Reliable Densely Connected Network with Machine Learning based Diabetic Retinopathy Grading Approach

Lately, diabetic retinopathy (DR) is deliberated to be the major cause of blindness in today's world. There are various reasons for loss of eyesight but DR has greatly contributed to blindness. The main challenge of DR is that it is incurable at an early stage, therefore basic analysis is of great significance. Consequently, it is mentioned as a classic difficulty in medical schemes due to minimum number of medicinal experts and facilities and massive patients. Accordingly, it resulted in deploying automated diagnosing methodology that is very beneficial for the analysis of DR. With this aspect, this article develops a reliable Densely Connected Network with Machine Learning based Diabetic Retinopathy Grading Approach, called RDCN-MLDR. The presented RDCN-MLDR model follows two major stages namely feature extraction and DR classification. Initially, the DenseNet model processes the input fundus images. In addition, squirrel search algorithm (SSA) is applied for the hyperparameter tuning of the DenseNet model. As soon as the features are created, they are fed to the LR module for proper identification of DR. The performance assessment of the RDCN-MLDR model is tested using open access dataset from Kaggle dataset and it produces the 97.8% accuracy. The attained outcomes highlighted that the RDCN-MLDR model can accomplish enhanced DR classification performance.