Comparison of early diabetic retinopathy staging in asymptomatic patients between autonomous AI-based screening and human-graded ultra-widefield colour fundus images

Abstract

Introduction Comparison of diabetic retinopathy (DR) severity between autonomous Artificial Intelligence (AI)-based outputs from an FDA-approved screening system and human retina specialists’ gradings from ultra-widefield (UWF) colour images. Methods Asymptomatic diabetics without a previous diagnosis of DR were included in this prospective observational pilot study. Patients were imaged with autonomous AI (IDx-DR, Digital Diagnostics). For each eye, two 45° colour fundus images were analysed by a secure server-based AI algorithm. UWF colour fundus imaging was performed using Optomap (Daytona, Optos). The International Clinical DR severity score was assessed both on a 7-field area projection (7F-mask) according to the early treatment diabetic retinopathy study (ETDRS) and on the total gradable area (UWF full-field) up to the far periphery on UWF images. Results Of 54 patients included (n = 107 eyes), 32 were type 2 diabetics (11 females). Mean BCVA was 0.99 ± 0.25. Autonomous AI diagnosed 16 patients as negative, 28 for moderate DR and 10 for having a vision-threatening disease (severe DR, proliferative DR, diabetic macular oedema). Based on the 7F-mask grading with the eye with the worse grading defining the DR stage 23 patients were negative for DR, 11 showed mild, 19 moderate and 1 severe DR. When UWF full-field was analysed, 20 patients were negative for DR, while the number of mild, moderate and severe DR patients were 12, 21, and 1, respectively. Conclusions The autonomous AI-based DR examination demonstrates sufficient accuracy in diagnosing asymptomatic non-proliferative diabetic patients with referable DR even compared to UWF imaging evaluated by human experts offering a suitable method for DR screening.

Variability of Grading DR Screening Images among Non-Trained Retina Specialists

“…There have been numerous studies looking into the performance of AI DR software against human graders of different specialty and training levels, and AI DR algorithms show significant performance differences [ 11 , 12 ]. Recently Sedova and colleagues compared the performance of IDx-DR with seven-field ETDRS protocol fundus images and ultra-wide-field images graded by two retinal specialists [ 13 ]. They found IDx-DR results to have a 66.6% and 66.7% agreement with seven-field and ultra-wide-field examinations, respectively, with respective kappa scores of 0.40 and 0.38 [ 13 ].…”

“…Recently Sedova and colleagues compared the performance of IDx-DR with seven-field ETDRS protocol fundus images and ultra-wide-field images graded by two retinal specialists [ 13 ]. They found IDx-DR results to have a 66.6% and 66.7% agreement with seven-field and ultra-wide-field examinations, respectively, with respective kappa scores of 0.40 and 0.38 [ 13 ]. Of note, in the current study, both AI and human graders had access to the same information (only standard color macula- and disc-centered fundus images), whereas Sedova et al compared IDx-DR output to graders using more comprehensive imaging studies.…”

Advances in Teleophthalmology Screening for Diabetic Retinopathy

“…[20][21][22] Studies done at other institutions produced similar results, demonstrating the real-world utility of these systems. 84,[88][89][90] For example, Mehra et al reported that the integration of IDx-DR in a primary care clinic DR screening program resulted in 100% sensitivity and 89% specificity for mtmDR. 88 Kim et al combined EyeArt with smartphone-based retinal photography, achieving a lower sensitivity but higher specificity than trained expert graders.…”

Progress of Imaging in Diabetic Retinopathy—From the Past to the Present

“…Comparative studies between non-mydriatic or mydriatic UWF images and ETDRS seven-field standard images were investigated [ 32 , 33 , 34 , 35 , 36 , 37 ]. Automated DR diagnosis in fundus photography, one of the software with artificial intelligent was recently approved by FDA, adopts two-fields photographs [ 38 ], and such cost-effective screening of this disease is another hot topic.…”

“…This system diagnosis with two conventional 45-degree photographs of macular-centered and disc-centered and was reported to offer sufficient quality of detecting referable DR compared to human grading [ 16 ]. Many studies also reported high rate of sensitivity and specificity of autonomous-AI diagnosis system in diagnosing any stage of DR [ 38 , 90 , 91 , 92 , 93 , 94 , 95 ]. Diagnostic sensitivity of the algorithm (2iRetinex) for DR to sight threatening vision was examined, and the sensitivity is nearly identical to diagnosis by physician [ 92 ].…”