invited paper wirelesssensornetworks forhealthcarejgko/papers/proc10.pdf · invited paper...

INV ITEDP A P E R

Wireless Sensor Networksfor HealthcareIn healthcare, there is a strong need to collect physiological data and sensornetworking; this paper reviews recent studies and points to the needfor future research.

By JeongGil Ko, Chenyang Lu, Mani B. Srivastava, John A. Stankovic, Fellow IEEE,

Andreas Terzis, and Matt Welsh

ABSTRACT | Driven by the confluence between the need to

collect data about people’s physical, physiological, psycholog-

ical, cognitive, and behavioral processes in spaces ranging from

personal to urban and the recent availability of the technologies

that enable this data collection, wireless sensor networks for

healthcare have emerged in the recent years. In this review, we

present some representative applications in the healthcare

domain and describe the challenges they introduce to wireless

sensor networks due to the required level of trustworthiness

and the need to ensure the privacy and security of medical data.

These challenges are exacerbated by the resource scarcity that

is inherent with wireless sensor network platforms. We outline

prototype systems spanning application domains from physi-

ological and activity monitoring to large-scale physiological and

behavioral studies and emphasize ongoing research challenges.

KEYWORDS | Healthcare monitoring; medical information

systems; wireless sensor networks

I . INTRODUCTION

Driven by technology advances in low-power networkedsystems and medical sensors, we have witnessed in recentyears the emergence of wireless sensor networks (WSNs)in healthcare. These WSNs carry the promise of drasticallyimproving and expanding the quality of care across a widevariety of settings and for different segments of the pop-ulation. For example, early system prototypes have demon-strated the potential of WSNs to enable early detection ofclinical deterioration through real-time patient monitoringin hospitals [13], [43], enhance first responders’ capabilityto provide emergency care in large disasters throughautomatic electronic triage [24], [50], improve the lifequality of the elderly through smart environments [72],and enable large-scale field studies of human behavior andchronic diseases [45], [58].

At the same time, meeting the potential of WSNs inhealthcare requires addressing a multitude of technicalchallenges. These challenges reach above and beyond theresource limitations that all WSNs face in terms of limitednetwork capacity, processing and memory constraints, aswell as scarce energy reserves. Specifically, unlike appli-cations in other domains, healthcare applications imposestringent requirements on system reliability, quality ofservice, and particularly privacy and security. In thisreview paper, we expand on these challenges and provideexamples of initial attempts to confront them.

These examples include: 1) network systems for vitalsign monitoring that show that it is possible to achievehighly reliable data delivery over multihop wirelessnetworks deployed in clinical environments [13], [43];2) systems that overcome energy and bandwidth limita-tions by intelligent preprocessing of measurementscollected by high data rate medical applications such asmotion analysis for Parkinson’s disease [49]; 3) an analysisof privacy and security challenges and potential solutions

Manuscript received January 29, 2010; revised May 2, 2010; accepted May 21, 2010.Date of publication September 13, 2010; date of current version October 20, 2010. Thework of J. Ko and A. Terzis was supported in part by the National Science Foundation(NSF) under Grant CNS-085591. The work of C. Lu was supported by the NationalCenter for Research Resources (NCRR) under Grant UL1-RR024992, part of the NationalInstitutes of Health (NIH) and NIH Roadmap for Medical Research. The work ofM. B. Srivastava was supported in part by the University of California at Los Angeles(UCLA) Center for Embedded Networked Sensing (CENS), and the NSF under AwardsCNS-0627084 and CNS-0910706. The work of J. A. Stankovic was supported by theNSF under Grants IIS-0931972 and EECS-0901686. The work of M. Welsh wassupported by the NSF under Grants CNS-0546338 and CNS-0519675 and also byMicrosoft, Intel, Sun, IBM, ArsLogica, and Siemens.J. Ko and A. Terzis are with the Department of Computer Science, Johns HopkinsUniversity, Baltimore, MD 21218 USA (e-mail: [email protected]; [email protected]).C. Lu is with the Department of Computer Science and Engineering, WashingtonUniversity in St. Louis, St. Louis, MO 63130 USA (e-mail: [email protected]).M. B. Srivastava is with the Electrical Engineering Department, University of Californiaat Los Angeles, Los Angeles, CA 90095 USA (e-mail: [email protected]).J. A. Stankovic is with the Department of Computer Science, University of Virginia,Charlottesville, VA 22904 USA (e-mail: [email protected]).M. Welsh is with the School of Engineering and Applied Sciences, Harvard University,Cambridge, MA 02138 USA (e-mail: [email protected]).

Digital Object Identifier: 10.1109/JPROC.2010.2065210

Vol. 98, No. 11, November 2010 | Proceedings of the IEEE 19470018-9219/$26.00 !2010 IEEE

in assisted living environments [72]; and 4) technologiesfor dealing with the large-scale and inherent data qualitychallenges associated with in-field studies [45], [58].

The remainder of the paper is structured as follows.The next section reviews background material in medicalsensing and wireless sensor networks, while Section IIIdescribes several promising healthcare applications forwireless sensor networks. We highlight the key technicalchallenges that wireless sensor networks face in thehealthcare domains in Section IV and describe represen-tative research projects that address different aspects ofthese challenges in Section V. We conclude with an outlineof the remaining challenges and future directions forwireless sensor networks in healthcare.

II . BACKGROUND

A. Medical SensingThere is a long history of using sensors in medicine and

public health [2], [74]. Embedded in a variety of medicalinstruments for use at hospitals, clinics, and homes, sensorsprovide patients and their healthcare providers insight intophysiological and physical health states that are critical to thedetection, diagnosis, treatment, andmanagement of ailments.Much of modern medicine would simply not be possible norbe cost effective without sensors such as thermometers, bloodpressure monitors, glucose monitors, electrocardiography(EKG), photoplethysmogram (PPG), electroencephalography(EEG), and various forms of imaging sensors. The ability tomeasure physiological state is also essential for interventionaldevices such as pacemakers and insulin pumps.

Medical sensors combine transducers for detecting elec-trical, thermal, optical, chemical, genetic, and other signalswith physiological origin with signal processing algorithms toestimate features indicative of a person’s health status.Sensors beyond those that directly measure health state havealso found use in the practice of medicine. For example,location and proximity sensing technologies [39] are beingused for improving the delivery of patient care and workflowefficiency in hospitals [22], tracking the spread of diseases bypublic health agencies [28], and monitoring people’s health-related behaviors (e.g., activity levels) and exposure tonegative environmental factors, such as pollution [58].

There are three distinct dimensions along which ad-vances in medical sensing technologies are taking place. Weelaborate on each of the three in the paragraphs that follow.

• Sensing modality: Advances in technologies such asmicroelectromechanical systems (MEMS), imag-ing, and microfluidic and nanofluidic lab-on-chipare leading to new forms of chemical, biological,and genomic sensing and analyses available outsidethe confines of a laboratory at the point of care. Byenabling new inexpensive diagnostic capabilities,these sensing technologies promise to revolution-ize healthcare both in terms of resolving public

health crisis due to infectious diseases [79] andalso enabling early detection and personalizedtreatments.

• Size and cost: Most medical sensors have tradi-tionally been too costly and complex to be usedoutside of clinical environments. However, recentadvances in microelectronics and computing havemade many forms of medical sensing more widelyaccessible to individuals at their homes, workplaces, and other living spaces./ The first to emerge [2] were portable medical

sensors for home use (e.g., blood pressure andblood glucose monitors). By enabling frequentmeasurements of critical physiological datawithout requiring visits to the doctor, theseinstruments revolutionized the managementof diseases such as hypertension and diabetes.

/ Next, ambulatory medical sensors, whose smallform factor allowed them to be worn or car-ried by a person, emerged [2]. Such sensorsenable individuals to continuously measurephysiological parameters while engaged rou-tine life activities. Examples include wearableheart rate and physical activity monitors andHolter monitors. These devices target fitnessenthusiasts, health conscious individuals, andobserve cardiac or neural events that may notmanifest during a short visit to the doctor.

/ More recently, embedded medical sensors builtinto assistive and prosthetic devices for geria-trics [78] and orthotics [18] have emerged.

/ Finally, we are seeing the emergence ofimplantable medical sensors for continuouslymeasuring internal health status and physio-logical signals. In some cases, the purpose is tocontinuously monitor health parameters thatare not externally available, such as intraoc-cular pressure in glaucoma patients [20]. Thegoal in other cases is to use the measurementsas triggers for physiological interventions thatprevent impending adverse events (e.g., epi-leptic seizures [62]) and for physical assis-tance (e.g., brain-controlled motor prosthetics[47]). Given their implantable nature, thesedevices face severe size constraints and needto communicate and receive power wirelessly.

