Manuscript received June 17, 2025; revised July 5, 2025; accepted August 13 , 2025; published January 16, 2026.

Abstract—This paper addresses the dual challenges of efficient compr ession and secure transmission for medical images, par -ticularly in band width-constrained telemedicine scenarios, like tele-radiography. We propose an end-to-end pipeline combining deep learning-based compr ession with chaos-based encryption. A Convolutional Autoencoder (CAE), optimized with a Structural Similarity Index Measure (SSIM) loss function and incorporat-ing residual connections and batch normalization, achieves an 8:1 (87.5%) compr ession ratio on Chest X-ray images while maintaining a high fidelity of 96% SSIM and 36 dB Peak Signal-to-Noise Ratio (PSNR). To secure the compact latent representation generated by the CAE, we introduce a lightweight, chaos-based encryption scheme operating directly on the latent space. This scheme utilizes a logistic map for confusion and secure permutations for diffusion. The experimental results confirm the effectiveness of the compression module in preserving high-frequency details and the encryption scheme’ s resistance against statistical attacks, by achieving high entr opy (7.92), strong randomness (0.99), without correlation (close to 0 in horizontal, vertical, and diagonal directions), and very sensiti ve to small changes in the key (1 single bit change conduct to a completely different k eystream). Our work offers a promising solution for secure and efficient transmission of medical images over constrained networks.

Keywords—convolutional autoencoder, chaotic cryptosystem, image compression, protected latent space, SSIM, secure transmission

Cite: Rayane Cherifi, Mahdi Madani*, El-Bay Bourennane, and Karima Amara, "Secure and Efficient Tele-Radiography Based on the Fusion of a Convolutional Autoencoder and Chaotic Latent Encryption," Journal of Image and Graphics, Vol. 14, No. 1, pp. 49-57, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC-BY-4.0).