Diabetic Retinopathy Recognition Using CNN

In this study, diabetic retinopathy has been taken into account when analyzing fundus images. An automated knowledge model is proposed to identify the key antecedents of Diabetic Retinopathy (DR) in this paper. A Convolutional Neural Network (CNN) was initially used to train the proposed model, but after testing it with a CPU-trained neural network, it shows the lowest accuracy because it has only one hidden layer, while deep learning outperforms NN. The weight of the patient's eye will be determined by the model by calculating the weights. By integrating a large selection of well-known classification algorithms into one sophisticated diagnostic model, an ensemble-based learning strategy was evaluated. The proposed framework achieved the highest accuracy rates among all other common classification algorithms. Identifying the threshold for each feature class is the most challenging part of this study. In order to identify the target class thresholds, we used a weighted fuzzy C-means algorithm. Using the model, diabetic retinopathy images can be classified according to their severity.