In an emergency healthcare situation, delay between injury and treatment is one of the most critical parameters with regard to survivability. Reduction in diagnosis/pre-treatment time by processing real-time ambulance data while en route to hospital can cut back the delay in treatment of the patient. However, several research challenges arise in accessing real-time patient data from ambulance to hospital while moving along different Road Side Units (RSUs). Due to the severity of medical data, there is a need to minimize computational losses along with costs due to migration and ambulance perceived latency. Considering the above scenarios, this paper formulates an average cost minimization problem keeping latency, energy, and loss function into deliberation as NP-hard. To solve the formulated problem, Minimum Cost Algorithm (MCA) using Federated Averaging (FedAvg) algorithm utilizing RSUs for effectively transferring real-time patient data to hospitals has been proposed considering above stated constraints altogether. Moreover, to handle imbalances in health data across different hospitals during processing, FedAvg algorithm combines augmentation techniques. Through experimental and prototype demonstration, the efficacy of proposed framework is shown by achieving$12.5 \%, 27 \%,$and$38 \%$reduction in an average total cost compared to other state-of-the-art techniques on real-world data sets, respectively.

The privacy-focused concept of Federated Learning (FL) allows local data processing without disclosing patients’ health details to a central server. However, its vulnerability to privacy breaches through shared model weights and susceptibility to a single point of failure remain concerns. Energy constraints of Wireless Body Area Networks (WBANs) necessitate considering computation and transmission energy in the FL process. Thus, this article introduces a smart healthcare system prioritizing energy efficiency and privacy through a blockchain-backed FL model. Yet, WBAN users might be unwilling to share data without adequate incentives, and miners might hesitate due to the high energy usage associated with maintaining the blockchain. Therefore, an optimization problem is formulated to maximize system utility while considering energy, WBAN incentives, miner revenue, and FL loss. A computationally efficient stable matching-based algorithm is proposed for optimizing utility via associating WBANs and miners. Associated WBANs use Quantized Neural Networks (QNNs) to minimize computation energy. Moreover, this work integrates Differential Privacy (DP) and Homomorphic Encryption (HE) mechanisms to prevent information leakage by adding noise to gradients before updating model weights and encrypting consequences before transmitting them to miners. Real-world experiments validate the framework, yielding an average of 15.1%, 9.03%, and 15.35% improvements over existing methods.

This work explores the coupling of Machine Learning (ML) and Wireless Body Area Network (WBAN) data to develop highly effective models. To support resource-constrained WBANs, we propose the integration of Drones-as-a-Service (DaaS) for on-demand data collection and model training. However, the growing number of WBAN users with varying 5G radio resources may cause interference and degrade system performance when transmitting data to Unmanned Aerial Vehicles (UAVs), hindering data sharing among independent UAVs. To address these challenges and enable privacy-preserving collaborative ML, we adopt Federated Learning (FL) framework, enabling independent UAV service providers to collaborate without sharing sensitive data. Furthermore, we aim to maximize the revenue of both WBANs, which contribute data, and UAVs, which perform model training. This requires careful resource allocation, considering minimum and maximum Physical Resource Block (PRB) requirements for transmitting critical and complete physiological data to UAVs underlying 5G networks. To tackle this complex problem, we propose an optimization framework that maximizes overall revenue while considering interference among WBANs. We apply stable matching and graph coloring-based heuristics to solve the problem efficiently. Extensive simulations and real-world data prototype demonstrate our proposed model's effectiveness, achieving an average revenue of 92.8% of the optimal value, outperforming existing state-of-the-art approaches.