AGU, JOHN DAVY CHIDERA (2026) INTELLIGENT SYSTEM FOR EARLY CATARACTDETECTION. Other thesis, Godfrey Okoye University, Enugu.
![[img]](https://eprints.gouni.edu.ng/style/images/fileicons/text.png) |
Text
Intelligent_System_For_Early_cataract_Detection.pdf Download (2MB) |
Abstract
Cataracts have been found among the main causes of avoidable blindness worldwide andform a large portion of visual impairment in sub-Saharan Africa owing to the delayeddiagnosis and lack of specialist eye services. Early identification is very important for theproper action to be taken in order to stop the progression of the problem. In this study, weintroduce CataractAI, which is a web-based intelligent cataract detection platform thatallows users to sign up and authenticate, and upload their retinal fundus images forautomatic cataract screening. For increasing the efficiency of learning from an unbalanceddata set, Synthetic Minority Over-Sampling Technique (SMOTE) was used during trainingto create synthetic examples for the minority class to make class distribution balanced andimprove the classification process. The performance of the trained model was tested usinga separate test set with 750 retinal fundus images. According to experimental results, theproposed model successfully classified 747 out of 750 retinal images. Thus, it achieved anaccuracy of 99.60%, precision of 99.60%, recall of 99.60%, and an F1-score of 99.60%. Also, the model obtained the ROC-AUC score equal to 99.97%, which means the excellent discriminating ability between cases with cataracts and without them. The developedsystem can be used in primary health care centers, rural clinics, and eye care services inNigeria and the sub-Saharan Africa region generally, where there is a lack of access toophthalmologists and diagnostic equipment. With the ability to detect cataracts accuratelyand efficiently, the system has the potential to assist healthcare providers in detectingcataracts early, refer them for appropriate treatment and help in preventing blindness dueto cataracts. The results indicate that the use of a deep learning system based onEfficientNetB0 is a practical and accurate method for automated cataract detection.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | Q Science > Q Science (General) |
| Divisions: | Faculty of Computing And Information Technology (FACIT) |
| Depositing User: | Uchenna Eneogwe |
| Date Deposited: | 29 Jul 2026 15:31 |
| Last Modified: | 29 Jul 2026 15:31 |
| URI: | http://eprints.gouni.edu.ng/id/eprint/6083 |
Actions (login required)
 |
View Item |