• Connectivity: Driven by advances in informationtechnology, medical sensors have become increas-ingly interconnected with other devices. Earlymedical sensors were largely isolated with inte-grated user interfaces for displaying their measure-ments. Subsequently, sensors became capable ofinterfacing to external devices via wired interfacessuch as RS 232, USB, and Ethernet. More recently,medical sensors have incorporated wireless con-nections, both short range, such as Bluetooth,

Ko et al.: Wireless Sensor Networks for Healthcare

1948 Proceedings of the IEEE | Vol. 98, No. 11, November 2010

Zigbee, and near-field radios to communicatewirelessly to nearby computers, personal digitalassistants, or smartphones, and long range, such asWiFi or cellular communications, to communicatedirectly with cloud computing services. Besides theconvenience of tetherless operation, such wirelessconnections permit sensor measurements to besent to caregivers while patients go through theirdaily work life away from home, thus heralding anage of ubiquitous real-time medical sensing. Wenote that with portable and ambulatory sensors,the wired or wireless connectivity to cloudcomputing resources is intermittent (e.g., connec-tivity may be available only when the sensor is incellular coverage area or docked to the user’s homecomputer). Therefore, such sensors can also recordmeasurements in nonvolatile memory for upload-ing at a later time when they can be shared withhealthcare personnel and further analyzed.

B. Wireless Sensor PlatformsRecent years have witnessed the emergence of various

embedded computing platforms that integrate processing,storage, wireless networking, and sensors. These embed-ded computing platforms offer the ability to sense physicalphenomena at temporal and spatial fidelities that werepreviously impractical. Embedded computing platformsused for healthcare applications range from smartphonesto specialized wireless sensing platforms, known as motes,that have much more stringent resource constraints interms of available computing power, memory, networkbandwidth, and available energy.

Existing motes typically use 8- or 16-b microcontrollerswith tens of kilobytes of RAM, hundreds of kilobytes ofROM for program storage, and external storage in the formof Flash memory. These devices operate at a few milliwattswhile running at about 10 MHz [61]. Most of the circuitscan be powered off, so the standby power can be about1 !W. If such a device is active for 1% of the time, itsaverage power consumption is just a few microwattsenabling long-term operation with two AA batteries.Motes are usually equipped with low-power radios suchas those compliant with the IEEE 802.15.4 standard forwireless sensor networks [33]. Such radios usually trans-mit at rates between 10 and 250 Kb/s, consume about20–60 mW, and their communication range is typicallymeasured in tens of meters [6], [71]. Finally, motes includemultiple analog and digital interfaces that enable them toconnect to a wide variety of commodity sensors.

These hardware innovations are paralleled by advancesin embedded operating systems [21], [30], component-based programming languages [25], and networking pro-tocols [9], [26].

In contrast to resource-constrainedmotes, smartphonesprovide more powerful microprocessors, larger data stor-age, and higher network bandwidth through cellular and

IEEE 802.11 wireless interfaces at the expense of higherenergy consumption. Their complementary characteristicsmake smartphones and motes complementary platformssuitable for different categories of healthcare applications,which we discuss in the section that follows.

III . HEALTHCARE APPLICATIONS

Wirelessly networked sensors enable dense spatio–temporal sampling of physical, physiological, psychologi-cal, cognitive, and behavioral processes in spaces rangingfrom personal to buildings to even larger scale ones. Suchdense sampling across spaces of different scales is resultingin sensory information based healthcare applicationswhich, unlike those described in Section II-A, fuse andaggregate information collected from multiple distributedsensors. Moreover, the sophistication of sensing hasincreased tremendously with the advances in cheap andminiature, but high-quality sensors for home and personaluse, the development of sophisticated machine learningalgorithms that enable complex conditions such as stress,depression, and addiction to be inferred from sensoryinformation, and finally the emergence of pervasiveInternet connectivity facilitating timely dissemination ofsensor information to caregivers.

In what follows, we introduce a list of healthcareapplications enabled by these technologies.

• Monitoring in mass-casualty disasters: While triageprotocols for emergency medical services alreadyexist [31], [70], their effectiveness can quicklydegrade with increasing number of victims. More-over, there is a need to improve the assessment ofthe first responders’ health status during suchmass-casualty disasters. The increased portability,scalability, and rapidly deployable nature ofwireless sensing systems can be used to automat-ically report the triage levels of numerous victimsand continuously track the health status of firstresponders at the disaster scene more effectively.

• Vital sign monitoring in hospitals: Wireless sensingtechnology helps address various drawbacks asso-ciated with wired sensors that are commonly usedin hospitals and emergency rooms to monitorpatients [43]. The all too familiar jumble of wiresattached to a patient is not only uncomfortable forpatients leading to restricted mobility and moreanxiety, but is also hard to manage for the staff.Quite common are deliberate disconnections ofsensors by tired patients and failures to reattachsensors properly as patients are moved around in ahospital and handed off across different units.Wireless sensing hardware that are less noticeableand have persistent network connectivity to back-end medical record systems help reduce thetangles of wires and patient anxiety, while alsoreducing the occurrence of errors.

Ko et al. : Wireless Sensor Networks for Healthcare

Vol. 98, No. 11, November 2010 | Proceedings of the IEEE 1949

• At-home and mobile aging: As people age, theyexperience a variety of cognitive, physical, andsocial changes that challenge their health, inde-pendence, and quality of life [76]. Diseases such asdiabetes, asthma, chronic obstructive pulmonarydisease, congestive heart failure, and memorydecline are challenging to monitor and treat. Thesediseases can benefit from patients taking an activerole in the monitoring process. Wirelessly net-worked sensors embedded in people’s living spacesor carried on the person can collect informationabout personal physical, physiological, and behav-ioral states and patterns in real-time and every-where. Such data can also be correlated with socialand environmental context. From such Blivingrecords,[ useful inferences about health and well-being can be drawn. This can be used for self-awareness and individual analysis to assist inmaking behavior changes, and to share withcaregivers for early detection and intervention. Atthe same time such procedures are effective andeconomic ways of monitoring age-related illnesses.

• Assistance with motor and sensory decline: Anotherapplication of wireless networked sensing is toprovide active assistance and guidance to patientscoping with declining sensory and motor capabili-ties. We are seeing the emergence of new types ofintelligent assistive devices that make use of infor-mation about the patient’s physiological and physicalstate from sensors built in the device, worn or evenimplanted on the user’s person, and embedded in thesurroundings. These intelligent assistive devices cannot only tailor their response to individual users andtheir current context, but also provide the user andtheir caregivers crucial feedback for longer termtraining. Traditional assistive devices such as canes,crutches, walkers, and wheel chairs can fuseinformation from built-in and external sensors toprovide the users with continual personalizedfeedback and guidance towards the correct usage ofthe devices. Such devices can also adapt the physicalcharacteristics of the device with respect to thecontext and a prescribed training or rehabilitationregimen [78]. Furthermore, wireless networkedsensing enables new types of assistive devices suchas way finding [17] and walking navigation [8] for thevisually impaired.

• Large-scale in-field medical and behavioral studies:Body-worn sensors together with sensor-equippedInternet-connected smartphones have begun torevolutionize medical and public health researchstudies by enabling behavioral and physiologicaldata to be continually collected from a large num-ber of distributed subjects as they lead their day today lives. With their ability to provide insight intosubject states that cannot be replicated in con-

trolled clinical and laboratory settings and thatcannot be measured from computer-assisted retro-spective self-report methods, such sensing systemsare becoming critical to medical, psychological, andbehavioral research. Indeed, a major goal of theexposure biology program under National Instituteof Health (NIH) Genes and Environment Initiative(GEI) is to develop such field deployable sensingtools to quantify exposures to environment (e.g.,psychosocial stress, addiction, toxicants, diet, phys-ical activity) objectively, automatically, and for daysat a time in the participants’ natural environments.Researchers, both within the GEI program (e.g.,[35], [45], and [58]) and elsewhere (e.g., [27], [55],and [63]), have also recognized the utility of suchsensing in making measurements for longitudinalstudies ranging from the scale of individuals tolarge populations.

As the four examples above show, the applicationsenabled by wireless networked sensing technologies aredistributed across multiple dimensions. One dimension isthe spatial and temporal scope of distributed sensing. Thespatial scope can range from sensory observations of healthstatus made when an individual is confined to a building(e.g., home, hospital) or a well-defined region (e.g., disastersite) to observations made as an individual moves aroundduring the course of daily life. The temporal scope can rangefrom observations made for the duration of an illness or anevent to long-term observations made for managing a long-term disease or for public health purposes. Different spatialand temporal scopes place different constraints on theavailability of energy and communications infrastructure,and different requirements on ergonomics.

A second dimension is that of the group size, which canrange from an individual patient at home, to groups ofpatients at a hospital and victims at disaster sites, and allthe way to large dispersed population of subjects in amedical study or an epidemic.

The last critical dimension is the type of wirelessnetworking and sensing technologies that are used: on-body sensors with long-range radios, body-area networksof short-range on-body sensors with a long-range gateway,sensors implanted in-body with wireless communicationand power delivery, wireless sensors embedded in assistivedevices carried by individuals, wireless sensors embeddedin the environment, and sensors embedded in the ubiq-uitous mobile smartphones. Clearly, there is a rich diver-sity of wireless sensing technology with complementarycharacteristics and catering to different applications.Typically, more than one type of sensing technology getsused for a single application.

IV. TECHNICAL CHALLENGES

In the paragraphs that follow, we describe some of the corechallenges in designing wireless sensor networks for



healthcare applications. While not exhaustive, the chal-lenges in this list span a wide range of topics, from corecomputer systems themes such as scalability, reliability,and efficiency, to large-scale data mining and data asso-ciation problems, and even legal issues.

