Fractal Analysis and Convolutional Neural Network Based Diabetic Retinopathy Grade Classification

Abstract

Diabetic retinopathy is a disease caused in eyes of the diabetic patients. The images of retina of the eyes are taken from the MESSIDOR dataset. Features are extracted by considering the fields using fractal dimension. The extracted features are then given as input to the Convolutional Neural Network for training the classifier. The classifier classifies the normal person and the person with diabetes. Then few more features are extracted to grade the diabetic patients in addition to finding normal and abnormal patients. The feature extraction has supported the classifier in yielding better results when compared to the other existing methods for grading the diabetic patients.

 Keywords: diabetic retinopathy, convolutional neural network,fractal dimension, feature, grading