ASRS Annual Meeting 2022

On the 1st stage of a social project aimed at detecting and treating diabetic retinopathy (DR) in the hinterlands of Brazil, we employed a previously validated artificial intelligence (AI) system to screen patients with a portable retinal camera. To report on a proof of concept of the clinical use of a portable handheld retinal camera with an embedded AI platform, and a remote, real-time, retina specialist confirmation for the screening of DR in an underserved rural area.

Retrospective analysis of a social project developed in partnership with Retina Global; All participants signed an informed consent form for the overall screening process and usage of their images/data as a requirement for the social project; Setting: municipality of Canindé de São Francisco (total population: 30,000), in the hinterlands of Northeastern Brazil; Population: the whole municipality’s diabetic population (n=740), according the local registry database, was invited by various advertising strategies; A portable handheld retinal camera, with a previously validated, highly sensitive, embedded AI system designed to detect any retinal changes, was used; Macula-centered and disc-centered images were obtained; Immediate and automatic push notifications were remotely sent out to four retina specialists whenever significant abnormal findings were detected by AI; Physicians would classify images as referable or non-referable in real time; Referral criteria were: more than mild DR, glaucoma or cataract suspicion, and those with poor image quality; All exams considered normal by the AI system were later reviewed to check up for false negatives; All referred patients were scheduled for a complete ophthalmic work-up and subsequent treatment.

400 patients were screened, accounting for 54% of the known diabetic population; The AI screening indicated that 111 individuals met the referral criteria: 57 with more than mild DR or poor-quality images preventing DR classification; 45 with suspected cataract; 5 with suspected glaucoma; All altered exams were checked by a retina specialist in real time; Subject was informed of his/her status instantaneously by the technician after remote physician feedback; After a new exam by a retina specialist (in-person), 30 out of 57 with more than mild DR or poor-quality images were treated for diabetic macular edema or proliferative DR, while 6 required cataract surgery, and 2 had other sources of media opacities; Retina specialist review of the non-referable cohort confirmed there were no false negatives.

Proof of concept: portable device + AI + real-time medical validation; Only 25% of the individuals needed to be referred for review, thus saving time for patients and physicians; Immediate validated feedback - higher acceptance by the screened population; Technological advances will likely play a major role in the prevention of blindness from DR, especially in scenarios of social inequalities.