A Conceptual Artificial Intelligence-Driven Framework for Detection of Eye Diseases in Nigeria

Abstract

Unavailability of accurate and timely diagnosis did not allow efforts made by health professionals to yield any meaningful result in Nigeria. Nigeria is the most populous country in Africa and home to over 200 million people, making it one of the most densely populated countries in the world. It is estimated that 4.25 million adults aged ≥ 40 years have moderate to severe eye disease or blindness. Allergic conjunctivitis (17.8%), age-related macular degeneration, cataracts, glaucoma (16.7%), cornea opacity, refractive errors (14.3%), and ocular trauma (21.7%) are known to be the major causes of eye disorders in Nigeria. Ocular injury was more common in males (p=0.002) and children aged 6–10 years, and 87.1% of these cases were closed-globe injuries, 80% of blindness cases are preventable and curable. Thus, artificial intelligence (AI) techniques, especially deep learning (DL) have emerged as important technique in handling very complex tasks. However, AI and DL are yet to be maximized in ocular medicine. This paper explored the concepts of deep learning methods in eye disease detection and classification. The exploration proposed a single modality and multimodal models, the models would be trained on a large multiclass dataset comprising of age, symptoms, laboratory test results, x-ray image results, diseases diagnosed and medical imaging (OCT, Fundus image, etc.) respectively. These would be obtained from the medical records of previously diagnosed and treated eye diseases. The techniques will improve accurate and timely diagnosis of eye diseases.