Authors : Amol Prataprao Bhatkar and G. U.Kharat
Page Nos : 74-83
Description :
Diabetic retinopathy is becoming the major cause of blindness in today’s world. Early detection of diabetic retinopathy is very essential to avoid further evidences. Retinal images are digital images of the eye fundus is sensitive and specific for early signs of diabetic retinopathy. An automated image analysis method to detect early signs of diabetic retinopathy is greatly desired for testing. This paper focuses on Generalized Feed Forward Neural Network (GFFNN) method to detect diabetic retinopathy in retinal images. In this paper the authors used GFFNN classifier to classify the retinal images into normal and abnormal. Discrete Cosine Transform (DCT), Fast Fourier Transform (FFT), Singular Value decomposition (SVD) with 9 different statistical parameters form three different feature vectors. These feature vectors are used to train GFFNN. The % classification accuracy is 95.45%, 100% and 95.83% for DCT, FFT and SVD feature vectors designed GFFNN classifier.
Keywords: Generalized Feed Forward neural network (GFFNN), Retinal images database DIARETDB0, Discrete Cosine Transform (DCT), Fast Fourier Transform (FFT), Singular Value decomposition (SVD).
