Multi-label classification of Retinal Disorders in Optical Coherence Tomography using Deep Learning

Abstract

We have developed a model with an intent of identifying the presence of each of three disorders, namely the Diabetic Macular Edema (DME), the Choroidal Neovascularization (CNV) and the Drusen, and if otherwise, then categorizing them as normal macula. OCT or Optical Coherence Tomography is a non-invasive imaging technique that produces high-resolution cross-sectional images of biological tissues. according to the literature survey, the identification of retinal disorders in OCT images of the retina is one of the most pressing issues in this domain. Keen on doing something about it we started researching and found that there have been very few recent publications in this domain on Deep Learning. Hence, we went ahead with working towards building a process of classifying the retinal disorders in OCT scans using Deep Learning. We have proposed a Deep-Learning-based detection system for the purpose of screening patients with blinding retinal diseases that can be treated if identified early. To achieve this, Convolutional Neural Network (CNN) is used. An accuracy of 99.38% with a precision amounting to 0.9938 is achieved. The Area Under Curve assesses to an average of 1.00.