Analysis of Ophthalmic Features for Glaucoma Detection

The selection of appropriate ophthalmic feature’s extraction technique is a crucial pre-requisite for glaucoma detection. There are many features that can be extracted from an image which has been a long-standing problem. This step is essential during the classification process to isolate relevant information from the image. There are various feature extraction techniques to extract the features and feature selection techniques to select best features among them on the basis of the disease type. This paper proposes the best possible subset of various features that are required to be extracted for glaucoma diagnosis using DRISTHI-GS dataset. In the first step, various features like structural features such as cup-to-disc ratio (CDR) and disc damage likelihood scale (DDLS), textural features such as grey level co-occurrence matrix (GLCM) and Grey level run length matrix (GLRM), statistical features such as mean, variance and skewness, wavelet-based features were extracted from retinal images and in the second step RFE method used for the feature selection. Total sixteen features are extracted from the first step and nine features are selected from the Recursive Feature Elimination (RFE) method. Selected nine features are applied on the different classifiers. Performance of the proposed model tested on K-nearest neighbors (KNN), Random Forest (RF), Naïve Bayes (NB) and Support Vector Machine (SVM) classifiers, which provides 98.6% accuracy on RF, 98% accuracy on SVM, 91 % accuracy on KNN and 96% accuracy on NB

Feature extraction · Feature selection · Glaucoma · Naive Bayes classifier · Random forest · Support vector machine · Computer Science · Digital Imaging for Blood Diseases · Glaucoma and retinal disorders · Retinal Imaging and Analysis · Artificial Intelligence