Novel Technique for Segmentation and Identification of Diabetic Retinopathy in Retinal Images Using GMM and Compare Accuracy with SVM

Abstract

The main objective of this work is to detect blood vessel segmentation of diabetic retinopathy in low-resolution retinal images using two machine learning algorithms and perform accuracy comparison. Materials and methods: Gaussian Mixture Model (GMM) and Support Vector Machine (SVM) are implemented to detect and segment the blood vessel in retinal images dataset with 40 samples (20 per group). Results: From the MATLAB simulation result, GMM classified the image with better accuracy of 95% compared to SVM classifier accuracy 89%, attained a significant accuracy ratio (p=0.020) in statistical analysis. Conclusion: GMM provides better accuracy compared to SVM classifiers in segmentation and detection.