Feature Extraction Process on Early Diabetic Retinopathy Identification Process

Abstract

The clinical procedures available nowadays are mostly useful in preventing many eye related issues in human being. Various studies based on clinical reports are very clear to make a statement that, patients with diabetes has highest risk to Diabetic Retinopathy (DP). Almost 80 to 85% of patients suffering from Diabetics has highest probabilistic chance of getting Diabetic Retinopathy (DP).  The collected retinal fundus images from eye are mostly used for detecting the severity level of infections and also used for analyzing the diabetic retinopathy effects. Analyzing retinal fundus images can be useful in finding different vision problems such as Strabismus, Glaucoma, Cataract, Diabetic Retinopathy, Amblyopia, Refractive Errors and Macular Degeneration. The collected retinal images from eye may has many irrelevant information, which is not needed for Diabetic Retinopathy identification process. The irrelevancy found on the collected images can affect the proper working of learning algorithm and can create problem in execution. The stage followed in pre-processing removes all the irrelevancies present in the collected retinal images and few techniques were followed for enhancing the quality of the processed retinal images. This research article is very specific in explaining the steps followed after the pre-processing stage such as feature extraction process on retinal images. The retinal fundus images collected from Kaggle are preprocessed and selected for feature extraction process. The feature extracted from database is tested and evaluated by considering mean and standard deviation value. The class difference for the collected retinal fundus images are divided into early stage, normal and severe.