Novel blood vessel segmentation methods of fundus images for diabetic retinopathy and accuracy comparison using Gmm and Svm classifiers

Aim-The main purpose of this work is to present a novel blood vessel segmentation method of fundus images for diabetic retinopathy using blurred images with better accuracy. Materials & Methods-Two different algorithms; Gaussian Mixture Model (GMM) and Support Vector Machine (SVM) are used to se...

Full description

Saved in:
Bibliographic Details
Main Authors: Reddy, N. Sreekanth, Ramkumar, G.
Format: Conference Proceeding
Language:English
Subjects:
Algorithms
Blood vessels
Classifiers
Datasets
Image segmentation
Probabilistic models
Software
Support vector machines
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Aim-The main purpose of this work is to present a novel blood vessel segmentation method of fundus images for diabetic retinopathy using blurred images with better accuracy. Materials & Methods-Two different algorithms; Gaussian Mixture Model (GMM) and Support Vector Machine (SVM) are used to segment an image from a trained dataset consisting of 40 blurred images. Results-By training these datasets in MATLAB software with add-ons installed GMM obtained 97% accuracy which is far better compared to SVM classifier. Attained significant accuracy of (P=0.005) in SPSS IBM software tool. Conclusion-For the trained dataset of blurred images GMM obtained better accuracy compared to SVM classifiers.
ISSN:0094-243X
1551-7616
DOI:10.1063/5.0177463