Social choice considerations in cloud-assisted WBAN architecture for post-disaster healthcare: Data aggregation and channelization

Abstract

In a cloud-assisted Wireless Body Area Network (WBAN), health data of patients are transmitted from the Local Data Processing Units (LDPUs) to the health-cloud platform through a set of mobile monitoring nodes. In a post-disaster scenario, it is important that the data acquired by the body sensor nodes are aggregated in an efficient manner, and channelized to the cloud platform with minimum delay. Our work focuses on two fundamental research issues in this context – aggregation of health data transmitted by the LDPUs within the mobile monitoring nodes, and channelization of the aggregated data by dynamic selection of the cloud gateways. While the existing literature mainly center around the aggregation of the data broadcast through the body sensor nodes, our focus is on the aggregation of the data transmitted by the LDPUs. Aggregation takes place at the mobile monitoring nodes, which makes the problem challenging due to reasons attributed to mobility and health data prioritization. Our proposed solution, Body Area Network Data Aggregation Algorithm (Banag), generates a preference order for each patient, based on the severity and acuteness of his/her health. For each patient, an Exigency Factor, which measures health criticality, is associated. Additionally, a pseudo-cluster based “fair” aggregation policy is proposed on the basis of the Theory of Social Choice. Cloud gateways are also dynamically allocated to channelize the prioritized and aggregated health data in a “fair” manner using The Optimal Channelization Algorithm (OCA). The simulation results illustrate that the proposed pseudo-cluster based aggregation scheme provides improved performance in terms of reliability of node selection, number of packets transmitted, redundancy during transmission, and probability of congestion, when compared with cluster-based, tree-based, and structure-free aggregation methods. The health data channelization algorithm proposed in the paper also demonstrates that the choice of packets to be transmitted to the cloud is biased towards data criticality, and the dynamic selection of gateways is biased towards gateway capacity and reduced communication cost.

Introduction

The world witnesses many indomitable disasters, natural or human-induced, such as earthquake, avalanche, flood, tsunami, and terrorist bomb blast, which take a deep toll upon the society and mankind. The deadly consequences of these disasters lead to enormous agony and cause heavy distress among the victims. Contemporary disaster-relief operations are generally active and prompt, but are insufficient in managing large-scale disasters with high number of injured or severely wounded victims. It becomes increasingly difficult for the healthcare units to diagnose, monitor, and provide medical facilities to a huge mass of affected patients. The aftermath of a disaster typically leads to utter confusion and mismanagement. One of the main reasons behind such anarchy is that the healthcare units function as individual isolated entities. A sufficiently collaborative and cooperative approach, in respect of both technology and management, would improve the efficacy of service in such situations of medical emergency.

In this paper, we look at some of the technological prospectives to mitigate the post-disaster healthcare problems, and organize the efforts of the medical teams in a post-disaster scenario. Our work considers disaster management in a cloud-assisted Wireless Body Area Network (WBAN) environment. Our system model specifically considers WBAN-equipped patients in the lower layer of hierarchy, communicating with the health-cloud platform through a multi-tiered intermediate architecture. The rationale behind the choice of such a model is discussed in Section 3.

WBAN is evolving as a promising technology in the domain of ubiquitous and remote healthcare in the recent times [17], [40], [3]. It has found its widespread admissibility in many application domains such as battlefield, disaster healthcare, and biomedical applications [23], [52], [11], [32]. Conventionally, in a WBAN, nodes are embedded with sensors that are capable of sensing and monitoring certain physiological attributes of a human body such as heart rate, body temperature, blood pressure, blood oxygen saturation level, and respiratory rate. These sensor nodes can be either mounted on the human body, or implanted within, thereby, forming a scattered network topology [45], [20]. The data acquired by these wireless sensor nodes are then broadcast to the LDPU, over the wireless medium. All subsequent patient specific computations are performed by the LDPU.

In the proposed cloud-assisted WBAN architecture, the on-body sensors corresponding to each patient communicate with the LDPU. The LDPUs, in turn, communicate with a base station (BS). A BS might be several hops away from a patient, which may result in unwanted transmission delay or communication difficulties. This can subsequently have fatal and distressful implications on the affected victims. However, in emergency scenarios, the process of monitoring and tracking a patient’s health should be eminently prioritized, and, hence, the health data transmission is well optimized. Moreover, instead of maintaining all the healthcare centers as isolated units, we propose, in this paper, an efficient and dynamic mapping of the healthcare centers with the cloud framework. This facilitates concurrence and cooperative functioning of the individual centers as an extensive healthcare unit, thereby rendering better medical support and service.

It is difficult to have each LDPU communicate with a BS, specially in medical emergency situations. Such type of communication not only increases the cost, but also exhausts the energy of a body sensor, and introduces latency in transmissions. It is important to mention that the nodes are responsible for carrying data of the critically ailing patients. So, we must ensure to minimize cost and delay and simultaneously maximize throughput of the data communication. Therefore, the LDPUs of different patients must be partitioned into several subsets. Corresponding to every subset, an aggregated form of data is transmitted to an intermediate entity or a mobile monitoring node. These monitoring nodes communicate with the cloud platform through the cloud gateways. Due to the mobility of the monitoring nodes, mapping a mobile node to a cloud gateway for communication is an additional problem. The data from the cloud may be further transmitted to a board of online doctors or medical experts, who are responsible for the treatment of disaster-affected patients.

