A Study on Fluid based Retinal abnormalities Analysis from OCT Images using SVM Classifier

摘要

Blindness is one of the most commonly occurring diseases in human. The reason being abnormality in the various layers of retina due to abnormal amount of fluid either excess accumulation or deficit. This paper aims at identifying and evaluating the such abnormality in the retinal layers by using classifier to diagnose the abnormality that may serve as prior stages to blindness. SVM Classifier was evaluated for around 114 images extracted from OCT Modality with common features using MATLAB. The SVM classifier was designed and evaluated to classify the input images as Normal and Abnormal, based on which the sensitivity and specificity were estimated to evaluate the system performance. Abnormalities taken into consideration are Cystoid Macular Edema (CME), Choroidal Neo Vascular Membrane (CNVM) and Macular Hole (MH) images for classification. The overall performances prove that the proposed system has an accuracy of 87.98%, say with sensitivity and specificity of 87.50% and 88.46% respectively.