Scholarlink

<p>The large-scale digitalization of healthcare infrastructures has accelerated the deployment of Medical Cyber-Physical Systems (MCPS), which integrate wearable sensors, diagnostic imaging devices, hospital information systems, and cloud-edge computing platforms. However, modern MCPS environments face critical challenges, including heterogeneous multimodal data fusion, real-time analytics under high-throughput medical streams, interoperability constraints, and increasing vulnerability to quantum-enabled cyber threats. Conventional cryptographic schemes such as RSA and ECC are no longer considered future-resilient in the post-quantum era. To address these limitations, this study proposes a Post-Quantum Secure Multimodal Transformer-based MCPS framework designed for intelligent, scalable, and quantum-resilient healthcare data digitalization. The proposed architecture integrates CRYSTALS-Kyber lattice-based key encapsulation and CRYSTALS-Dilithium digital signature schemes to ensure post-quantum secure communication among edge medical devices and centralized clinical servers. A Multimodal Cross-Attention Transformer (MCAT) is introduced to jointly learn global contextual representations from structured Electronic Health Records (EHR), physiological signal streams, and medical imaging data. Unlike conventional CNN-RNN pipelines, the proposed transformer architecture models long-range dependencies and cross-modal correlations through hierarchical self-attention and adaptive modality weighting mechanisms. Additionally, Bayesian Optimization (BO), Grey Wolf Optimization (GWO), and Whale Optimization Algorithm (WOA) are employed to enhance model convergence stability and predictive generalization under heterogeneous clinical distributions. The framework is experimentally validated using publicly available healthcare datasets, including MIMIC-III (clinical tabular data), PhysioNet ECG datasets, and MedMNIST imaging benchmarks. The proposed system achieves 98.36% classification accuracy, 97.82% precision, 98.11% recall, 97.96% F1-score, and an AUC of 0.991 for disease risk prediction and anomaly detection tasks. Security evaluation demonstrates 36.4% improved resilience against simulated quantum cryptanalytic attacks compared to ECC-based MCPS implementations, while maintaining acceptable encryption latency suitable for real-time clinical environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Intelligent medical cyber-physical systems in digital healthcare: a post-quantum secure multimodal transformer framework

  • 期刊论文

  • 发表日期：

    2026年4月30日

查看全文
  • Ahed Abugabah

摘要

The large-scale digitalization of healthcare infrastructures has accelerated the deployment of Medical Cyber-Physical Systems (MCPS), which integrate wearable sensors, diagnostic imaging devices, hospital information systems, and cloud-edge computing platforms. However, modern MCPS environments face critical challenges, including heterogeneous multimodal data fusion, real-time analytics under high-throughput medical streams, interoperability constraints, and increasing vulnerability to quantum-enabled cyber threats. Conventional cryptographic schemes such as RSA and ECC are no longer considered future-resilient in the post-quantum era. To address these limitations, this study proposes a Post-Quantum Secure Multimodal Transformer-based MCPS framework designed for intelligent, scalable, and quantum-resilient healthcare data digitalization. The proposed architecture integrates CRYSTALS-Kyber lattice-based key encapsulation and CRYSTALS-Dilithium digital signature schemes to ensure post-quantum secure communication among edge medical devices and centralized clinical servers. A Multimodal Cross-Attention Transformer (MCAT) is introduced to jointly learn global contextual representations from structured Electronic Health Records (EHR), physiological signal streams, and medical imaging data. Unlike conventional CNN-RNN pipelines, the proposed transformer architecture models long-range dependencies and cross-modal correlations through hierarchical self-attention and adaptive modality weighting mechanisms. Additionally, Bayesian Optimization (BO), Grey Wolf Optimization (GWO), and Whale Optimization Algorithm (WOA) are employed to enhance model convergence stability and predictive generalization under heterogeneous clinical distributions. The framework is experimentally validated using publicly available healthcare datasets, including MIMIC-III (clinical tabular data), PhysioNet ECG datasets, and MedMNIST imaging benchmarks. The proposed system achieves 98.36% classification accuracy, 97.82% precision, 98.11% recall, 97.96% F1-score, and an AUC of 0.991 for disease risk prediction and anomaly detection tasks. Security evaluation demonstrates 36.4% improved resilience against simulated quantum cryptanalytic attacks compared to ECC-based MCPS implementations, while maintaining acceptable encryption latency suitable for real-time clinical environments.