Detection of Dry and Wet Age-Related Macular Degeneration Using Deep Learning

Age-related macular degeneration (AMD) is a retinal disease in elderly people which deteriorate the central part of the retina.It is one of the leading causes of vision loss in the ageing persons.Every day, massive retinal images of patients with AMD are generated using the Optical Coherence Tomography (OCT) and other retinal imaging modalities.It is critical that these images are automatically analysed, so as to reduce the time consumption and over reliance on clinical professionals.The advance stage of AMD which usually causes loss of sight occurs in either dry or wet form.Most of the models developed in previous studies focuses on the classification of AMD infected and normal retinal images.However, in the later stages of AMD it is necessary to determine whether the AMD is dry or wet.Ability to classify between dry and wet Age-Related Macular Degeneration is very crucial to ophthalmologists in therapeutic indication.It determines whether a patient receives Anti-VEGF injection therapy treatment.The objective of this study is to develop a convolutional neural network model that will classify between dry and wet AMD.A pretrained Deep Residual Neural Network with 50-layers (ResNet50) was used to train the model using the KERMANY dataset consisting of 32,931 OCT images of dry and wet AMD.The model was evaluated and it performed with an accuracy of 96.56%, 98.20% Specificity and 89.45% sensitivity respectively.