Blood vessel segmentation and diabetic retinopathy recognition: an intelligent approach

This paper intends to develop an enhanced framework for Diabetic retinopathy (DR) recognition. Accordingly, here the proposed model is performed under two phases, first is the blood vessel segmentation and second one is the DR recognition. In vessel segmentation, two thresholded binary images are at...

Full description

Saved in:
Bibliographic Details
Published inComputer methods in biomechanics and biomedical engineering. Vol. 8; no. 2; pp. 169 - 181
Main Authors Nair, Arun T., Muthuvel, K.
Format Journal Article
LanguageEnglish
Published Taylor & Francis 03.03.2020
Subjects
Classification
Cuckoo Search Algorithm
Diabetic Retinopathy
Feature Extraction
Retinal Fundus Image
Segmentation
Step size
Online AccessGet full text

Cover

More Information
Summary:This paper intends to develop an enhanced framework for Diabetic retinopathy (DR) recognition. Accordingly, here the proposed model is performed under two phases, first is the blood vessel segmentation and second one is the DR recognition. In vessel segmentation, two thresholded binary images are attained using High-Pass Filtering (HPF) and tophat by reconstruction of the red portions in the green plane image. The areas that are found similar to two binary images are extracted as the major vessels, and the leftover pixels in both binary images are merged to form a vessel sub-image that is subjected to a Gaussian Mixture Model (GMM) classification. Consequently, the entire pixels in the sub-image, which are classified as vessels are merged with the major vessels to get the segmented vasculature. Further, the GLCM and GLRM features are extracted from the segmented blood vessel, which are then classified using Neural Network (NN). To improve the accuracy, training is done by Modified Cuckoo Search with New Step size (MCS-NS) algorithm, such that the error between predicted and actual output should be minimal.
ISSN:2168-1163
2168-1171
DOI:10.1080/21681163.2019.1647459