Citation

Rahman, Md. Naimur and Sadeque, Md. Golam and Al Emran, Md. and Rahul, Md. Kornel Ahmed and Sarker, Md Tanjil and Ramasamy, Gobbi (2025) Comparative Analysis of CNN Architectures for Eye Disease Diagnosis Using Retinal Images. In: 2025 Multimedia University Engineering Conference, MECON 2025, 21 July 2025 - 23 July 2025, Cyberjaya, Malaysia.

Text 23.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

Detecting eye diseases is crucial in medical diagnostics, particularly for preventing vision loss or blindness through timely intervention. This study aims to find the most effective DL model for the automated classification of common eye conditions from retinal fundus images. Specifically, we evaluate five advanced Convolutional Neural Network (CNN) architectures DenseNet201, MobileNetV2, NASNetMobile, ResNet50V2, and Xception, for classifying six eye diseases (Diabetic Retinopathy, Disc Oedema, Macular Degeneration, Myopia, Retinal Detachment, and Retinitis Pigmentosa) along with healthy cases. A total of 4,200 labeled fundus images were used, and transfer learning with fine-tuning was performed to optimize each model. The review process involved measuring classification accuracy, precision, recall, and F1-score. DenseNet201 got the best performance, with the highest test accuracy of 98.0% and superior metrics across all disease classes. While ResNet50V2 and Xception also showed strong performance, MobileNetV2 and NASNetMobile offered competitive results with lower computational cost, making them suitable for deployment in resource-constrained environments. This study shows the potential of CNNs to support clinical decision-making in ophthalmology and highlights directions for future work in model interpretability and generalizability using ensemble techniques and larger datasets.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Eye disease classification, transfer learning.
Subjects:	R Medicine > R Medicine (General) > R858-859.7 Computer applications to medicine. Medical informatics
Divisions:	Faculty of Artificial Intelligence & Engineering (FAIE)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	17 Mar 2026 07:19
Last Modified:	19 Mar 2026 03:21
URII:	http://shdl.mmu.edu.my/id/eprint/15527

Downloads

Edit (login required)