JoWUA

Abstract

The delivery of medical information is the core of safe and stable contemporary healthcare, especially in telemedicine, continuous patient monitoring, and AI-controlled diagnosis. The reliability is, however, problematic because it suffers from two nagging problems, namely noise-degraded medical imaging, which makes diagnostic accuracy lower and the increased susceptibility of classical cryptography schemes to emergent quantum attacks, which is a threat to long-term patient confidentiality. To overcome them, the present research proposes a Hybrid Post-Quantum Encryption and Denoising Framework, combining deep-learning-based noise removal with a lattice-based, steganography-improved secure communication pipeline. The framework operates in two interrelated phases. Transformer-enhanced U-Net (TransUNet). First, TransUNet uses pre-encryption denoising and self-attention, along with hierarchical feature aggregation, to learn to capture local textures and long-range structural dependencies more efficiently than either wavelets or shallow Filters. Bayesian Optimisation gives rise to task-specific adaptivity because the model hyperparameters are tuned on a variety of noise profiles and imaging modalities. Denoising performance is strictly evaluated based on PSNR (12.887 dB to 9.787 dB), SSIM (0.3924 dB to 0.3965 dB), MSE (0.105998 to 0.054714) and edge-preservation measures to ensure that structures with diagnostic value are maintained. Second, the Hybrid Post-Quantum Stego-Crypto Module substitutes the traditional AES encryption with the lattice-based key-encapsulation, which is quantum attack-resistant. The steganography model is a GAN-based approach that covertly embeds the resulting ciphertext within innocent carrier data to enhance its confidentiality and detection resistance. Bit error rate, embedding distortion, detectability, key sensitivity, cypher entropy and adversarial robustness are used to assess system performance. On the whole, the suggested framework is likely to achieve better denoising fidelity, enhanced post-quantum security, and lower overall end-to-end latency than conventional pipelines, and has been tested on a variety of medical datasets and artificial network conditions.