A. TrustworthinessHealthcare applications impose strict requirements on

end-to-end system reliability and data delivery. Forexample, pulse oximetry applications, which measure thelevels of oxygen in a person’s blood, must deliver at leastone measurement every 30 s [37]. Furthermore, end usersrequire measurements that are accurate enough to be usedin medical research. Using the same pulse oximetryexample, measurements must deviate at most 4% fromthe actual oxygen concentrations in the blood [37]. Finally,applications that combine measurements with actuation,such as control of infusion pumps and patient controlledanalgesia (PCA) devices, impose constraints on the end-to-end delivery latency. We term the combination of datadelivery and quality properties the trustworthiness of thesystem and claim that medical sensing applications requirehigh levels of trustworthiness.

A number of factors complicate the systems’ ability toprovide the trustworthiness that applications require.First, medical facilities, where some of these systems willbe deployed, can be very harsh environments for radio-frequency (RF) communications. This harshness is theresult of structural factors such as the presence of metaldoors and dividers as well as deliberate effort to provideradiation shielding, for example, in operating rooms thatuse fluoroscopy for orthopedic procedures. In fact, Ko et al.recently found that packet losses for radios following theIEEE 802.15.4 standard is higher in hospitals thanother indoor environments [42]. Moreover, devices thatuse 802.15.4 radios are susceptible to interference fromWiFi networks, Bluetooth devices, and cordless phones allof which are heavily used in many hospitals.

The impact of obstacles and interference is exacerbatedby the fact that most wireless sensor network systems uselow-power radios to achieve long system lifetimes (i.e.,maximizing the battery recharging cycle). The other impli-cation of using low-power radios is that the networkthroughput of these devices is limited. For example, thetheoretical maximum throughput of IEEE 802.15.4 radiosis 250 Kb/s and much lower in practice due to constraintsposed by medium access control (MAC) protocols andmultihop communications. Considering that applicationssuch as motion and activity monitoring capture hundredsof samples per second, these throughput limits mean that anetwork can support a small number of devices or that onlya subset of the measurements can be delivered in real time.

In some cases, the quality of the data collected fromwireless sensing systems can be compromised not bysensor faults and malfunctions, but by user actions. This istrue even for smartphone-based sensing systems for which

many of the above mentioned RF challenges are lesssevere. Considering that wireless sensing systems forhealthcare will be used by the elderly and medical staffwith little training, loss in quality due to operator misuse isa big concern. Moreover, because wireless sensing enablescontinuous collection of physiological data under condi-tions not originally envisioned by the sensors’ developers,the collected measurements may be polluted by a variety ofartifacts. For example, motion artifacts can have an impacton the quality of heart rate and respiration measurements.Therefore, estimating the quality of measurements col-lected under uncertain conditions is a major challenge thatWSNs for healthcare must address. In turn, this challengemeans that WSNs need to employ techniques for auto-mated data validation and cleansing and interfaces tofacilitate and verify their correct installation. Last but notleast, WSNs in healthcare should provide metadata thatinform data consumers of the quality of the data delivered.

B. Privacy and SecurityWireless sensor networks in healthcare are used to

determine the activities of daily living (ADL) and providedata for longitudinal studies. It is then easy to see that suchWSNs also pose opportunities to violate privacy. Further-more, the importance of securing such systems willcontinue to rise as their adoption rate increases.

The first privacy challenge encountered is the vaguespecification of privacy. The Heath Insurance Portabilityand Accountability Act (HIPPA) by the U.S. Government isone attempt to define this term [1]. One issue is thatHIPPA as well as other laws define privacy using humanlanguage (e.g., English), thus, creating a semantic night-mare. Nevertheless, privacy specification languages havebeen developed to specify privacy policies for a system in aformal way. Once the privacy specifications are specified,healthcare systems must enforce this privacy and also beable to express users’ requests for data access and thesystem’s policies. These requests should be evaluatedagainst the predefined policies in order to decide if theyshould be granted or denied. This framework gives rise tomany new research challenges, some unique to WSNs, aswe describe in the paragraphs that follow.

• Since context can affect privacy, policy languagesmust be able to express different types of contextfrom the environment such as time, space, physi-ological parameter sensing, environmental sensing,and stream based noisy data. Moreover, most of thecontext must be collected and evaluated in realtime. Since context is so central it must also beobtained in a secure and accurate manner.

• There is a need to represent different types of dataowners and request subjects in the system as wellas external users and their rights when differentdomains such as assisted living facilities, hospitals,and pharmacies interact. One of the more difficultprivacy problems occurs when interacting systems



have their own privacy policies. Consequently,inconsistencies in such policies may arise acrossdifferent systems. For this reason, online consis-tency checking and notification along with resolu-tion schemes are required.

• There is a need to represent high-level aggregatingrequests such as querying the average, maximum,or minimum reading of specified sensing data. Thisprivacy capability must be supported by anonymiz-ing aggregation functions. This need arises forapplications related to longitudinal studies andsocial networking.

• There is a need to support not only adherence toprivacy for data queries (e.g., data pull requests),but also the security for push configurationrequests to set system parameters (e.g., for privateuse or configuring specific medical actuators).

• Because WSNs monitor and control a large varietyof physical parameters in different contexts, it isnecessary to tolerate a high degree of dynamics andpossibly even allow temporary privacy violations inorder to meet functional, safety, or performancerequirements. For example, an individual wearingan EKG might experience heart arrhythmia and thereal-time reporting of this problem takes prece-dence over some existing privacy requirements. Inother words, to send an emergency alert quickly itmay be necessary to skip multiple privacy protec-tions. Whenever such violations occur, corehealthcare staff members must be notified ofsuch incidents.

In addition to policy and database query privacyviolations, WSNs are susceptible to new side channelprivacy attacks that gain information by observing theradio transmissions of sensors to deduce private activities,even when the transmissions are encrypted. This physicallayer attack needs only the time of transmission and thefingerprint of each message, where a fingerprint is a set offeatures of an RF waveform that are unique to a particulartransmitter. Thus, this is called the fingerprint and timing-based snooping (FATS) attack [67].

To execute a FATS attack, an adversary eavesdrops onthe sensors’ radio to collect the timestamps and finger-prints of all radio transmissions. The adversary then usesthe fingerprints to associate each message with a uniquetransmitter, and uses multiple phases of inference todeduce the location and type of each sensor. Once this isknown, various private user activities and health condi-tions can be inferred.

For example, Srinivasan et al. introduce this uniquephysical layer privacy attack and propose solutions withrespect to a smart home scenario [67]. Three layers ofinference are used in their work. First, sensors in the sameroom are clustered based on the similarity of theirtransmission patterns. Then, the overall transmissionpattern of each room is passed to a classifier, which

automatically identifies the type of room (e.g., bathroomor kitchen). Once the type of room is identified, thetransmission pattern of each sensor is passed to anotherclassifier, which automatically identifies the type of sensor(e.g., a motion sensor or a refrigerator door). From thisinformation, the adversary easily identifies several activ-ities of the home’s residents such as cooking, showering,and toileting, all with consistently high accuracy. Fromsuch information, it is then possible to infer the residents’health conditions.

Fortunately, many solutions with different tradeoffsare possible for this type of physical layer attack. Suchsolutions include 1) attenuating the signal outside of thehome to increase the packet loss ratio of the eavesdropper,2) periodically transmitting radio messages whether or notthe device has data to be sent, 3) randomly delaying radiomessages to hide the time that the corresponding eventsoccurred, 4) hiding the fingerprint of the transmitter, and5) transmitting fake data to emulate a real event.

Unfortunately, an adversary can combine informationavailable from many (external) sources with physical layerinformation to make inferences even more accurate andinvasive. New solutions that are cost effective, addressphysical layer data, protect against inferences based oncollections of related data, and still permit the originalfunctionality of the system to operate effectively areneeded.

A related fundamental problem, yet unsolved in WSNs,is dealing with security attacks. Security attacks areespecially problematic to low-power WSN platformsbecause of several reasons including the strict resourceconstraints of the devices, minimal accessibility to thesensors and actuators, and the unreliable nature of low-power wireless communications. The security problem isfurther exacerbated by the observation that transient andpermanent random failures are common in WSNs andsuch failures are vulnerabilities that can be exploited byattackers. For example, with these vulnerabilities, it ispossible for an attacker to falsify context, modify accessrights, create denial of service, and, in general, disrupt theoperation of the system. This could result in a patient beingdenied treatment, or worse, receiving the wrong treatment.

Having in mind such unique challenges, new light-weight security solutions that can operate in these openand resource-limited systems are required. Solutions thatexploit the considerable amount of redundancy found inmany WSN systems are being pursued. This redundancycreates great potential for designing WSN systems thatcontinuously provide their target services despite theexistence of failures or attacks. In other words, to meetrealistic system requirements that derive from long livedand unattended operation, WSNs must be able to continueto operate satisfactorily and effectively recover fromsecurity attacks. WSNs must also be flexible enough toadapt to attacks not anticipated during design or deploy-ment time. Work such as the one proposed by Wood et al.



provides an example of how such problems are addressed,by proposing to design a self-healing system with thepresence and detection of attacks, rather than trying tobuild a completely secure system [77].

