Quantum-Assisted Optimization and Security for Trustworthy AI-Driven Healthcare

Digital health increasingly relies on artificial intelligence for clinical decision-making, patient management, and distributed learning, yet two persistent challenges remain: the computational complexity of optimization tasks, such as operating-room scheduling and genomic feature selection, and the need for robust security in federated and Internet-of-Medical-Things (IoMT) systems that are vulnerable to poisoning and spoofing attacks. Existing approaches often address these issues separately and struggle with scalability, interpretability, and resilience to advanced adversaries. This paper introduces a Hybrid Quantum-Classical Architecture (HQCA) that integrates the Quantum Approximate Optimization Algorithm (QAOA) for constrained healthcare decision-making with Quantum Digital Signatures (QDS) for secure authentication. The layered framework combines classical preprocessing with NISQ-era quantum modules and is validated under calibrated IBM Eagle-class noise models. We evaluate four representative tasks, including operating-room scheduling, genomic feature selection, federated-learning integrity, and IoMT firmware authentication, using short-depth QAOA (p≤ 3) and SWAP-test based QDS verification. HQCA reduces operating-room makespan by approximately 7.5% and halves overtime compared to classical baselines, while maintaining over 99.5% scheduling feasibility. In genomic analysis, it improves classification performance while selecting 32% less features. For security, it reduces successful poisoning in federated learning from 20.7% to 3.8%, and no spoofing events were observed in 30 trials for IoMT telemetry, with practical verification latency. These findings indicate that a unified hybrid stack can improve decision quality while providing information-theoretic authenticity guarantees with operationally feasible verification latency, demonstrating a practical pathway toward secure and efficient digital health as quantum hardware matures.