An Integrated Fusion based Feature Extraction and Classification Model for Diabetic Retinopathy Diagnosis

摘要

Diabetic retinopathy (DR) becomes widespread among diabetic patient and cause vision loss in maximum number of people. The DR can be avoidance by controlling blood glucose level and proper prescription. As the DR involves different classes and complexity levels, it is hard to detect and diagnose DR. This paper introduces a new integrated fusion based feature extraction and classification model for DR diagnosis. The projected model involves a fusion model using gray level co-occurrence matrix and VGG-19 called FM-GLCM-VGG19 method. Initially, the input images undergo format conversion and then watershed based segmentation process gets executed. Then, the fusion model takes place to extract the useful set of features and finally softmax based classification process is carried out. The validation of the FM-GLCM-VGG19 Method takes place on Kaggle dataset and the simulation results revealed the betterment of the FM-GLCM-VGG19 Method by obtaining higher accuracy of 71.30%, sensitivity of 50.43% and specificity of 80.19% respectively.