Machine-Learning-Scheme to Detect Choroidal-Neovascularization in Retinal OCT Image

Abstract

Eye is a fundamental sensory organ and any disease in eye will severely affect the sensory signal evaluation and conclusion making capability of the brain. The Choroidal-Neovascularization (CNV) is one of the harsh eye diseases in which a new blood-vessel grow from the choroid. Usually, the major cause of CNV is due to wet Age-Related-Macular-Degeneration (ARMD) and the formed new vessel will cause a leak in fluid which makes the retinal wet. The untreated CNV will lead to vision loss. In this research, detection of CNV using Optical-Coherence-Tomography (OCT) is presented using 484 images (242 Healthy and 242 CNV). In this work, a Machine-Learning-Scheme (MLS) is developed to examine the resized OCT of 256x256 pixels and the stages of this MLS includes; pre-processing, feature extraction, Mayfly-Optimization-Algorithm (MFA) based feature reduction, and two-class classification. The experimental outcome of this technique confirmed that the Fine-Gaussian-SVM (SVM-FG) classifier helped to accomplish an improved classification accuracy (>92%) compared to the alternative classifiers of this study.