International Journal of Computing and Digital Systems | 2026
Authors: Mediani H.; Tayeb S.; Mekouar S.; Himmi M.M.
DOI: 10.12785/ijcds/1571160064
Journal: International Journal of Computing and Digital Systems
Year: 2026
Publisher: University of Bahrain
Document Type: Article
Open Access: All Open Access; Gold Open Access
Cited by: 0
The objective of this project is to develop a fully automated, real-time magnetic resonance imaging (MRI) diagnostic system that reduces diagnostic variability and human error by integrating image enhancement and anomaly classification into a unified framework. We present ClassGAN, a hybrid architecture that combines a convolutional neural network (CNN) for anomaly detection with a generative adversarial network (GAN) for image enhancement, addressing the fragmentation of existing approaches, which typically treat super-resolution (SR) and classification as separate tasks. To generate high-resolution (HR) reconstructions from low-resolution (LR) inputs while preserving clinically relevant anatomical structures, the GAN module is guided by features extracted from the CNN-based classifier. The system improves both image quality and diagnostic performance by enabling automated preprocessing, feature extraction, and diagnosis within a single end-to-end pipeline, eliminating the need for human intervention. In addition to qualitative evaluations showing realistic detail restoration and effective noise reduction, experimental results on MRI datasets demonstrate that ClassGAN achieves strong and consistent quantitative performance in terms of reconstruction fidelity and classification metrics. The model’s robustness across different imaging conditions is confirmed by improvements in standard evaluation measures, including peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and classification accuracy. These findings indicate that the proposed integrated CNN–GAN framework provides a reliable, efficient, and real-time solution for medical image analysis. Overall, the results suggest that unified deep learning architectures can effectively bridge the gap between image enhancement and automated diagnosis, thereby supporting clinical decision-making and improving patient outcomes in modern medical imaging systems. © 2026, University of Bahrain. All rights reserved.
Deep Neural Network; Generative Adversarial Networks (GAN); High-resolution (HR); Low-resolution (LR); Magnetic Resonance Imaging (MRI); Super-Resolution (SR)