NWANKWO, NESTA O (2026) EARLY DIABETIC RETINOPATHY DETECTION SYSTEM USING RETINAL IMAGES. Other thesis, Godfrey Okoye University, Enugu Nigeria.
|
Text
Nwankwo_EARLY_DIABETIC_RETINOPATHY_DETECTION_SYSTEM_USING_RETINAL_IMAGES.pdf Download (1MB) | Preview |
Abstract
Diabetic retinopathy (DR) is known to be one of the leading causes of blindness in the world especially among diabetics, whereby it is very crucial for early detection of the disease in order to prevent going blind. Conventional approaches to diagnose this condition are always slow, expensive, and require the presence of a qualified ophthalmologist. This research project will concentrate on designing an efficient and automatic system that will detect and monitor early development of diabetic retinopathy from retinal fundus images using machine learning and computer vision algorithms. This system involves pre-processing of the image in order to enhance its quality, followed by feature extraction and ultimately detection of certain clinical symptoms, including microaneurysms, abnormal blood vessels, and the optic disc. Classification of this image will be done using a random forest classifier, which will categorize images into normal and early stage of DR. This process will be done on the APTOS 2019 Blindness Detection dataset while handling the class imbalance problem. The system implementation will be carried out using a webbased framework known as Flask, which enables users to upload his/her retinal images for processing. Performance evaluation is carried out in terms of accuracy, 5 sensitivity, specificity, and F1-score, indicating appropriate performance of the system for the initial phase of detecting diabetic retinopathy. The initial estimates of the method indicate that its effectiveness is high, as indicated by an accuracy of 92.3%, precision of 91.5%, sensitivity (recall) of 93.1%, specificity of 94.2%, and F1-score of 92.3%. This study reveals that machine learning and image processing techniques can be applied to develop an efficient, inexpensive, and scalable solution for early detection of diabetic retinopathy.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | Q Science > Q Science (General) |
| Divisions: | Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science |
| Depositing User: | Cynthia Ugwuoti |
| Date Deposited: | 22 Jul 2026 11:24 |
| Last Modified: | 22 Jul 2026 11:24 |
| URI: | http://eprints.gouni.edu.ng/id/eprint/5994 |
Actions (login required)
 |
View Item |