C. Resource ScarcityIn order to enable small device sizes with reasonable

battery lifetimes, typical wireless sensor nodes make use oflow-power components with modest resources. Fig. 1shows a typical wearable sensor node for medical appli-cations, the SHIMMER platform [34]. The SHIMMERcomprises an embedded microcontroller (TI MSP430;8-MHz clock speed; 10-KB RAM; 48-KB ROM) and a low-power radio (Chipcon CC2420; IEEE 802.15.4; 2.4 GHz;250-Kb/s PHY data rate). The total device power budget isapproximately 60 mW when active, with a sleep powerdrain of a few microwatts. This design permits small,rechargeable batteries to maintain device lifetimes ofhours or days, depending on the application’s duty cycles.

The extremely limited computation, communication,and energy resources of wireless sensor nodes lead to anumber of challenges for system design. Software must bedesigned carefully with these resource constraints in mind.The scant memory necessitates the use of lean, event-driven concurrency models, and precludes conventionalOS designs. Computational horsepower and radio band-width are both limited, requiring that sensor nodes tradeoff computation and communication overheads, forexample, by performing a modest amount of on-boardprocessing to reduce data transmission requirements.Finally, application code must be extremely careful withthe node’s limited energy budget, limiting radio commu-nication and data processing to extend the battery lifetime.While smartphone-based systems typically enjoy moreprocessing power and wireless bandwidth, the fact that

they are less flexible compared to customizable moteplatforms limits their capability to aggressively conserveenergy. This leads to shorter recharge cycles and can limitthe types of applications that smartphones can support.

Another consideration for low-power sensing platformsis the fluctuation in the resource load experienced by sensornodes. Depending on the patient’s condition, the sensor databeing collected, and the quality of the radio link, sensornodes may experience a wide variation in communicationand processing load over time. As an example, if sensornodes perform multihop routing, a given node may berequired to forward packets for one or more other nodesalong with transmitting its own data. The network topologycan change over time, due to node mobility and environ-mental fluctuations in the RF medium, inducing unpre-dictable patterns of energy consumption for which theapplication must be prepared.

V. SYSTEMS

Next, we present several wireless sensing system proto-types developed and deployed to evaluate the efficacy ofWSNs in some of the healthcare applications described inSection III. While wireless healthcare systems usingvarious wireless technologies exist, this work focuses onsystems based on low-power wireless platforms for physi-ological and motion monitoring studies, and smartphone-based large-scale studies.

A. Physiological MonitoringIn physiological monitoring applications, low-power

sensors measure and report a person’s vital signs (e.g.,pulse oximetry, respiration rate, temperature). These appli-cations can be developed and deployed in different contextsranging from disaster response, to in-hospital patient mon-itoring, and long-term remote monitoring for the elderly.

While triage protocols for disaster response already exist(e.g., [31] and [70]), multiple studies have found that theycan be ineffectual in terms of accuracy and the time totransport as the number of victims increases in multicasualtyincidents [5], [65]. Furthermore, studies in hospitals reportthat patients are left undermonitored [15] and emergencydepartments today operate at or over capacity [4]. Finally,anecdotal evidence suggest that this lack of patientmonitoring can lead to fatalities [14], [54], [68].

Therefore, systems that automate patient monitoringhave the potential to increase the quality of care both indisaster scenes and clinical environments. Systems such asCodeBlue [50], MEDiSN [43], and the Washington Uni-versity’s vital sign monitoring system [13] target these appli-cation scenarios. Specifically, CodeBlue [50] aims toimprove the triage process during disaster events with thehelp of WSNs comprising motes with IEEE 802.15.4 radios.The CodeBlue project integrated various medical sensors[e.g., EKG, SpO2, pulse rate, electromyography (EMG)] withmote-class devices and proposed a publish/subscribe-based

Fig. 1. The SHIMMER wearable sensor platform. SHIMMER

incorporates a TI MSP430 processor, a CC2420 IEEE 802.15.4 radio,

a triaxial accelerometer, and a rechargeable Li-polymer battery.

The platform also includes a MicroSD slot supporting up to 2 GB of

Flash memory.



network architecture that also supports priorities andremote sensor control [11]. Finally, victims with CodeBluemonitors can be tracked and localized using RF-basedlocalization techniques [48].

Ko et al. proposed MEDiSN to address similar goals asCodeBlue (e.g., improve the monitoring process of hospitalpatients and disaster victims as well as first responders),but using a different network architecture [43]. Specifi-cally, unlike the ad hoc network used in CodeBlue,MEDiSN employs a wireless backbone network of easilydeployable relay points (RPs). RPs are positioned at fixedlocations and they self-organize into a forest rooted at oneor more gateways (i.e., PC-class devices that connect to theInternet) using a variant of the collection tree protocol(CTP) [26] tailored to high data rates. Motes that collectvital signs, known as miTags (see Fig. 2), associate with RPsto send their measurements to the gateway. The dedicatedbackbone architecture that MEDiSN incorporates signif-icantly reduces the routing overhead compared to a mobilead hoc network architecture and results in two majorbenefits. First, it allows the network’s operator to expandits coverage and engineer its performance by altering thenumber and position of RPs in the backbone. Second, sincemiTags do not have to route other nodes’ data, theyaggressively duty cycle their radio to conserve energy. TheWashington University’s patient monitoring has adopted asimilar wireless backbone network to take advantage ofsimilar benefits [12], [13].

The systems described above were deployed in disastersimulations [24] and hospital pilot studies [13], [41], [42].These studies showed that wireless sensing systems can infact overcome the challenging RF conditions that exist inthese environments to meet the applications’ stringenttrustworthiness requirements [42].

Chipara et al. found that another source of unreliabilityin clinical environments is the outage of the sensingcapability itself [13]. The authors show that the distribu-tion of sensing outages is heavy-tailed containing pro-longed outages caused by sensor disconnections. Theirexperience reveals that the use of automatic sensor discon-nection alarms and oversampling can enhance systemreliability. Finally, the pilot studies above also report thatthe satisfaction levels of healthcare personnel and userssuch as patients or disaster victims is high and concludethat the systems are practically feasible.

While the systems introduced above deal with improv-ing the quality of patient care in hospitals or disasterscenarios, researchers and practitioners noticed that thecoming worldwide silver tsunami [69], where a largenumber of retiring elders overload the capacity of currenthospitals, is stressing the traditional concept of healthcare,which is focused on clinical and emergency medicalservice (EMS) settings. Specifically, it is economically andsocially advantageous to reduce the burden of diseasetreatment by enhancing prevention and early detectionwhile allowing people to stay at home for as long aspossible. This requires a long-term shift from a centralized,expert-driven, crisis-care model to one that permeatespersonal living spaces and involves informal caregivers,such as family, friends, and members of the community.

A typical home healthcare system based on WSN isAlarmNet [75], [76], an assisted-living and residentialmonitoring network for pervasive, adaptive healthcare.AlarmNet is a system based on an extensible, heteroge-neous network architecture targeting ad hoc, wide-scaledeployments. It includes custom and commodity sensors,an embedded gateway, and a back-end database withvarious analysis programs. The system includes protocolssuch as context-aware protocols informed by circadianactivity rhythm analysis for smart power management. Itsupports real-time online sensor data streaming and aninference system to recognize anomalous behaviors aspotential indicators of medical problems. Privacy control isbased on access control lists and all queries are logged.Future work is planned to use data mining on the querylogs to detect privacy attacks. All messages are encryptedto ensure data confidentiality.

Intel Research Seattle and the University ofWashingtonhave built a prototype system that can infer a person’sADLs [59]. In their system, sensor tags (both passive andactive) are placed on everyday objects such as a toothbrushor a coffee cup. The system tracks the movement of taggedobjects with tag readers. Their long-term goal is to developa computerized and unobtrusive system that helps manageADLs for the senior population [38].

University of Rochester is building the Smart MedicalHome [46], which is a five-room Bhouse[ outfitted withinfrared sensors, computers, biosensors, and video camerasfor use by research teams to work on research subjects as theytest concepts and prototype products. Researchers observe

Fig. 2. Medical information tag, or miTag for short, used in

MEDiSN [43]. The miTag is a Tmote mini-based [53] patient monitor

that includes a pulse oximetry sensor with LEDs, buttons and an

LCD screen. The miTag is powered using a rechargeable 1200-mAh

3.7-V Li–Ion battery and external finger tip sensors are used to make

the pulse oximetry measurements.



and interact with subjects from two discreet observationrooms integrated into the home. Their goal is to develop anintegrated personal health system that collects data for24 h a day and presents it to the healthcare professionals.

Georgia Institute of Technology built an Aware Home[40] as a prototype for an intelligent space. This spaceprovides a living laboratory that is capable of gatheringinformation about itself and the different types of activitiesof its inhabitants. The Aware Home combines context-aware and ubiquitous sensing, computer-vision-basedmonitoring, and acoustic tracking all together for ubiqui-tous computing of everyday activities while remainingtransparent to its users.

The Massachusetts Institute of Technology is workingon the PlaceLab initiative [36], which is a part of theHouse_n project. The mission of House_n is to conductresearch by designing and building real living environ-mentsVBliving labs[Vthat are used to study technologyand design strategies in context. The PlaceLab is a one-bedroom condominium with hundreds of sensors installedin nearly every part of the home.

The systems introduced above provide useful physiolog-ical information to medical personnel using resourceconstrained devices. Nevertheless, these systems deal withonly the simplest aspects of medical data security. Forexample, MEDiSN performs 128-b advanced-encryption-standard-based encryption and authentication to secure allphysiological data [32] but does not provide any of the policycontrols described above. Another limitation of existingsystems is the small number of sensors that each mobiledevice can support due to hardware constraints. Developingnew platforms that integrate stronger security and privacymechanisms with more diverse sensing and processingcapabilities is likely to increase the range of physiologicalmonitoring applications that WSNs can support.