The proposed work focuses on three aspects:

• Dynamic subset formation of the LDPUs, and mapping of the subsets to the mobile monitoring nodes.

• Performing lossless data aggregation on behalf of every subset.

• Establishment of a communication map for every mobile monitoring node to the cloud gateway.

The first two aspects are fundamentally important from a doctor’s perspective. Lossless data of patients should be available to the medical expert or doctors to perform further analysis. It is essential to ascertain the proper choice of LDPUs which will lead to the establishment of a group that communicates with the mobile monitoring node causing minimal cost and delay. Moreover, the aggregation mechanism should be sensitive to the patient’s health criticality and the aggregation should be lossless in nature. The third aspect channelizes the health data, and manages the data traffic to alleviate network congestion. It tries to maintain an overall optimality in the process of communication. Health data transmission and its subsequent analysis can be overlaid on the cloud environment. In this work, the health-cloud renders dynamic monitoring and medical diagnosis facilities to the patients [8], [42].

The aforesaid concerns not only insist on making a critical patient’s data readily available to medical experts, but also raises the need to channelize and improve the overall network performance. In this paper, we propound the Body Area Network Data Aggregation Algorithm (Banag)¹ and Optimal Channelization Algorithm (OCA), which attempt to address the aforesaid issues in an efficient manner. The proposed algorithms are anchored in the Theory of Social Choice [5].

Dynamic data aggregation is a very important issue in the proposed system architecture. Based on the grouping of the LDPUs, the application performance can be improved. We select (or ignore) an LDPU to be inside (or outside) a group based on an aggregation function. A proper aggregation function necessarily needs to be “fair”, so that none of the eligible elements are ignored unjustly. Motivated by the generalized criteria of fairness, we discuss the properties postulated in the social welfare literature [34]:

• Majority: An aggregation function violates the majority criterion, if some node has a majority of the first place preferences, but eventually after aggregation, it remains overlooked.

• Condorcet: An aggregation function violates the condorcet criterion, if some node is preferred constantly against every other candidate, but ends up not being the winner of the selection.

• Irrelevant Alternatives: An aggregation function violates the irrelevant alternatives criterion, if having a loser node drop out of the race, changes the winner of the selection.

• Monotonicity: An aggregation function violates the monotonicity criterion, if one can transform the winner into a loser by moving the winner up the preference list on some of the individual preferences.

An aggregation function is expected to be “fair”, so that the outcome of the aggregation remains consistent with the individual preference of each LDPU. Furthermore, many LDPUs might get wrongly grouped, if the fairness criteria are not conformed with. Also, in such cases, the method of choosing a winner may also become unfair.

Following the aggregation, the data are transmitted to the cloud platform for subsequent analyses [59]. However, in a post-disaster environment, it is required to monitor patients’ health conditions remotely. This includes ambulatory healthcare services where the health status of a patient is examined continuously over time, while the patient is being moved to the emergency healthcare center. Therefore, the gateway through which the health data is transmitted to the cloud changes along with the global position of the patient. It is important to select gateway in such a way it has the capacity of forwarding the health data, and, also, incurs minimal energy expenditure while communication with the LDPU of the patient. This motivates the significance of data channelization. If the data is not properly channelized, it causes some gateways to be over-loaded, while some to remain idle. Data passing through over-loaded gateways may introduce unnecessary delay due to queuing within a gateway, and may increase communication cost significantly.

The importance of social choice is perceptible. We discuss and analyze how Social Choice theory helps to improve fairness, accuracy, and energy efficiency from a system’s point of view. The contributions of our work are catalogued below.

• The work focuses on the formation of pseudo-clusters so that the aggregation is not biased towards the leader nodes. Each patient is considered to be a member of a democratic society. This provides the opportunity to every sensor to transmit its data to the LDPU. However, the data from LDPU is processed and analyzed to a superficial extent before further transmission to the health-cloud.

• Data aggregation among the LDPUs is done in a “fair” manner, as discussed in Section 1.1. The health data packets are organized based on the acuteness or the severity of the health data of the disaster affected patients. Health-based priority is established, based on which the data is transmitted to the medical teams in an orderly manner.

• Aggregation is performed at mobile aggregation centers, thereby increasing the flexibility and scalability of the system. Load-sharing with optimal packet transmissions are ensured to speed-up the healthcare operation.

• After the aggregation of data, the gateways are allocated dynamically. At this stage, the packets are re-prioritized, so that the nodes with the most critical data are attended first. Gateways are allocated to nodes by optimizing the communication cost, health data criticality, and gateway capacity.

The remaining part of the paper is organized as follows. Section 2 contains a discussion on the related work. Section 3 elaborates the details of the architectural design. The workflow of the proposed algorithms are illustrated in Sections 4 , 5 . Section 6 presents some of the examples that illustrate the proposed solution. Section 7 illustrates the results of simulation. Finally, Section 8 concludes the work with discussions about how it can be enhanced in the future.