Multiple Retinal Disease Diagnosis Using Customized Deep Learning Models and Hybrid Learning Model from Retinal Oct Images

Abstract

The human eye is a sophisticated organ that enables vision, allowing us to perceive the world around us. It consists of the iris, pupil, sclera, cornea, lens, retina, and optic nerve. Medical image processing and analysis can help the doctors to identify abnormalities, diagnosing diseases earlier, and make more accurate treatment decisions. Retinal ailments, including age-related macular degeneration, glaucoma, and diabetic retinopathy, can lead to significant vision loss or even blindness. Early detection and treatment are crucial for preserving vision. Traditionally, retinal diseases are diagnosed through a combination of methods, including optical coherence tomography (OCT), Indocyanine green angiography (ICGA), ultra-wide field imaging, and fluorescein angiography. An optical coherence tomography, which is a diagnostic method that does not involve any invasive procedures, is capable of producing cross-sectional scans of the retina in real time. In the realm of clinical diagnosis and therapy, medical image processing is essential. However, the conventional techniques utilized to evaluate these images require a significant investment of effort and time. This research aims to develop lightweight customized Convolution Neural Networks (CNN s) and hybrid models for detecting significant retinal diseases and simplifying system complexity to enable real-time application newline