Evaluation of an AI System for the Detection of Diabetic Retinopathy from Images Captured with a Handheld Portable Fundus Camera: the MAILOR AI study

Diagnosis of Diabetic Retinopathy in Ethiopia: Before the Deep Learning based Automation

“…Although not publicly available, retinal images have been collected using portable fundus cameras for automated classifiers' performance evaluation (Rogers et al, 2019) and retinal images quality assessment purposes (Wang et al, 2015). To this date, and to the best of the author's knowledge, the CHASE DB1 (Kingston, 2011) is the only publicly available retinal pictures dataset, which have used handheld fundus cameras for data collection.…”

“…Higher FOV and smaller diameter are preferred camera properties. Even though conventional fundus cameras that are currently on the market seem to achieve these specifications, which helps in classifying DR into different stages, portable fundus cameras output images of low quality; hence their use in binary classification only (Rogers et al, 2019;Sosale et al, 2020). Since identification of DR severity scale provides practitioners with the necessary information for treatment, automated fundus images classification systems should provide degree of DR by performing multi-class classification.…”