Locust based genetic classifier for the diagnosis of diabetic retinopathy

Abstract

Due to the ongoing advancement in detection of critical diseases, there is a need in revamping the accurate diagnosis of diabetic retinopathy (DR). In this current study, locust based genetic classifier plays a crucial role in early screening of DR and to determine the exact location of the affected region of retina. Initially, preprocessing is performed to remove the obnoxious information such as noise present in the image and helps to transform the RGB format to gray scale image. It is done by applying wiener filter technique. After removing the obnoxious information, exudate segmentation is performed. After splitting out the image into samples, feature extraction is applied by Gabor based region covariance matrix. It helps to reduce the feature in the DIARETDB1 dataset by obtaining a new feature. After obtaining the feature, whale optimization is performed to pick out the best features such as mean, exudate area, optic distance and standard deviation and finally locust based genetic classifier is used to analogize between trained and test set data and it provides infallible information to the ophthalmologist to provide timely treatments. Comparative analysis is performed. It reveals the significant performance of the current approach over other existing SVM and CNN classifier. The results obtained from the current study shows a promising future and it achieves an accuracy of 98.9%.