Prediction of Retina Damage in Optical Coherence Tomography Image Using Xception Architecture Model

Abstract

One of the most vital human organs is the retina. Most of the data is gathered via our eyesight. Thus, maintaining good eye health is crucial for happy and healthy life and eyes. Unfortunately, hazardous eye conditions, including choroidal neovascularization (CNV), Drusen, and diabetic macular edema (DME), directly harm the retina. They are typically discovered late, making it frequently impossible to cure and restore vision. Significant vision loss or even complete blindness may result from this. Ophthalmologists can view the inner structure of the retina using the sophisticated medical imaging technology known as noninvasive retinal optical coherent tomography (OCT), which relies on the visual reflection of the tissues inside the eye. However, in practice, errors continue to occur in diagnosing illnesses in general and eye ailments in particular. Thus, we develop a deep learning model to help physicians diagnose CNV, Drusen, and DME more correctly and lessen medical examination and treatment mistakes. About 8,000 images from the Large Dataset of Labeled Optical Coherence Tomography (OCT) Images were used to train with the Xception's architecture model. The result of this classification study for three types of DME, CNV, and Drusen diseases showed an accuracy of 93%.