TY  - JOUR
AU  - Haque, A. K. M. Fazlul
AU  - Hridoy, Raihan Ahmed
AU  - Hossain, Meherab
AU  - Hasan, Nazia
AU  - Hridoy, Afzal Hossain
AU  - Akter, Taslima
PY  - 2026
DA  - 2026/04/09
TI  - An Intelligent and Secure Application for Early Detection of Eye Disease
JO  - ICCK Transactions on Mobile and Wireless Intelligence
T2  - ICCK Transactions on Mobile and Wireless Intelligence
JF  - ICCK Transactions on Mobile and Wireless Intelligence
VL  - 2
IS  - 1
SP  - 44
EP  - 55
DO  - 10.62762/TMWI.2026.590606
UR  - https://www.icck.org/article/abs/TMWI.2026.590606
KW  - eye disease detection
KW  - mobile intelligence
KW  - convolutional neural network
KW  - modified mobileNet
KW  - deep learning models
KW  - medical image analysis
KW  - predictive analytics
KW  - grad-cam
AB  - This paper presents a deep learning-based intelligent web application for the early detection of eye diseases using retinal fundus images. The dataset used in this study consists of 4,216 retinal fundus images collected from Kaggle, representing multiple eye disease categories. Multiple deep learning architectures, including CNN, DenseNet, InceptionV3, and ResNet, were evaluated and compared with a proposed modified MobileNetV2 architecture. The proposed architecture enhances the baseline MobileNetV2 by optimizing feature extraction and classification layers for improved performance in multi-class eye disease detection. Experimental results show that the proposed model achieved an overall classification accuracy of 88%, outperforming other evaluated models, including CNN (85.87%), DenseNet (87.19%), InceptionV3 (82.88%), and ResNet (50.67%), while maintaining low computational complexity. To improve transparency and interpretability of the deep learning model, Grad-CAM was integrated to visualize the discriminative regions of fundus images influencing the model’s predictions. The optimized model was deployed within a smart web application using a Flask backend and Streamlit frontend, enabling real-time disease prediction and visual explanation of results. The proposed system demonstrates the effectiveness of lightweight deep learning architectures combined with explainable AI techniques for accessible and reliable web-based ophthalmic diagnostic support systems. The optimized lightweight model was deployed as a mobile-friendly web application using Flask and Streamlit, enabling real-time inference on resource-constrained mobile devices and supporting wireless remote access for ophthalmic screening in underserved areas.
SN  - 3069-0692
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Haque2026An,
  author = {A. K. M. Fazlul Haque and Raihan Ahmed Hridoy and Meherab Hossain and Nazia Hasan and Afzal Hossain Hridoy and Taslima Akter},
  title = {An Intelligent and Secure Application for Early Detection of Eye Disease},
  journal = {ICCK Transactions on Mobile and Wireless Intelligence},
  year = {2026},
  volume = {2},
  number = {1},
  pages = {44-55},
  doi = {10.62762/TMWI.2026.590606},
  url = {https://www.icck.org/article/abs/TMWI.2026.590606},
  abstract = {This paper presents a deep learning-based intelligent web application for the early detection of eye diseases using retinal fundus images. The dataset used in this study consists of 4,216 retinal fundus images collected from Kaggle, representing multiple eye disease categories. Multiple deep learning architectures, including CNN, DenseNet, InceptionV3, and ResNet, were evaluated and compared with a proposed modified MobileNetV2 architecture. The proposed architecture enhances the baseline MobileNetV2 by optimizing feature extraction and classification layers for improved performance in multi-class eye disease detection. Experimental results show that the proposed model achieved an overall classification accuracy of 88\%, outperforming other evaluated models, including CNN (85.87\%), DenseNet (87.19\%), InceptionV3 (82.88\%), and ResNet (50.67\%), while maintaining low computational complexity. To improve transparency and interpretability of the deep learning model, Grad-CAM was integrated to visualize the discriminative regions of fundus images influencing the model’s predictions. The optimized model was deployed within a smart web application using a Flask backend and Streamlit frontend, enabling real-time disease prediction and visual explanation of results. The proposed system demonstrates the effectiveness of lightweight deep learning architectures combined with explainable AI techniques for accessible and reliable web-based ophthalmic diagnostic support systems. The optimized lightweight model was deployed as a mobile-friendly web application using Flask and Streamlit, enabling real-time inference on resource-constrained mobile devices and supporting wireless remote access for ophthalmic screening in underserved areas.},
  keywords = {eye disease detection, mobile intelligence, convolutional neural network, modified mobileNet, deep learning models, medical image analysis, predictive analytics, grad-cam},
  issn = {3069-0692},
  publisher = {Institute of Central Computation and Knowledge}
}