High-performance hybrid AI systems with quantum-secure protocols for cyber-physical remote healthcare applications

Abstract

Digital Twin (DT) technology is increasingly important for real-time healthcare monitoring and predictive analytics. However, existing healthcare systems face critical challenges, including excessive computational load, high network latency, vulnerability to quantum cyberattacks, and inefficient strategies for distributing tasks across cloud and edge environments. Existing solutions often fail to scale efficiently, protect sensitive health data against future cyberattacks, or deliver reliable performance under dynamic conditions. To address these challenges, this work proposes an integrated healthcare framework that advances the state-of-the-art across multiple dimensions. First, to tackle the computational bottleneck in wearable healthcare devices, a lightweight one-dimensional convolutional neural network (CNN) accelerator was designed and implemented on field-programmable gate arrays (FPGAs), leveraging shift-based computation and pipelined architecture. This achieved a classification throughput of 1145 GOPS (Giga operations per second), enabling ultra-low-latency and energy-efficient biosignal analysis. Second, to enhance system responsiveness and scalability, a cloud-edge Digital Twin healthcare system was developed, leveraging secure Internet of Things (IoT) communication, dynamic telemetry optimization using Pyomo mathematical programming, and real-time predictive analytics. Third, to address emerging data security threats, a novel quantum-secure healthcare Digital Twin model was introduced, leveraging Quantum Key Distribution (QKD) protocols and hybrid artificial intelligence (AI) models combining multilayer perceptrons (MLP), extreme gradient boosting (XGBoost), and generative adversarial networks (GANs) for data augmentation. Finally, to optimize system resilience under dynamic healthcare conditions, a dynamic task offloading strategy was proposed, leveraging multi-agent reinforcement learning (MAPPO), adaptive cybersecurity protection (ACTO), and quantum-enhanced task preprocessing (AQDT-IoT). Experimental results demonstrate that the FPGA accelerator achieves 1145 GOPS throughput for real-time biosignal classification, while the proposed cloud-edge healthcare system reduces network latency by 40% and improves throughput by 30%. The hybrid AI model achieves an average prediction accuracy of 97.48% across health indicators under 10-fold crossvalidation. Moreover, the adaptive task offloading framework increases task success rates by 32% and reduces error rates by 80%, significantly improving operational efficiency and system robustness. Compared to previous approaches, the proposed framework delivers a highly scalable, secure, and intelligent Digital Twin Healthcare system, significantly strengthening patient monitoring, predictive decision-making, and preparedness against future quantum-era cybersecurity threats.