Intelligent Real‐Time Medical Image Transmission for Healthcare‐IoT via Dual‐Stage Probabilistic Restoration and Lightweight Nonlinear Encryption

ABSTRACT Telemedicine and healthcare‐IoT require medical image transmission that preserves diagnostic quality and confidentiality under noise, packet loss and cyber threats. Existing methods often separate restoration from security and lack real‐time efficiency. To address these limitations, this paper proposes an intelligent framework combining dual‐stage probabilistic restoration with lightweight nonlinear encryption. The restoration component, probabilistic image loss‐noise restoration (PILNR), operates in two phases: before encryption, structural preservation and redundancy embedding improve resilience to data loss; after decryption, normal probability‐based outlier correction (NPOC) uses vectorised z ‐score and CDF‐based operations to detect and correct corrupted structures efficiently. For confidentiality, the framework introduces a global‐local encryption architecture based on a nonlinear invertible N‐dimensional translation‐reflection substitution (ND‐TRS) mechanism. ND‐TRS maps groups of pixels into high‐dimensional Cartesian points and combines key‐dependent geometric reflection, modular translation and lightweight S‐box substitution to enhance confusion, diffusion and resistance to statistical, differential, brute‐force and chosen‐plaintext attacks. Experiments on large medical images demonstrate near‐ideal entropy (> 7.999), NPCR/UACI within theoretical bounds, PSNR near 30 dB under 50% data loss and 25% noise, strong sensitivity and fast execution, outperforming several recent state‐of‐the‐art encryption and restoration methods. These results support real‐time healthcare‐IoT medical image transmission with preserved diagnostic quality and data confidentiality.