Secure near-lossless medical image compression and encryption with image-specific post-quantum key management and blockchain-based verification

Secure image transfer is a critical requirement in telemedicine and Picture Archiving and Communication Systems (PACS), where diagnostic integrity and patient confidentiality must be simultaneously ensured. High-resolution medical images, especially Magnetic Resonance Imaging (MRI), are often transferred over bandwidth-limited networks without impairing diagnostic quality or confidentiality against unauthorized access, manipulation, and replay. Existing approaches generally rely on a cascaded pipeline where image compression, encryption, and verification are considered as separate, independent operations. These distributed designs are prone to cryptographic key management issues, suffer from non-deterministic error control, lack meaningful coupling between image identity and cryptographic state, and provide no verifiable provenance or post-quantum key migration support. In this paper, we propose a DICOM-compliant system that integrates [Formula: see text]-regularized near-lossless image compression, hyperchaotic permutation diffusion encryption, post-quantum secure key establishment, and blockchain-based provenance verification as a unified operation. JPEG 2000 is used for primary image compression, and a residual refinement layer is introduced to limit residual correction amplitudes by [Formula: see text]-regularization to retain diagnostically significant information. Hyperchaotic encryption is performed with a symmetric image key generated using the NIST-standard post-quantum Key Encapsulation Mechanism (ML-KEM) and extracted using the HMAC-based Key Derivation Function (HKDF), cryptographically linked to DICOM image metadata. Message integrity is ensured using a keyed-hash message authentication code (HMAC), while a permissioned blockchain records only non-sensitive cryptographic metadata to enable verifiable provenance. Experimental results on clinical MRI and susceptibility-weighted imaging datasets demonstrate high reconstruction quality (PSNR: 46-51 dB, SSIM: 0.92-0.97), compression ratios ranging from 7.14:1 to 23.03:1 depending on modality and [Formula: see text], and strong cryptographic properties (NPCR: 99.6%, UACI: 33.46%, entropy: 7.9993). The evaluation focuses on MRI-based modalities, reflecting the intended scope of this study.