.3 megapixel camera, as well as the Diabetic Retinopathy Study seven standard 30° stereoscopic color fields taken through pharmacologically dilated pupils onto film (F). Each eye was graded as to the severity of DR on a modified Early Treatment Diabetic Retinopathy Study scale and for the presence of hemorrhages and microaneurysms, hard exudates, soft exudates, intraretinal microvascular abnormalities, venous beading, new vessels on the disc, new vessels elsewhere, fibrous proliferations, preretinal/vitreous hemorrhage, and macular edema. The percent of exact agreement and the kappa statistic were used to assess comparability between digital and film grading. Results: When categorizing DR into none (level 10), mild (level 21–31), moderate (level 41–51), and proliferative (level ≥ 60), the exact agreement between the F and DP grading was 83% with a kappa (standard error (SE)) of 0.76(0.06). The corresponding agreement between the F and DA grading was 71% with a kappa of 0.61(0.07). Exact agreements varied from 69% to 100% for specific retinopathy lesions for F and DP gradings and from 67% to 97% for F and DA gradings. In addition, where disagreements were present, lesions tended to be detected more frequently in film compared to digital images. We estimated that 34% of the bias was due to the smaller area of retina included in the digital images. Conclusions: Grading of two nonstereoscopic digital fields (DA or DP) results in lower sensitivity in detecting retinopathy lesions and assigning an overall DR severity compared to grading of seven standard stereoscopic film fields. However, grades from DP images were more like those from film compared to grades from DA images. Because of increased sensitivity for detecting DR, when feasible, pharmacologic dilation should be used when imaging the retina in epidemiologic studies.
Keywords: clinical (human) or epidemiologic studies: systems/equipment/techniques • diabetic retinopathy