An Automatic Diabetic Retinopathy Detection Using Artificial Neural Network
No Thumbnail Available
Date
2018-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
AAU
Abstract
Diabetic retinopathy is a disease that affects the eye of a diabetic patient. Diabetics being
the abnormal level of glucose in the human blood, it is a cause for several types of organ
disorders in the body. One of them being the eye. Diabetic retinopathy is caused when the blood
vessels in the eye start liking blood in to the surface of the eye. Detection of diabetic retinopathy
at different stages of the disease is an invaluable input for physicians in the medical field.
In this papera research has been made for detection of diabetic retinopathy exudates by
using an artificial neural network. The process of image detection starts from image preprocessing,
next is image segmentation then feature selection and finally classification. In the
image pre-processing a median filtering and adaptive histogram equalization is used. The image
blood vessels and Optic disk are then segmented out using morphological analysis.
Accuracy of segmentation of the optic disk is 90%, 87% and 90 % for exudates
containing images, normal images and all images that contain all features of diabetic retinopathy
respectively.
The gray level co-occurrencematrix of the image is then calculated to evaluate the textual
features. Fourteen features are then used to feed the artificial neural network. Two fundus image
data sets(STARE 402 images, and DIABETRETDB1 with 130 images) were obtained from two
medical universities have been used for the detection process. Using the proposed methodology,
a detection performance of 84% sensitivity and 63% specificity is obtained.The implementation
is done by using mathlab 2015.
Description
Keywords
Diabetic Retinopathy Detection, Artificial Neural Network