B. Motion and Activity MonitoringAnother application domain for WSNs in healthcare is

high-resolution monitoring of movement and activitylevels. Wearable sensors can measure limb movements,posture, and muscular activity, and can be applied to arange of clinical settings including gait analysis [60], [64],[73], activity classification [29], [52], athletic performance[3], [51], and neuromotor disease rehabilitation [49], [57].In a typical scenario, a patient wears up to eight sensors(one on each limb segment) equipped with MEMS accel-erometers and gyroscopes. A base station, such as aPC-class device in the patient’s home, collects data fromthe network. Data analysis can be performed to recover thepatient’s motor coordination and activity level, which is inturn used to measure the effect of treatments.

In such studies, the size and weight of the wearablesensors must be minimized to avoid encumbering thepatient’s movement. The SHIMMER sensor platformshown in Fig. 1 measures 44.5! 20! 13 mm and weighsjust 10 g, making it well suited for long-term wearable use.

In contrast to physiological monitoring, motion anal-ysis involves multiple sensors on a single patient eachmeasuring high-resolution signals, typically six channelsper sensor, sampled at 100 Hz each. This volume of sensordata precludes real-time transmission, especially overmultihop paths, due to both bandwidth and energy con-straints. The SHIMMER platform incorporates a MicroSDinterface, permitting up to 2 GB of storageVenough tostore up to a month of continuously sampled sensor data.While the energy consumption of flash input/output isnonnegligible, it is about one-tenth the energy cost totransmit the same amount of data over the radio. As aresult, it is necessary to carefully balance data sampling,storage, processing, and communication to achieveacceptable battery lifetimes and data fidelity.

Two systems, SATIRE [23] and Mercury [49], typifythe approach to addressing these challenges. SATIRE isdesigned to identify a user’s activity based on acceler-ometers and global positioning system (GPS) sensorsintegrated into Bsmart attire[ such as a winter jacket.SATIRE nodes, which are based on the MICAz [16]platform, measure accelerometer data and log it to localflash. These data are opportunistically transmitted usingthe low-power radio when the SHIMMER node is withincommunication range with the base station. Once the dataare collected at the base station, the collected data areprocessed offline to characterize the user’s activity pat-terns, such as walking, sitting, or typing. Sensor nodesperform aggressive duty cycling to reduce power con-sumption, extending lifetimes from several days to severalweeks.

The goal of the Mercury system is to permit long-termstudies of a patient’s motor activity for neuromotor diseasestudies, including Parkinson’s disease, stroke, and epilep-sy. Energy is far more constrained in Mercury than inSATIRE, due to the use of lightweight sensor nodes withsmall batteries. Mercury builds upon SATIRE’s approachto energy management and integrates several energy-aware adaptations, including dynamic sensor duty cycling,priority-driven data transmissions, and on-board featureextraction. Mercury is being used in several studies ofParkinson’s and epilepsy patients [49].

While SATIRE and Mercury show the feasibility ofusing low-power wireless platforms to perform longitudi-nal studies of human activity, issues related to improvingnode lifetime and providing stronger security and privacyguarantees remain areas of active research.

C. Large-Scale Physiological and Behavioral StudiesThe final application of WSNs in healthcare that we

discuss is their use in conducting large-scale physiologicaland behavioral studies. The confluence of body-area net-works of miniature wireless sensors (such as the previouslymentioned miTag and SHIMMER platforms), always-connected sensor-equipped smartphones, and cloud-baseddata storage and processing services is leading to a new



paradigm in population-scale medical research studies,particularly on ailments whose causes and manifestationsrelate to human behavior and living environments.

Traditionally such studies are either conducted incontrolled clinical laboratory settings with artificial stimuli,or rely on computer-assisted retrospective self-reportmethods. Both of these approaches have drawbacks: thecomplex subtleties of real-life affecting human behavior canrarely be recreated accurately in a laboratory, and self-reportmethods suffer from bias, errors, and lack of compliance.However, the combination of body-area wireless sensornetworks, smartphones, and cloud services permits physical,physiological, behavioral, social, and environmental data tobe collected from human subjects in their natural environ-ments continually, in real time, unattended, and in anunobtrusive fashion over long periods. Typically, data arecollected from wireless sensors worn by subjects, wirelessmedical instruments, and sensors embedded in devices suchas smartphones. After local validation, artifact removal, andlocal processing, sensor data are wirelessly transmitted usingcellular or WiFi networks to cloud-based services forsubsequent analysis, visualization, and sharing by research-ers. Such systems provide insight into subject states thattraditional study methods simply cannot achieve. Conse-quently, research efforts such as the exposure biologyprogram under NIH’s GEI are developing field-deployablewireless sensing tools to quantify exposures to environments(e.g., psychosocial stress, addiction, toxicants, diet, physicalactivity) objectively, automatically, and for multiple days inparticipants’ natural environments.

One example of such systems is AutoSense [45], inwhich objective measurements of personal exposure topsychosocial stress and alcohol are collected in the studyparticipants natural environments. A field-deployable suiteof wireless sensors form a body-area wireless network andmeasure heart rate, heart rate variability, respiration rate,skin conductance, skin temperature, arterial blood pres-sure, and blood alcohol concentration. From these sensorreadings, which after initial validation and cleansing at thesensor are sent to a smartphone, features of interest indi-cating onset of psychosocial stress and occurrence ofalcoholism are computed in real time. The collected infor-mation is then disseminated to researchers answeringbehavioral research questions about stress, addiction, andthe relationship between the two. Moreover, by also cap-turing time-synchronized information about a subject’sphysical activity, social context, and location, factors thatlead to stress can also be inferred, and this information canpotentially be used to provide personalized guidance aboutstress reduction.

A second example is a portable system called thephysical activity and location measurement system(PALMS) developed at the University of California SanDiego [58]. PALMS aims at monitoring study subjects ineveryday life for long enough periods of time to detectpatterns in physical activity and energy expenditure. These

information (collected from combined heart rate andmotion sensors) and location (from GPS units) are col-lected in the natural environment of the study participants.The system helps answer questions about the energy usedby a person during different activities in the course of theday and the variance across a population of subjects. Thesynchronized geolocation information permits under-standing how physical activity and energy expenditurevaries by location and is influenced by environmentalfactors such as the built environment, crime, the avail-ability of parks, and recreation facilities, or terrain.

These systems for population-scale medical studies arestill in their early stages, and several technical andalgorithmic challenges remain to be addressed. Energy iscertainly one challenge. While some on-body sensors havehigh sampling rates leading to significant energy con-sumption (i.e., low battery life), the desire to facilitate easycompliance with the study protocols preclude a frequentcharging schedule.

However, a bigger challenge with this technology is theissue of information privacy, and its tension with the qualityand value of information [19]. Contemporary privacypractices center on the notion of Bpersonally identifiableinformation[ and Binformed consent.[However, with thesesystems, the traditional intuitive notion of privacy is notenough. Privacy is not just about removing explicitidentifiers, encrypting data, using trusted software, andsecuring servers. These are easily done, though imperfectly.Sensory information traces captured by these systems arehighly personal. Embedded in them is information thatcorrelates with our identity and our behaviors. Whencombined with publicly available innocuous factsVthe so-called Bdigital footprints[ and Binformation breadcrumbs[that we all leave behind as we lead our livesVthese sensorinformation traces can be deanonymized, and subjects’identities and life patterns can be inferred statistically.

For example, Chaudhuri and Mishra showed thatpersonal information may be identified even fromanonymized and sanitized population level data sets [10].Similarly, Krumm has shown that location traces can bedeanonymized via statistical analysis to infer subjects’home location with high probability [44], which then canbe used to reveal their identity using information that isfreely available on the web such as reverse white-pagelookup.

Likewise, traditional prior informed consent is notadequate when sensors may capture data in unanticipatedsituations and the sheer amount and nature of sensor datamakes information leakage risks hard to comprehend.Collecting data continuously as subjects go through theirdaily lives at their homes, offices, and other places meansthat it is impossible to anticipate upfront, and accordinglyinform subjects about, the complete nature of informationthat the sensor data may reveal. Some of the seeminglyinnocuous sensor data thus collected in relatively uncon-trolled settings may capture information about confidential



aspects of subjects’ life patterns, personal habits, andmedical condition.

One answer to these problems can be to allow the studysubjects and patients to retain control over their raw sensordata throughout its life cycle: its capture, sharing, retention,and reuse [7], [66]. However, giving study subjects controlover data raises concern about quality of data for researchers.As it is, ensuring high-quality trustworthy information fromsensors out in the real world is hard due to malfunctions,misbehaviors, and lack of compliance. Letting subjectsselectively hide or perturb data raises the issue of bias andavailability, and thus utility. Quoting P. Ohm from a recentarticle: BData can either be useful or perfectly anonymous,but never both[ [56]. Technology assists such as automatedvalidation procedures, audit traces, and incentive mechan-isms to ensure compliance and encourage sharing mayprovide further help.

