Frontiers in digital health · 2026
Autonomous AI-driven point-of-care screening for diabetic retinopathy compared to reading center multi-expert clinical review: results from three prospective controlled pivotal validation studies with AEYE-DS in over 1,200 patients.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Dvey-Aharon Zack, Livne Moshe, Margalit Dan, Wohl Amit, Haupt Adi, Aviv Rachelle, Stein Ahava, Rachelson Adi, Nussdorf Jonathan, Ianchulev Tsontcho
Аннотация
PURPOSE: Diabetic retinopathy (DR) remains the leading cause of blindness among working-age adults and requires regular screening to detect progression among the growing global diabetic population. This study evaluated the performance of AEYE-DS, an autonomous artificial intelligence (AI) system designed for high-throughput, point-of-care analysis of retinal images, in detecting more-than-mild diabetic retinopathy (mtmDR) during routine screening of patients with diabetes who had not previously been diagnosed with DR.
PRINCIPAL RESULTS: AEYE-DS was tested across three prospective clinical studies using two FDA-cleared non-mydriatic retinal cameras: the handheld Aurora and the desktop Topcon NW400. The algorithm autonomously analyzed retinal images and determined mtmDR presence. Diagnostic outcomes were compared to a reference standard based on the Early Treatment for Diabetic Retinopathy Study (ETDRS) severity grading performed by multi-expert review at an independent reading center. Sensitivity and specificity were 93% and 91% in AEYE-1 (95% CI: 83%-97% and 88%-94%), 92% and 94% in AEYE-2 (95% CI: 79%-97% and 90%-96%), and 93% and 89% in AEYE-3 (95% CI: 80%-97% and 85%-92%). Imageability was >99% in all studies. Intra-operator repeatability exceeded 99% for both devices. Between-operator reproducibility was 98% for the desktop camera and 95% for the handheld device, while between-device reproducibility reached 99% and 97%, respectively.
CONCLUSIONS: AEYE-DS demonstrated high diagnostic accuracy, imageability, reliability, and reproducibility across different operators and devices in non-mydriatic settings. Findings support autonomous AI system use for scalable, point-of-care DR screening, potentially expanding access, streamlining workflows, and reducing the global burden of diabetic eye disease.
PRINCIPAL RESULTS: AEYE-DS was tested across three prospective clinical studies using two FDA-cleared non-mydriatic retinal cameras: the handheld Aurora and the desktop Topcon NW400. The algorithm autonomously analyzed retinal images and determined mtmDR presence. Diagnostic outcomes were compared to a reference standard based on the Early Treatment for Diabetic Retinopathy Study (ETDRS) severity grading performed by multi-expert review at an independent reading center. Sensitivity and specificity were 93% and 91% in AEYE-1 (95% CI: 83%-97% and 88%-94%), 92% and 94% in AEYE-2 (95% CI: 79%-97% and 90%-96%), and 93% and 89% in AEYE-3 (95% CI: 80%-97% and 85%-92%). Imageability was >99% in all studies. Intra-operator repeatability exceeded 99% for both devices. Between-operator reproducibility was 98% for the desktop camera and 95% for the handheld device, while between-device reproducibility reached 99% and 97%, respectively.
CONCLUSIONS: AEYE-DS demonstrated high diagnostic accuracy, imageability, reliability, and reproducibility across different operators and devices in non-mydriatic settings. Findings support autonomous AI system use for scalable, point-of-care DR screening, potentially expanding access, streamlining workflows, and reducing the global burden of diabetic eye disease.