Sebastian, Jins, Deny, J., and Kumar, S.N. (2025) AI-Optimized Autoencoder Architecture for Secure and Efficient Brain MRI Image Encryption and Compression on EDGE Platforms. KSII Transactions On Internet And Information Systems, 19 (10). pp. 3325-3344.

PDF (Published Version) - Published Version Restricted to Repository staff only

Abstract

The advancements in imaging techniques have improved the diagnostics in health care and medicine. The magnitude of MRI volumetric data sets remains a challenge with storage space, real-time processing, and communication, especially in edge computing environments that have limited resources. To tackle these issues, we suggest a new artificial intelligence-based optimisation of an autoencoder network architecture that combines spectral data augmentation with feature extraction using discrete wavelet transform on the brain MRI images for data compression and encryption. The system is optimised for edge deployment on NVIDIA Jetson TX2 platforms, allowing real-time processing at low power consumption. The proposed convolutional autoencoder encodes and decodes the MRI images while minimizing the reconstruction loss. It uses spectral data augmentation to improve overfitting and generalisation, while DWT further improves compression efficiency through multi-resolution feature extraction. The proposed model's compression performance and image reconstruction accuracy are evaluated with various parameters, achieving a compression ratio of 33.25 and a compression factor of 0.0625 with a mean squared error of 0.000445, a peak signal-to-noise ratio of 33.51 dB. Additionally, a structural similarity index of 0.9956 confirms that the model reconstructs images with a high preservation of detail, which is necessary for capturing critical diagnostic data. The deployment of the developed model on the NVIDIA Jetson TX2 GPU platform emphasizes computational efficiency and real-time execution, making the system suitable for telemedicine, remote diagnostics, and embedded AI healthcare applications.

Item ID:	89604
Item Type:	Article (Research - C1)
ISSN:	1976-7277
Copyright Information:	Copyright ⓒ 2025 KSII
Date Deposited:	17 Nov 2025 03:13
FoR Codes:	40 ENGINEERING > 4003 Biomedical engineering > 400304 Biomedical imaging @ 100%
SEO Codes:	28 EXPANDING KNOWLEDGE > 2801 Expanding knowledge > 280110 Expanding knowledge in engineering @ 100%
Downloads:	Total: 1 Last 12 Months: 1
	More Statistics