A Transfer Learning Approach For Retinal Disease Classification

Diagnosing retinal disease in an earlier stage using fundus images is a complicated, error-prone, time-consuming, and challenging process. Therefore, a computerized retinal disease detection system with advances in technology is required to identify various eye disorders in fundus images. The proposed work creates a dataset that comprises of fundus images with some of the retinal diseases such as Diabetic retinopathy (DR), Age-related Macular Degeneration (AMD), Glaucoma (GA), Hemorrhages (HG), Epiretinal membrane (EM), and No disease (NOD) and it is named as “Multi Disease Dataset” (MUD). To identify the disease in retinal images, the created dataset is evaluated using different transfer learning techniques. Compared to state-of-the-art methods, experimental analysis demonstrates that the proposed method achieves an accuracy of 89.11% using Inceptionv3 on the MUD dataset and is capable of detecting five diseases.