CLASSIFICATION OF RETINAL DISORDERS USING OPTICAL COHERENCE TOMOGRAPHY IMAGES BASED ON MEDICAL EXPERT SYSTEMS

Abstract

Vision impairment is increasing at an alarming rate. Diagnosis and classification of retinal disorders is a significant challenge in ophthalmological applications. The thesis aims to classify the optical coherence tomography images into four classes: Choroidal Neovascularization, Diabetic Macular Edema, Drusen, and normal cases. The thesis proposed a robust method based on both machine learning and deep learning approaches. Deep learning-based platform has been proposed using two novel techniques; InceptionV3 and SqueezeNet convolutional neural networks to classify the data and a hybrid machine-deep learning platform using Support Vector Machine (SVM), K-nearest neighbor (K-NN), Decision Tree (DT), and Ensemble Model (EM) has been proposed also to solve the same problem with another method. The proposed models are presented as a medical expert system that classifies the optical coherence tomography images into the main retinal disorders. The thesis introduces nine evaluation criteria for performance computation.