Explainable Deep Learning for Lesion-Level Detection of Diabetic Retinopathy: A Segmentation Approach Using Fundus Images Graded as Mild-to-Moderate Nonproliferative Diabetic Retinopathy

Data Availability

This study used publicly available datasets from Kaggle competitions: the APTOS 2019 Blindness Detection dataset (https://www.kaggle.com/competitions/aptos2019-blindness-detection) and the Diabetic Retinopathy Detection dataset provided by EyePACS (https://www.kaggle.com/competitions/diabetic-retinopathy-detection). Both datasets consist of de-identified fundus images. No additional data were generated or analyzed during the current study.

https://www.kaggle.com/competitions/diabetic-retinopathy-detection

https://www.kaggle.com/competitions/aptos2019-blindness-detection