Date on Master's Thesis/Doctoral Dissertation

5-2019

Document Type

Master's Thesis

Degree Name

M.S.

Department

Computer Engineering and Computer Science

Degree Program

Computer Science, MS

Committee Chair

El-Baz, Ayman

Committee Co-Chair (if applicable)

Elmaghraby, Adel

Committee Member

Imam, Ibrahim

Author's Keywords

diabetic retinopathy; medical imaging; convolutional neural networks; deep learning

Abstract

Diabetic Retinopathy is an eye disease that affects the blood vessels of the retina tissue found in the back of the eye. The blood vessels in the retina discharge fluid or cause hemorrhage when affected by diabetic retinopathy. It leads to loss of eye vision with people with diabetics and also among working adults. The disease may advance in four stages which include Mild nonproliferative, moderate nonproliferative, severe nonproliferative and proliferative diabetic retinopathy. In the mild stage also known as background retinopathy, lumps occur in the blood vessel and distort a little amount of blood. At this stage, you are at high risk of experiencing visual problems in the future, and to avoid the problem from escalating it is advised you take care and be more cautious. In the second stage, the blood vessels bulge and leak the blood, and there might be a high risk of vision infection. After the second stage, new blood vessels are formed causing severe bleeding and leading retina to pull away from the back of the eye in the third stage. In the proliferative diabetic retinopathy, the blood vessels in the macula found in the center of the retina distort and block. At this phase, you are advised to see a hospital specialist and having an often visit to monitor the eyes. Early detection of this disease is vital as it can prevent blindness from people suffering from diabetic retinopathy. Nowadays, using computer skills to diagnose eye illnesses is very common. In this study, by using a Low Complexity Convolutional Neural Network, you could address the stages of Diabetic Retinopathy. CNN accurately detect the phases and segmentation of images of the disease by using spatial analysis with a higher diagnostic accuracy, sensitivity, and specificity as 89.1%, 86.6%, and 96.4%, respectively.

Download

Share

COinS