Abstract: Diabetic retinopathy (DR) diagnosis by color fundus images needs trained practitioners to recognise the existence and significance of many minor abnormalities, which, combined with a complex grading system, makes this a challenging and time-consuming procedure. In this study, we present a CNN approach to detecting DR from digital fundus images and properly grading its severity. We create a network with CNN, Densnet 121 architecture, and data augmentation that can identify the intricate elements involved in the classification task, such as micro-aneurysms, exudate, and hemorrhages on the retina, and then deliver a diagnosis automatically and without user input.We train this network on the publically accessible Kaggle dataset with a high-end graphics processor unit (GPU) and achieve outstanding results, particularly for a high-level classification task. Our proposed CNN achieves a sensitivity of 95% on the data set of 80,000 photos used.
Keywords: Diabetic retinopathy (DR), Colour fundus images, Convolutional neural network (CNN), graphics processor unit (GPU), Kaggle dataset.
|
DOI:
10.17148/IJARCCE.2023.124193
