Diabetic Retinopathy Detection and Classification using Hybrid Multiclass SVM classifier and Deeplearning techniques

Abstract

In the medical field, earlier disease detection is the effective treatment. The diabetic retinopathy causes the lesion on retina this leads to vision loss. Early detection of diabetic retinopathy reduces the vision loss. Detection and classification of diabetic retinopathy can be performed by using AI approaches like machine learning and deep learning techniques. Here the diabetic feature analysis determines the various stages of retinopathy infection caused by unbalanced diabetics. Conventional approaches use the deep learning with convolutional neural network model for retinopathy image classification. Large dataset processing difficulty, complex training and computation time are the major drawbacks of existing work. In this proposed research work, the multi class Support Vector Machine learning technique with lesion and vessel analysis of retinopathy images are performed and the Deep learning technique is used to detect and classify the diabetic retinopathy images. Multi layer performance of deep learning technique is performed to classify the normal and abnormal diabetic retinopathy image data. The objective of this research is to develop a multiclass algorithm by using hybrid machine learning and deep learning techniques for prediction and classification of diabetic retinopathy. The proposed algorithm has been applied on open source diabetic retinopathy image data set. Experimental results are compared with conventional techniques and the proposed work yields better results with Accuracy of 93.33, Sensitivity of 96.71 and Specificity of 99.22. From the results, it was observed that proposed algorithm produced relatively better results than existing conventional algorithms. So, it was decided that the proposed algorithm can be used as diagnosis tool for early detection of diabetic retinopathy.