Deep Learning and Optimization Approaches in Secure and Energy-Efficient MRI Image Transmission via IoT Devices and Hybrid Physics-Guided Neural Networks: A Review

Abstract

The rapid growth of IoT-enabled healthcare has transformed MRI-based medical imaging, but secure and energy-efficient transmission remains challenging due to large image sizes, limited device resources, and increasing cyber threats. This review examines recent advances in deep learning, optimization techniques, and hybrid physics-guided neural networks for secure MRI image transmission. Deep learning models, including convolutional neural networks, autoencoders, and generative adversarial networks, have demonstrated excellent performance in image compression, reconstruction, and encryption while preserving diagnostic quality. Hybrid cryptographic techniques, such as chaotic encryption and deep learning-based security frameworks, further enhance data confidentiality and integrity. Physics-guided neural networks improve model generalization, interpretability, and robustness by incorporating physical constraints into the learning process, reducing dependence on large training datasets. The integration of optimization algorithms enables energy-efficient communication and intelligent resource management for IoT devices. Despite these advancements, challenges related to computational complexity, real-time implementation, scalability, and privacy preservation remain. Future research should focus on lightweight, explainable, and energy-aware hybrid AI frameworks that combine deep learning, optimization, and physics-guided modeling to achieve secure, reliable, and efficient MRI image transmission in next-generation IoT healthcare systems.