Combining CNNs for the Detection of Diabetic Retinopathy

Diabetes is a major public health issue that affects approximately forty million individuals in the United States alone. A common side effect of diabetes is vision loss and blindness caused by diabetic retinopathy (DR). The goal of this research is to introduce a robust deep learning approach for the early detection of DR from retinal images into five categories namely No DR, Mild DR, Moderate DR, Severe DR, Non-Proliferative DR. Here, we propose a fusion of CNNs with an optimal weighting scheme to improve classification accuracy. The dataset used is Kaggle APTOS 2019 and for cross dataset validation we used IDRiD dataset. The proposed weighted twin CNN algorithm is implemented using a pair of pre-trained deep networks namely the DenseNet-169 and the InceptionV3. Such a hybrid combination provided a robust and an optimized architecture. A total of 98.43% sensitivity and 88.78% specificity are recorded with a Kappa score and accuracy of 95.8% and 94.3%. Our research has achieved a significant 11.90% improvement as compared to state of the art, showcasing remarkable performance in this field.