VI. FUTURE DIRECTIONS

Driven by user demand and fueled by recent advances inhardware and software, the first generation of wirelesssensor networks for healthcare has shown their potential toalter the practice of medicine. Looking into the future, thetussle between trustworthiness and privacy and the abilityto deploy large-scale systems that meet the applications’requirements even when deployed and operated inunsupervised environments is going to determine theextent that wireless sensor networks will be successfullyintegrated in healthcare practice and research. h

Acknowledgment

The authors would like to thank the anonymousreviewers for their comments in improving the quality ofthis paper.

REFERENCES

[1] 104th Congress. (1996). Health InsurancePortability and Accountability Act of 1996.Public Law 104-191. [Online]. Available:http://www.cms.hhs.gov/hipaageninfo/downloads/hipaalaw.pdf.

[2] P. A. Aberg, T. Togawa, and F. A. Spelman, Eds.,Sensors inMedicine and Healthcare. New York:Wiley, 2002.

[3] A. Ahmadi, D. D. Rowlands, and D. A. James,BInvestigating the translational and rotationalmotion of the swing using accelerometers forathlete skill assessment,[ in Proc. 5th IEEEConf. Sensors, Oct. 2006, pp. 980–983.

[4] American Hospital Association. (2005, Oct.).The state of America’s hospitalsVTaking thepulse. [Online]. Available: http://www.ahapolicyforum.org/ahapolicyforum/reports/.

[5] G. Asaeda, BThe day that the START triagesystem came to a STOP: Observations fromthe World Trade Center disaster,[ Acad.Emergency Med., vol. 9, no. 3, pp. 255–256,2002.

[6] Atmel Corporation. AT86RF230: Low power2.4 GHz transceiver for ZigBee, IEEE 802.15.4,6LoWPAN, RF4CE and ISM applications.[Online]. Available: http://www.atmel.com/dyn/resources/prod_documents/doc5131.pdf.

[7] J. Benaloh, M. Chase, E. Horvitz, andK. Lauter, BPatient controlled encryption:Patient privacy in electronic medicalrecords,[ in Proc. ACM Cloud Comput. SecurityWorkshop, Chicago, IL, 2009, pp. 103–114.

[8] S. Bohonos, A. Lee, A. Malik, C. Thai, andR. Manduchi, BUniversal real-timenavigational assistance (URNA): An urbanbluetooth beacon for the blind,[ in Proc. 1stACM SIGMOBILE Int. Workshop Syst. Netw.Support Healthcare Assisted Living Environ.,New York, 2007, pp. 83–88.

[9] M. Buettner, G. Yee, E. Anderson, and R. Han,BX-MAC: A short preamble MAC protocol forduty-cycled wireless sensor networks,[ inProc. 4th Int. Conf. Embedded Netw. Sensor Syst.,Boulder, CO, 2006, pp. 307–320.

[10] K. Chaudhuri and N. Mishra, BWhen randomsampling preserves privacy Advances inCryptologyVCRYPTO 2006, vol. 4117,Berlin, Germany: Springer-Verlag, 2006,pp. 198–213.

[11] B. Chen, K. K. Muniswamy-Reddy, andM. Welsh, BAd-hoc multicast routing onresource-limited sensor nodes,[ in

Proc. 2nd Int. Workshop Multi-Hop Ad HocNetw.: From Theory to Reality, Florence, Italy,2006, pp. 87–94.

[12] O. Chipara, C. Brooks, S. Bhattacharya, C. Lu,R. D. Chamberlain, G.-C. Roman, andT. C. Bailey, BReliable real-time clinicalmonitoring using sensor networktechnology,[ in Proc. Amer. Med. Inf. Assoc.Annu. Symp., Nov. 2009, pp. 103–107.

[13] O. Chipara, C. Lu, T. C. Bailey, andG.-C. Roman, BReliable patient monitoring:A clinical study in a step-down hospital unit,[Dept. Comput. Sci. Eng., Washington Univ.St. Louis, St. Louis, MO, Tech. Rep.WUCSE-2009-82, Dec. 2009.

[14] CNN, ‘‘Death after two-hour ER wait ruledhomicide,’’ Sep. 2006. [Online]. Available:http://www.cnn.com/2006/US/09/15/er.homicide.ap/index.html.

[15] Coalition for American Trauma Care,‘‘Action needed to bolster nation’s emergencycare system,’’ Jun. 2006. [Online]. Available:http://www8.nationalacademies.org/onpinews/newsitem.aspx?RecordID=06142006.

[16] Crossbow Corporation. MICAz Specifications.[Online]. Available: http://www.xbow.com/Support/Support_pdf_files/MPR-MIB_Series_Users_Manual.pdf.

[17] J. Coughlan and R. Manduchi, BColor targets:Fiducials to help visually impaired people findtheir way by camera phone,[ J. Image VideoProcess., vol. 2007, no. 2, p. 10, 2007.

[18] F. Dabiri, A. Vahdatpour, H. Noshadi,H. Hagopian, and M. Sarrafzadeh,BUbiquitous personal assistive system forneuropathy,[ in Proc. 2nd Int. Workshop Syst.Netw. Support Health Care Assisted LivingEnviron., Breckenridge, CO, 2008,no. article 17.

[19] T. Das, P. Mohan, V. N. Padmanabhan,R. Ramjee, and A. Sharma, BPRISM: Platformfor remote sensing using mobilesmartphones,[ in Proc. 8th Int. Conf. MobileSyst. Appl. Services, San Francisco, CA,Jun. 2010, pp. 63–76.

[20] R. P. Dresher and P. P. Irazoqui, BA compactnanopower low output impedance CMOSoperational amplifier for wireless intraocularpressure recordings,[ in Proc. 29th Annu. Int.Conf. IEEE Eng. Med. Biol. Soc., Aug. 2007,pp. 6055–6058.

[21] A. Dunkels, B. Groonvall, and T. Voigt,BContikiVA lightweight and flexibleoperating system for tiny networked sensors,[in Proc. 29th Annu. IEEE Int. Conf. LocalComput. Netw., Nov. 2004, pp. 455–462.

[22] E. A. Fry and L. A. Lenert, BMASCAL: RFIDtracking of patients, staff and equipment toenhance hospital response to mass casualtyevents,[ in Proc. AMIA Annu. Symp.,Jan. 2005, pp. 261–265.

[23] R. Ganti, P. Jayachandran, T. Abdelzaher, andJ. Stankovic, BSATIRE: A softwarearchitecture for smart AtTIRE,[ in Proc. 4thInt. Conf. Mobile Syst. Appl. Services, Uppsala,Sweden, Jun. 2006, pp. 110–123.

[24] T. Gao, C. Pesto, L. Selavo, Y. Chen, J. Ko,J. Lim, A. Terzis, A. Watt, J. Jeng, B. Chen,K. Lorincz, and M. Welsh, BWireless medicalsensor networks in emergency response:Implementation and pilot results,[ in IEEEInt. Conf. Technol. Homeland Security, 2008,pp. 187–192.

[25] D. Gay, P. Levis, R. von Behren, M. Welsh,E. Brewer, and D. Culler, BThe nesClanguage: A holistic approach to networkedembedded systems,[ in Proc. Programm. Lang.Design Implement., San Diego, CA, Jun. 2003,pp. 1–11.

[26] O. Gnawali, R. Fonseca, K. Jamieson, D. Moss,and P. Levis, BCollection tree protocol,[ inProc. 7th ACM Conf. Embedded Netw. SensorSyst., Berkeley, CA, 2009, pp. 1–14.

[27] M. S. Goodwin, W. F. Velicer, and S. S. Intille,BTelemetric monitoring in the behaviorsciences,[ Behav. Res. Methods, vol. 40, no. 1,pp. 328–341, 2007.

[28] A. Hanjagi, P. Srihari, and A. S. Rayamane,BA public health care information systemusing GIS and GPS: A case study ofShiggaon,[ in GIS for Health and theEnvironment, P. C. Lai and S. H. Mak, Eds.New York: Springer-Verlag, 2007,pp. 243–255.

[29] J. He, H. Li, and J. Tan, BReal-time dailyactivity classification with wireless sensornetworks using hidden Markov model,[ inProc. 29th Annu. Int. Conf. IEEE Eng. Med. Biol.Soc., Aug. 2007, pp. 3192–3195.

[30] J. Hill, R. Szewczyk, A. Woo, S. Hollar,D. Culler, and K. Pister, BSystem architecturedirections for network sensors,’’ ACMSIGPLAN Notices, vol. 35, no. 11, pp. 93–104,Nov. 2000.



[31] T. J. Hodgetts and K. Mackaway-Jones, MajorIncident Medical Management and Support, thePractical Approach. London, U.K.: BMJ,1995.

[32] Hopkins Internetworking Research Group(HiNRG). (2008). CC2420 security supportfor TinyOS 2. [Online]. Available: http://hinrg.cs.jhu.edu/git/?p=jgko/tinyos-2.x.git.

[33] IEEE Standard for InformationTechnologyVTelecommunications andInformation Exchange Between SystemsVLocaland Metropolitan Area Networks, SpecificRequirements-Part 15.4: Wireless MediumAccess Control (MAC) and Physical Layer (PHY)Specifications for Low-Rate Wireless PersonalArea Networks (LR-WPANs), May 2003.[Online]. Available: http://www.ieee802.org/15/pub/TG4.html.

[34] Intel Corporation. (2006). The SHIMMERsensor node platform. [Online]. Available:http://www.eecs.harvard.edu/~mdw/proj/codeblue.

