A Lightweight CNN for Multiclass Retinal Disease Screening with Explainable AI
Journal of Imaging · 15.08.2025 · 11 (8) · 275
DOI: 10.3390/jimaging11080275 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Imaging, online erschienen am 15.08.2025. Verfasst von Arjun Kumar Bose Arnob und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Computer Science, American International University-Bangladesh, Dhaka 1229, Bangladesh. Die Arbeit wurde bisher 17-mal zitiert (Crossref) und nennt 50 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: University Kuala Lumpur (UniKL), Malaysia.
Abstract (Originalsprache)
Timely, balanced, and transparent detection of retinal diseases is essential to avert irreversible vision loss; however, current deep learning screeners are hampered by class imbalance, large models, and opaque reasoning. This paper presents a lightweight attention-augmented convolutional neural network (CNN) that addresses all three barriers. The network combines depthwise separable convolutions, squeeze-and-excitation, and global-context attention, and it incorporates gradient-based class activation mapping (Grad-CAM) and Grad-CAM++ to ensure that every decision is accompanied by pixel-level evidence. A 5335-image ten-class color-fundus dataset from Bangladeshi clinics, which was severely skewed (17–1509 images per class), was equalized using a synthetic minority oversampling technique (SMOTE) and task-specific augmentations. Images were resized to 150×150 px and split 70:15:15. The training used the adaptive moment estimation (Adam) optimizer (initial learning rate of 1×10−4, reduce-on-plateau, early stopping), ℓ2 regularization, and dual dropout. The 16.6 M parameter network converged in fewer than 50 epochs on a mid-range graphics processing unit (GPU) and reached 87.9% test accuracy, a macro-precision of 0.882, a macro-recall of 0.879, and a macro-F1-score of 0.880, reducing the error by 58% relative to the best ImageNet backbone (Inception-V3, 40.4% accuracy). Eight disorders recorded true-positive rates above 95%; macular scar and central serous chorioretinopathy attained F1-scores of 0.77 and 0.89, respectively. Saliency maps consistently highlighted optic disc margins, subretinal fluid, and other hallmarks. Targeted class re-balancing, lightweight attention, and integrated explainability, therefore, deliver accurate, transparent, and deployable retinal screening suitable for point-of-care ophthalmic triage on resource-limited hardware.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Computer Science, American International University-Bangladesh, Dhaka 1229, Bangladesh.
- Faculty of Computer Science and Informatics, Berlin School of Business and Innovation, 12043 Berlin, Germany.
- Malaysian Institute of Information Technology, Universiti Kuala Lumpur, Kuala Lumpur 50250, Malaysia.
ORCID-Profile: Arjun Kumar Bose Arnob, Muhammad Hasibur Rashid Chayon, Fahmid Al Farid, Mohd Nizam Husen, Firoz Ahmed
Förderung
- University Kuala Lumpur (UniKL), Malaysia
Zitieren
APA
Arnob, A. K. B., Chayon, M. H. R., Al Farid, F., Husen, M. N., & Ahmed, F. (2025). A Lightweight CNN for Multiclass Retinal Disease Screening with Explainable AI. Journal of Imaging, 11(8), 275. https://doi.org/10.3390/jimaging11080275
Vancouver
Arnob AKB, Chayon MHR, Al Farid F, Husen MN, Ahmed F. A Lightweight CNN for Multiclass Retinal Disease Screening with Explainable AI. Journal of Imaging. 2025;11(8):275. doi:10.3390/jimaging11080275
BibTeX
@article{Arnob2025,
author = {Arnob, Arjun Kumar Bose and Chayon, Muhammad Hasibur Rashid and Al Farid, Fahmid and Husen, Mohd Nizam and Ahmed, Firoz},
title = {A Lightweight CNN for Multiclass Retinal Disease Screening with Explainable AI},
journal = {Journal of Imaging},
year = {2025},
volume = {11},
number = {8},
pages = {275},
doi = {10.3390/jimaging11080275}
}
EUVIMED