Glaucoma Detection Using Image Channels and Discrete Wavelet Transform

Abstract

Glaucoma is a critical retinal disorder due to high intraocular pressure (IOP) within the eye. It causes irreversible damage of the optical nerve head (ONH). The available glaucoma detection methods using decomposition techniques with gray images or only green channel images are less accurate. In this paper, a more accurate method for glaucoma detection using image channels (ICs) and discrete wavelet transform (DWT) from fundus images is proposed. Firstly, input images are resized then red channel (RC), green channel (GC), blue channel (BC), and gray scale (Gs) images are extracted. Secondly, these four types of images are enhanced and decomposed separately into sub band images (SBIs) using second level (SL) DWT. Thirdly, most useful features are extracted from each of DWT SBIs. Fourthly, extracted features from each of RC, GC, BC, and Gs images are concatenated and normalized. Finally, robust features are selected and fed to the least square support vector machine (LS-SVM) classifier. The obtained glaucoma detection accuracy of the proposed method is 84.95% which is more than the existing methods using the same image database. The proposed method may become suitable for ophthalmologists to detect glaucoma with better accuracy.