Classification of retinal pathology from OCT images using a parametric tuned CNN
Authors
Singh, Shalini
Issue Date
2020
Degree
MSc in Data Analytics
Publisher
Dublin Business School
Rights holder
Rights
Items in eSource are protected by copyright. Previously published items are made available in accordance with the copyright policy of the publisher/copyright holder.
Abstract
Optometrists nowadays use optical biopsy to get cross sectional images of the retina infected by pathologies. This is also known as Optical Coherence Tomography (OCT). It is important to identify the retinal diseases at an early stage to prevent damage to the vision. There is a lot of research to be done to find a suitable method which can automatically detect retinal diseases. Therefore, we propose this research for automatic detection of retinal diseases by using a novel method of hyperparameter tuning instead of manually detecting the parameters of our Convolutional Neural Network (CNN). The Model is tested on metrics such as F1-score, precision, specificity, sensitivity, loss graph and accuracy. We also compare it with pretrained state-of-the-art model of Inception V3 and result shows that hyperparameter-tuned CNN gets better results. Being reliable, this proposed model can be used by optometrists to detect retinal disorders at an early stage.