[35] S. Intille. (2007). Enabling population-scalephysical activity measurement on commonmobile phones. NIH GEI project at MIT.[Online]. Available: http://www.gei.nih.gov/exposurebiology/program/docs/StephenIntille.pdf.

[36] S. S. Intille, K. Larson, E. M. Tapia,J. S. Beaudin, P. Kaushik, J. Nawyn, andR. Rockinson, BUsing a live-in laboratory forubiquitous computing research,’’ in LectureNotes in Computer Science, vol. 3968. Berlin,Germany: Springer-Verlag, 2006, p. 349.

[37] ISO 9919:2005 Medical ElectricalEquipment-Particular Requirements for theBasic Safety and Essential Performance of PulseOximeter Equipment for Medical Use,Publication of the ASTM F29.11.05 andISO TC 121 SC3 working group on pulseoximeters, Mar. 2005.

[38] S. Katz, A. B. Ford, R. W. Moskowitz,B. A. Jackson, and M. W. Jaffe, BStudies ofillness in the aged: The index of ADL: Astandardized measure of biological andpsychosocial function,[ J. Amer. Med. Assoc.,vol. 185, no. 12, p. 914, 1963.

[39] O. A. Khan and R. Skinner, Eds., GeographicInformation Systems and Health Applications.Hershey, PA: IGI Global, 2002.

[40] C. D. Kidd, R. Orr, G. D. Abowd,C. G. Atkeson, I. A. Essa, B. MacIntyre,E. Mynatt, T. E. Starner, and W. Newstetter,BThe aware home: A living laboratory forubiquitous computing research,’’ in LectureNotes in Computer Science. Berlin, Germany:Springer-Verlag, 1999, pp. 191–198.

[41] J. Ko, T. Gao, R. Rothman, and A. Terzis,BWireless sensing systems in clinicalenvironments: Improving the efficiency of thepatient monitoring process,[ IEEE Eng. Med.Biol. Mag., vol. 29, no. 2, pp. 103–109,Mar./Apr. 2010.

[42] J. Ko, T. Gao, and A. Terzis, BEmpirical studyof a medical sensor application in an urbanemergency department,[ in Proc. ICST 4th Int.Conf. Body Area Netw., Los Angeles, CA,Apr. 2009, article 10.

[43] J. Ko, J. Lim, Y. Chen, R. Musaloiu-E.,A. Terzis, G. Masson, T. Gao, W. Destler,L. Selavo, and R. Dutton, BMEDiSN: Medicalemergency detection in sensor networks,[ACM Trans. Embedded Comput. Syst., vol. 10,no. 1, pp. 11:1–11:29, 2010, article 11.

[44] J. Krumm, BInference attacks on locationtracks,[ in Proc. 5th Int. Conf. PervasiveComput., 2007, pp. 127–143.

[45] S. Kumar, ‘‘Autosense,’’ NIH GEI project atThe University of Memphis, 2007. [Online].

Available: http://www.cs.memphis.edu/~santosh/AutoSense.html.

[46] L. Liao, D. Fox, and H. Kautz, BLocation-basedactivity recognition using relationalMarkov networks,[ in Proc. 19th Int. JointConf. Artif. Intell., Edinburgh, Scotland, 2005,pp. 773–778.

[47] M. D. Linderman, G. Santhanam,C. T. Kemere, V. Gilja, S. O’Driscoll, B. M. Yu,A. Afshar, S. I. Ryu, K. V. Shenoy, andT. H. Meng, BSignal processing challenges forneural prostheses,[ IEEE Signal Process. Mag.,vol. 25, no. 1, pp. 18–28, Dec. 2008.

[48] K. Lorincz and M. Welsh, BMoteTrack: Arobust, decentralized approach to RF-basedlocation tracking,[ Personal UbiquitousComput., vol. 11, no. 6, pp. 489–503, 2007.

[49] K. Lorincz, B. Chen, G. W. Challen,A. R. Chowdhury, S. Patel, P. Bonato, andM. Welsh, BMercury: A wearable sensornetwork platform for high-fidelity motionanalysis,[ in Proc. 7th ACM Conf. EmbeddedNetw. Sensor Syst., 2009, pp. 353–366.

[50] D. Malan, T. Fulford-Jones, M. Welsh, andS. Moulton, BCodeBlue: An ad hoc sensornetwork infrastructure for emergency medicalcare,[ in Proc. MobiSys/ Workshop Appl. MobileEmbedded Syst., Jun. 2004, pp. 12–14.

[51] F. Michahelles and B. Schiele, BSensing andmonitoring professional skiers,[ IEEEPervasive Comput., vol. 4, no. 3, pp. 40–46,Jul.–Sep. 2005.

[52] E. Miluzzo, N. D. Lane, K. Fodor, R. Peterson,H. Lu,M.Musolesi, S. B. Eisenman, X. Zheng,and A. T. Campbell, BSensing meets mobilesocial networks: The design, implementationand evaluation of the cenceme application,[ inProc. 6th ACM Conf. Embedded Netw. SensorSyst., Raleigh, NC, 2008, pp. 337–350.

[53] MoteIV Corporation. (2007, May). TmoteMini. [Online]. Available: http://blog.moteiv.com/archives/2007/05/introducing_tmo.php.

[54] MSNBC News. (2006, Nov.). Tired ofWaiting for the Doctor? You Are Not Alone.[Online]. Available: http://www.msnbc.msn.com/id/15487676/page/2/.

[55] M. Mun, S. Reddy, K. Shilton, N. Yau,J. Burke, D. Estrin, M. Hansen, E. Howard,R. West, and P. Boda, BPEIR, the personalenvironmental impact report, as a platformfor participatory sensing systems research,[ inProc. 7th Int. Conf. Mobile Syst. Appl. Services,New York, 2009, pp. 55–68.

[56] P. Ohm, BBroken promises of privacy:Responding to the surprising failure ofanonymization,[ UCLA Law Rev., vol. 57,no. 6, 2010. [Online]. Available: http://ssrn.com/abstract=1450006.

[57] S. Patel, K. Lorincz, R. Hughes, N. Huggins,J. H. Growdon, M. Welsh, and P. Bonato,BAnalysis of feature space for monitoringpersons with Parkinson’s disease withapplication to a wireless wearable sensorsystem,[ in Proc.29th IEEE Eng. Med. Biol. Soc.Annu. Int. Conf., Aug. 2007, pp. 6290–6293.

[58] K. Patrick, ‘‘A tool for geospatial analysis ofphysical activity: Physical activity locationmeasurement system (PALMS),’’ NIH GEIproject at the University of California atSan Diego, 2007. [Online]. Available: http://www.gei.nih.gov/exposurebiology/program/docs/KevinPatrick.pdf.

[59] D. Patterson, D. Fox, H. Kautz, andM. Philipose, BExpressive, tractable andscalable techniques for modeling activities ofdaily living,[ in Proc. 2nd Int. WorkshopUbiquitous Comput. Pervasive Healthcare Appl.,2003.

[60] A. Pentland, BHealthwear: Medicaltechnology becomes wearable,[ Computer,vol. 37, no. 5, pp. 42–49, May 2004.

[61] J. Polastre, R. Szewczyk, and D. Culler,BTelos: Enabling ultra-low power wirelessresearch,[ in Proc. 4th Int. Conf. Inf. Process.Sensor Netw., Apr. 2005, pp. 364–369.

[62] S. Raghunathan, M. P. Ward, K. Roy, andP. P. Irazoqui, BA low-power implantableevent-based seizure detection algorithm,[ inProc. 4th Int. IEEE/EMBS Conf. Neural Eng.,May 2, 2009, pp. 151–154.

[63] S. Reddy, A. Parker, J. Hyman, J. Burke,D. Estrin, and M. Hansen, BImage browsing,processing, and clustering for participatorysensing: Lessons from a dietsense prototype,[in Proc. 4th Workshop Embedded Netw. Sens.,New York, 2007, pp. 13–17.

[64] A. Salarian, H. Russmann, F. J. G. Vingerhoets,C. Dehollain, Y. Blanc, P. R. Burkhard, andK. Aminian, BGait assessment in Parkinson’sdisease: Toward an ambulatory system forlong-term monitoring,[ IEEE Trans. Biomed.Eng., vol. 51, no. 8, pp. 1434–1443, Aug. 2004.

[65] J. D. Schenker, S. Goldstein, J. Braun,A. Werner, F. Buccellato, G. Asaeda, andD. J. Prezant, BTriage accuracy at amultiple casualty incident disaster drill:The emergency medical service, firedepartment of New York City experience,[J. Burn Care Res., vol. 27, no. 5, p. 570, 2006.

[66] K. Shilton, J. Burke, D. Estrin, M. Hansen,R. Govindan, and J. Kang, BDesigning thepersonal data stream: Enabling participatoryprivacy in mobile personal sensing,[Cntr. Embedded Netw. Sens., Univ.California Los Angeles, Tech. Rep., 2009.

[67] V. Srinivasan, J. Stankovic, andK. Whitehouse, BProtecting your dailyin-home activity information from a wirelesssnooping attack,[ in Proc. 10th Int. Conf.Ubiquitous Comput., Seoul, Korea, 2007,pp. 202–211.

