Deep Learning Approach to Predict the Structural Irregularity of Retinal Images

According to the World Health Organization, at least 2.2 billion people have impaired vision. It is crucial to note that in at least 1 billion of these cases, vision impairment could have been prevented or is yet to be addressed. There exists a wide range of retinal irregularities ranging from Glaucoma to Diabetic Retinopathy (DR), each with its symptoms and causative. With vision impairment booming in number, it is important to put a system in place which can digitize the optic tests done to detect various retinal irregularities. This paper strives to be a better iteration of the pre-existing models in place, in terms of accuracy and F1 score using processed retinal fundus images with Discrete Wavelet Transform (DWT) and Flower Pollination Algorithm (FPA) in Deep Learning (DL) based Ensemble Learning model using Convolutional Neural Networks (CNNs). Additionally, the integration of the output results into a cloud-based application for storage and statistical analysis makes it a better solution for easy retrieval, future reference and visualization.