A cost-optimized medical digital twin framework for secure and efficient patient data management in smart healthcare

The increasing demand for personalized, real-time healthcare necessitates efficient, secure patient data management. Digital Twins (DTs) enable AI-powered monitoring and decision support but also introduce challenges related to latency, computational cost, and security. This paper proposes a cost-optimized, AI-driven Medical Digital Twin (MDT) framework that manages task allocation across heterogeneous edge, fog, and cloud infrastructures. The system is formulated as a tri-objective optimization model that jointly minimizes latency and operational cost while maximizing security, subject to resource and clinical-priority constraints. To solve this problem, three complementary approaches are developed: (i) an exact Integer Linear Programming (ILP) model for optimal benchmarking, (ii) a Patient-Aware Task Intelligence Greedy (PATI-Greedy) heuristic algorithm for low-latency decision-making, and (iii) a Hybrid Q-Learning Enhanced Genetic Algorithm (HybridQeGA) for scalable, near-optimal performance in complex environments. Extensive simulations in a smart ICU scenario with 4, 8, and 12 patients demonstrate that ILP consistently achieves the best objective values but is computationally impractical for large instances. PATI-Greedy executes rapidly with polynomial complexity, achieving results within 5-[Formula: see text] of ILP for small- to medium-scale workloads. HybridQeGA offers the closest match to ILP in larger problem sizes, with less than [Formula: see text] deviation in overall objective value while maintaining scalability. Security-sensitive scenarios highlight HybridQeGA's adaptability, improving security scores by an average of [Formula: see text] compared to PATI-Greedy. These findings establish a balanced trade-off between accuracy and computational efficiency, positioning the proposed framework as a robust and deployable solution for intelligent and trustworthy digital health ecosystems.