Early Detection of Pathological Myopia in Fundus Images Using Deep Learning

High myopia or Pathological Myopia (PM) is a common cause of blindness worldwide, with a high prevalence in Asia. Pathological Myopia (PM) is a severe form of myopia that is accompanied by retinal degeneration and can result in irreversible visual loss. Patients will be able to delay or even prevent suffering from irreversible visual loss, if early discovery is followed by rapid treatment and active follow-up. Experts in Pathologic Myopia, on the other hand, are still only a few around the world. Myopic maculopathy, or Pathologic detection, is a difficult diagnosis for ophthalmologists and optometrists. As a result, automated detection utilizing Deep Learning can aid clinicians in identifying Pathologic Myopia in early stages using fundus images which are commonly used in basic eye care settings around the world, they are non-invasive and easy to obtain. Our dataset contains 473 myopic fundus images and 600 normal images. We proposed a new model based on the VGG-19 convolutional neural network with an adjusted densely connected classifier that is used in this study to achieve an accuracy of 99.069%.