Abstract:Retinal image analysis relies on the effectiveness of computational techniques to discriminate various abnormalities in the eye like diabetic retinopathy, macular degeneration and glaucoma. The onset of the disease is often unnoticed in case of glaucoma, the effect of which is felt only at a later stage. Diagnosis of such degenerative diseases warrants early diagnosis and treatment. In this work, performance of statistical and textural features in retinal vessel segmentation is evaluated through classifiers like extreme learning machine, support vector machine and Random Forest. The fundus images are initially preprocessed for any noise reduction, image enhancement and contrast adjustment. The two-dimensional Gabor Wavelets and Partition Clustering is employed on the preprocessed image to extract the blood vessels. Finally, the combined hybrid features comprising statistical textural, intensity and vessel morphological features, extracted from the image, are used to detect glaucomatous abnormality through the classifiers. A crisp decision can be taken depending on the classifying rates of the classifiers. Public databases RIM-ONE and high-resolution fundus and local datasets are used for evaluation with threefold cross validation. The evaluation is based on performance metrics through accuracy, sensitivity and specificity. The evaluation of hybrid features obtained an overall accuracy of 97% when tested using classifiers. The support vector machine classifier is able to achieve an accuracy of 93.33% on high-resolution fundus, 93.8% on RIM-ONE dataset and 95.3% on local dataset. For extreme learning machine classifier, the accuracy is 95.1% on high-resolution fundus, 97.8% on RIM-ONE and 96.8% on local dataset. An accuracy of 94.5% on high-resolution fundus 92.5% on RIM-ONE and 94.2% on local dataset is obtained for the random forest classifier. Validation of the experiment results indicate that the hybrid features can be deployed in supervised classifiers to discriminate retinal abnormalities effectively.

A Gaussian Scale Space Approach For Exudates Detection, Classification And Severity Prediction

An Artificial-Intelligence- and Telemedicine-Based Screening Tool to Identify Glaucoma Suspects from Color Fundus Imaging

Automated Detection and Segmentation of Exudates for the Screening of Background Retinopathy

Segmentation of exudates in fundus images applying color mathematical morphology

Machine Learning for Diabetic Retinopathy: Gaussian SVM Model for Accurate Analysis in Retinal Fundus Images

Discrete ripplet-II transform feature extraction and metaheuristic-optimized feature selection for enhanced glaucoma detection in fundus images using least square-support vector machine

Diagnosis of glaucoma from retinal fundus images using disc localization and sequential DNN model

A Location-to-segmentation Strategy for Automatic Exudate Segmentation in Colour Retinal Fundus Images

Analysis of hybrid statistical textural and intensity features to discriminate retinal abnormalities through classifiers

Fast and Robust Exudate Detection in Retinal Fundus Images Using Extreme Learning Machine Autoencoders and Modified KAZE Features

Investigations of severity level measurements for diabetic macular oedema using machine learning algorithms

Enhanced diabetic retinopathy detection and exudates segmentation using deep learning: A promising approach for early disease diagnosis

Sparse Dissimilarity-Constrained Coding for Glaucoma Screening

Retinal imaging based glaucoma detection using modified pelican optimization based extreme learning machine

A significant progress of computer aided diagnosis system using automated glaucoma screening

Fully Convolutional Neural Network for Semantic Segmentation of Anatomical Structure and Pathologies in Colour Fundus Images Associated with Diabetic Retinopathy

Three level automatic segmentation of optic disc using LAB color space contours and morphological operation

Effective recognition of glaucoma using SIFT and RFSO classifier

An attention enriched encoder–decoder architecture with CLSTM and RES unit for segmenting exudate in retinal images

Hemorrhage Detection in Fundus Image Based on 2D Gaussian Fitting and Human Visual Characteristics