[68] P. Stark, ‘‘Press release: Stark openingremarks at emergency care hearing,’’Jul. 2006. [Online]. Available: http://www.house.gov/stark/news/109th/pressreleases/20060727_emergencycare.htm.

[69] A. M. Summer, BThe silver tsunami: Oneeducational strategy for preparing to meetAmerica’s next great wave of underserved,[J. Health Care Poor Underserved, vol. 18, no. 3,p. 503, 2007.

[70] G. Super, S. Groth, and R. Hook, ‘‘START:Simple triage and rapid treatment plan,[Hoag Memorial Hospital Presbyterian,Newport Beach, CA, 1994.

[71] Texas Instruments. (2006). 2.4 GHz IEEE802.15.4/ZigBee-ready RF transceiver.[Online]. Available: http://www.ti.com/lit/gpn/cc2420.

[72] G. Virone, A. Wood, L. Selavo, Q. Cao,L. Fang, T. Doan, Z. He, and J. A. Stankovic,BAn advanced wireless sensor network forhealth monitoring,[ in Proc. TransdisciplinaryConf. Distrib. Diagnosis Home Healthcare,Apr. 2006, pp. 95–100.

[73] M. Visintin, H. Barbeau, N. Korner-Bitensky,and N. E. Mayo, BA new approach to retraingait in stroke patients through body weightsupport and treadmill stimulation,[ Stroke,vol. 29, no. 6, pp. 1122–1128, Jun. 1998.

[74] C. B. Wilson, BSensors in medicine,[ WesternJ. Med., vol. 171, no. 5–6, pp. 322–325, 1999.

[75] A. Wood, L. Selavo, and J. Stankovic, BSenQ:An embedded query system for streaming datain heterogeneous interactive wireless sensornetworks,[ in Proc. 4th IEEE Int. Conf. Distrib.Comput. Sensor Syst., Santorini Island, Greece,Jun. 2008, pp. 531–543.



[76] A. Wood, J. Stankovic, G. Virone, L. Selavo,Z. He, Q. Cao, T. Doan, Y. Wu, L. Fang, andR. Stoleru, BContext-aware wireless sensornetworks for assisted living and residentialmonitoring,[ IEEE Network, vol. 22, no. 4,pp. 26–33, Jul./Aug. 2008.

[77] A. D. Wood, L. Fang, J. A. Stankovic, andT. He, BSIGF: A family of configurable, secure

routing protocols for wireless sensornetworks,[ in Proc. 4th ACM WorkshopSecurity Ad Hoc Sensor Netw., Alexandria, VA,2006, pp. 35–48.

[78] W. Wu, L. Au, B. Jordan, T. Stathopoulos,M. Batalin, W. Kaiser, A. Vahdatpour,M. Sarrafzadeh, M. Fang, and J. Chodosh,BThe smartcane system: An assistive device

for geriatrics,[ in Proc. ICST 3rd Int. Conf. BodyArea Netw., Tempe, AZ, 2008, article 2.

[79] P. Yager, T. Edwards, E. Fu, K. Helton,K. Nelson, M. R. Tam, and B. H. Weigl,BMicrofluidic diagnostic technologies forglobal public health,[ Nature, vol. 442,no. 7101, pp. 412–418, 2006.

ABOUT THE AUTHORS

JeongGil Ko received the B.Eng. degree in com-

puter science and engineering from Korea Univer-

sity, Seoul, Korea, in 2007 and the M.S.E. degree in

computer science from the Johns Hopkins Univer-

sity, Baltimore, MD, in 2009, where he is currently

working towards the Ph.D. degree at the Depart-

ment of Computer Science.

His research interests include wireless medical

sensing systems, low-power embedded network

system and protocol design, and the deployment

of such embedded systems to real environments.

Mr. Ko is a recipient of the Abel Wolman Fellowship awarded by the

G.W.C. Whiting School of Engineering at the Johns Hopkins University.

Chenyang Lu received the B.S. degree from the

University of Science and Technology of China,

Hefei, Anhui, China, in 1995, the M.S. degree from

the Chinese Academy of Sciences, Beijing, China,

in 1997, and the Ph.D. degree from the University

of Virginia, Charlottesville, in 2001, all in com-

puter science.

He is an Associate Professor of Computer

Science and Engineering at Washington University

in St. Louis, St. Louis, MO. He authored and

coauthored more than 100 publications. His research interests include

real-time embedded systems, wireless sensor networks, and cyber-

physical systems.

Prof. Lu received a National Science Foundation (NSF) CAREER Award

in 2005 and a Best Paper Award at the International Conference on

Distributed Computing in Sensor Systems (DCOSS) in 2006. He is an

Associate Editor of the ACM Transactions on Sensor Networks, Real-Time

Systems, and the International Journal of Sensor Networks, and served as

Guest Editor of the Special Issue on Real-Time Wireless Sensor Networks

of Real-Time Systems and Special Section on Cyber-Physical Systems and

Cooperating Objects of the IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS.

He also served on the organizational committees of numerous confer-

ences such as Program and General Chair of the IEEE Real-Time and

Embedded Technology and Applications Symposium (RTAS) in 2008 and

2009, Track Chair on Sensor Networks of the IEEE Real-Time Systems

Symposium (RTSS) in 2007 and 2009, Program Chair of the International

Conference on Principles of Distributed Systems (OPODIS) in 2010, and

Demo Chair of the ACM Conference on Embedded Networked Sensor

Systems (SenSys) in 2005. He is a member of the Executive Committee of

the IEEE Technical Committee on Real-Time Systems.

Mani B. Srivastava received the B.Tech. degree

from the Indian Institute of Technology (IIT),

Kanpur, India, in 1985 and the M.S. and Ph.D.

degrees from the University of California at

Berkeley, Berkeley, in 1987 and 1992, respectively.

After working for five years at Bell Labs

Research, Murray Hill, NJ, he joined the University

of California at Los Angeles, Los Angeles, faculty in

1997 where he is currently Professor and Vice Chair

(Graduate Affairs) in Electrical Engineering, and

Professor of Computer Science. He is also affiliated with the National

Science Foundation (NSF) Science and Technology Center on Embedded

Networked Sensing where he co-leads the System Research Area. His

research interests are in embedded systems, low-power design, wireless

networking, and pervasive sensing.

Dr. Srivastava received the President of India s Gold Medal in 1985, the

NSF Career Award in 1997, and the Okawa Foundation Grant in 1998. He

currently serves as the Editor-in-Chief of the ACM Sigmobile Mobile

Computing and Communications Review, and as an Associate Editor for

the ACM/IEEE TRANSACTIONS ON NETWORKING, and the ACM Transactions on

Sensor Networks.

John A. Stankovic (Fellow, IEEE) received the Ph.D.

degree from Brown University, Providence, RI.

He is the BP America Professor in the Computer

Science Department, University of Virginia,

Charlottesville. He served as Chair of the depart-

ment for eight years. Before joining the University

of Virginia, he taught at the University of

Massachusetts where he won an outstanding

scholar award. He has also held visiting positions

in the Computer Science Department at Carnegie

Mellon University, Pittsburgh, PA, at INRIA, France, and Scuola Superiore

S. Anna, Pisa, Italy. His research interests are in real-time systems,

distributed computing, wireless sensor networks, and cyber physical

systems. He has built three wireless sensor networks: VigilNet, a military

surveillance system funded by Darpa and now being constructed by

Northrup-Gruman, Luster, an environmental science system for measur-

ing the effect of sunlight on plant growth, and AlarmNet, an emulation of

home health care.

Dr. Stankovic is a Fellow of the Association for Computing Machinery

(ACM). He won the IEEE Real-Time Systems Technical Committee’s Award

for Outstanding Technical Contributions and Leadership. He also won the

IEEE Technical Committee on Distributed Processing’s Distinguished

Achievement Award (inaugural winner). He has won four Best Paper

awards, including one for ACM SenSys 2006. He is ranked among the top

250 highly cited authors in CS by Thomson Scientific Institute. He has

given more than 23 Keynote talks at conferences and many Distinguished

Lectures at major Universities. He also served on the Board of Directors

of the Computer Research Association for nine years. He was the Editor-

in-Chief for the IEEE TRANSACTIONS ON DISTRIBUTED AND PARALLEL SYSTEMS

and was Founder and Co-Editor-in-Chief for the Real-Time Systems

Journal.

Andreas Terzis received the Ph.D. degree in

computer science from the University of California

at Los Angeles, Los Angeles.

He is an Associate Professor at the Department

of Computer Science, Johns Hopkins University,

Baltimore, MD, where he leads the Hopkins

interNetworking Research Group (HiNRG). His

research interests include the broad area of

wireless sensor networks, including protocol de-

sign, system support, and data management.

Dr. Terzis is a recipient of the National Science Foundation (NSF)

CAREER award.



Matt Welsh received the B.S. degree from Cornell

University, Ithaca, NY and the M.S. and Ph.D.

degrees from the University of California at

Berkeley, Berkeley.

He is currently a Gordon McKay Professor of

Computer Science at the Harvard School of

Engineering and Applied Sciences, Cambridge,

MA. His research interests involve operating

system, network, and language support for com-

plex distributed systems.

Dr. Welsh is a Senior Member of the Association for Computing

Machinery and serves on the Editorial Board for the ACM Transactions on

Sensor Networks.



invited paper wirelesssensornetworks forhealthcarejgko/papers/proc10.pdf · invited paper...

Documents