Deep Learning for Diabetic Retinopathy Detection

ResearchSpace/Manakin Repository

Show simple item record

dc.contributor.advisor Abdulla, WH en
dc.contributor.author Thanati, Haneesha en
dc.date.accessioned 2020-01-22T02:42:22Z en
dc.date.issued 2019 en
dc.identifier.uri http://hdl.handle.net/2292/49682 en
dc.description Full Text is available to authenticated members of The University of Auckland only. en
dc.description.abstract Diabetics retinopathy (DR) is a vision-threatening complication occurs due to damage of blood vessels in the retina. It is one of the common causes of blindness among the working class population. Early detection and suitable treatment are crucial to prevent sight loss. The screening for the disease is through an examination of fundus images by trained clinicians. The presence of lesions like microaneurysms, haemorrhages and hard exudates are indicative of a damaged eye. Microaneurysms (MAs) presence is usually an early sign of diabetic retinopathy. Automatic detection of DR is vital for early detection of the disease, and can help making healthcare a ordable and e cient. Machine learning has been used extensively for automation, and its powerful subset deep learning is being used everywhere for complex image recognition tasks. The research in this thesis is an investigation of deep learning methods for diabetic retinopathy detection. We propose an e cient algorithm for DR classi cation. There are numerous methods used previously in this particular area of research, and this thesis has been built to augment these methods to design an automatic DR screening to assist the management and control of diabetic retinopathy disease. In this thesis, we reviewed the main methods used for DR screening. The pros and cons of each of these methods have been thoroughly investigated. A new convolutional neural network (CNN) architecture is designed, and its e ciency is analysed. The proposed CNN has been tested with several public datasets. Training of a convolutional neural network requires several hyperparameters such as learning rate, lter size and strides, and there are also di erent choices for the designing networks, each has been examined and discussed at length. The designed CNN has been trained for several iterations and experimented as a binary and multi-class classi er and has achieved a signi cant accuracy and good score on other metrics relevant to classi cation. It is hoped that from this research, we have contributed towards automatic DR detection and have moved a little forward toward the goal of preventing invertible blindness. en
dc.publisher ResearchSpace@Auckland en
dc.relation.ispartof Masters Thesis - University of Auckland en
dc.relation.isreferencedby UoA en
dc.rights Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated. Previously published items are made available in accordance with the copyright policy of the publisher. en
dc.rights Restricted Item. Full Text is available to authenticated members of The University of Auckland only. en
dc.rights.uri https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm en
dc.rights.uri http://creativecommons.org/licenses/by-nc-sa/3.0/nz/ en
dc.title Deep Learning for Diabetic Retinopathy Detection en
dc.type Thesis en
thesis.degree.discipline Electrical and Electronic Engineering en
thesis.degree.grantor The University of Auckland en
thesis.degree.level Masters en
dc.rights.holder Copyright: The author en
pubs.elements-id 792686 en
pubs.record-created-at-source-date 2020-01-22 en


Full text options

This item appears in the following Collection(s)

Show simple item record

http://creativecommons.org/licenses/by-nc-sa/3.0/nz/ Except where otherwise noted, this item's license is described as http://creativecommons.org/licenses/by-nc-sa/3.0/nz/

Share

Search ResearchSpace


Advanced Search

Browse