Deep learning approach for retinopathy identification in combined clinical datasets

Abstract

A technique for automatically identifying different stages of diabetic retinopathy (DR) based on patient fundus photographs is presented in this paper. The "Dataset of fundus images for the study of diabetic retinopathy" version v3 CastilloBenitez21, a relatively young and underutilized image collection, served as the foundation for the system's construction. Clinical fundus photos obtained from Wroclaw Medical University's Department of Nephrology were used to augment the main dataset. Convolutional neural networks (CNNs) that were pre-trained on ImageNet data served as the foundation for the development of the diagnostic system. With a global accuracy of 83.15%, the CNN classifier—which is based on VGG16 with transfer learning—proved to be successful. An accuracy of 89.7%, a sensitivity of 94.9%, a specificity of 88.3%, and a Matthews Correlation Coefficient of 0.7665 were obtained from the assessment of discriminating between the non-DR and DR states.