Detecting Eye Diseases Via Hybrid Deep Learning Algorithms From Fundus Images

According to the World Health Organization's worldwide report published in 2020, at least 2.2 billion people worldwide currently have visual impairment, and at least 1 billion of them have a visual impairment that can be prevented or has not yet been addressed. The world faces significant eye health challenges, including inequalities in the coverage and quality of eye care prevention, treatment, and rehabilitation services, a lack of trained eye care providers, and poor integration of eye care services into health systems, among others. It is known that more than 80% of all visual disorders can be prevented or treated. An eye fundus examination must be performed by a retina specialist to make a correct diagnosis, but people only consult an ophthalmologist when they feel any discomfort. While typically symptoms progress so much that once a disease occurs, resulting in expensive treatments and surgeries, often the damage is irreversible, resulting in visual impairment or even permanent vision loss. Artificial intelligence is used to study and develop theories and methods that can help simulate and extend human intelligence, which have been used in many fields of research such as automatic diagnosis and medicine. In recent years, the intersection of artificial intelligence (AI) technology and modern medicine has made effective and rapid disease screening possible. EyeCheckup is an automated software device designed to analyze digital color photographs of the ocular fundus to quickly screen for retinal and optic disc diseases. The main aim of the research is to evaluate the performance of the automatic screening algorithm to detect steerable retinal and optic disc diseases based on color fundus images and to determine its sensitivity and specificity towards possible diseases. For the clinical validation of the system, the images will be evaluated by ophthalmologists and the results will be compared with the artificial intelligence algorithm. After exclusions, this study will enroll up to 1528 subjects that meet the eligibility criteria. Participants who meet the eligibility criteria will be recruited after obtaining written informed consent from primary health care providers. Subjects will undergo fundus photography per, Food and Drug Administration (FDA) cleared, ophthalmic cameras. Images will be taken according to a specific EyeCheckup imaging protocol provided to the ophthalmic camera operator and then analyzed by the EyeCheckup v2.0 device. Methods and tools to be used in the research: I. Fundus photo capturing with non-mydriatic cameras: Optic disc-centered and fovea-centered fundus images will be taken with Canon CR-2 AF, Topcon TRC-NW400 and Optomed Aurora Non-mydriatic fundus cameras. For volunteers whose non-mydriatic images cannot be obtained, pupil dilation will be achieved by instilling tropicamide drops, and then images will be taken. Canon CR-2 AF, Topcon TRC-NW400 and Optomed Aurora Non-mydriatic fundus cameras, from which retina images will be taken, are CE marked and FDA approved. Tests to be done: I. Fundus images obtained with three different cameras from each volunteer included in the study will be analyzed separately for both the right eye and the left eye by the EyeCheckup artificial intelligence algorithm on a camera-based basis. ii. Evaluation of Canon CR-2 AF images by retina and glaucoma specialists for clinical validation of the system and comparison of the results,