A portable retina fundus photos dataset for clinical, demographic, and diabetic retinopathy prediction

AI Generated Summary*

mBRSET is a public dataset of 5,164 fundus photographs from 1,291 diabetic patients, taken after pupil dilation with a handheld camera paired with a smartphone at a November 2022 diabetes campaign in Itabuna, Bahia, Brazil. Metadata cover diabetes duration, medications, comorbidities, age, sex, education and insurance; 65% of patients were female and 92.3% lacked health insurance. Two retina specialists graded images by ICDR criteria (weighted kappa 0.863 for retinopathy, 0.618 for macular edema). Among 4,885 graded images, 76.8% showed no retinopathy, 4.3% proliferative disease and 8.7% macular edema. Best benchmark F1 scores were 87.4 binary retinopathy, 80.3 three-class, 83.1 edema, 84.4 sex, 72.3 education, 76.1 insurance; one model's retinopathy F1 fell slightly versus a tabletop dataset.

*This summary was generated by AI and is published unedited. Oku does not alter these summaries. It may contain errors or omissions and is provided for general informational purposes only. Oku does not guarantee its accuracy, completeness, or reliability. For authoritative information, please refer to the original, peer-reviewed article.