Bright Lesion Detection in Retinal Fundus Images for Diabetic Retinopathy Detection Using Machine Learning Approach

Abstract

The objective of the project is the detection and classification of bright exudates lesions in  fundus images of eyes depending on the number of bright lesions present. Diabetic Retinopathy (DR) is a condition caused by diabetes which leads to microvascular difficulties and microaneurysm development. These are the primary underlying indicators of DR in the early stages. The proposed model helps make a prediction whether Diabetic Retinopathy can be classified on the basis of the amount of exudates present in the retinal fundus images. The model will also help in predicting whether there are less number or high number of exudates present. For the detection of exudates, several steps have been followed including resizing of images, extracting the blue channels, feature extraction using Local Binary patters (LBP), and classification using SVM. IDRID dataset has been used for performing the same task.