Pervasive Healthcare Overview A Personal Perspective (May 2007)

Save this PDF as:

Size: px
Start display at page:

Download "Pervasive Healthcare Overview A Personal Perspective (May 2007)"


1 1 Pervasive Healthcare Overview A Personal Perspective (May 2007) Sabrina Nefti Abstract The latest outstanding advances in mobile network technologies, wireless communications and medical sensors are propelling a new and exciting research for 21 st century pervasive healthcare systems. Such systems have the capacity to provide health services for patients regardless of time and place. The aim of this paper is to overview the state of the art pervasive healthcare technologies by exploring the conducted case studies. The paper conceptually maps the existing technologies into wearable and environmental systems with the aim of identifying, comparing and analysing the current challenges and limitations of each system. Finally, the paper concludes by shedding light on the future trends and thrusts that will direct research in this field. Index Terms Pervasive, ubiquitous, personalised healthcare, mobile health, pervasive technology, wearable devices, body area network, smart homes, independent lifestyle, wellbeing monitoring, remote health surveillance. T I. INTRODUCTION oday, the developed world is benefiting from extending life spans thanks to the outstanding medical developments. This in turn is fuelling an immense demographic shift towards an aging population. To set this into context, the UK population aged over 65 is estimated to rise from 9.3 million to 16.8 million over the next 50 years [12]. The downside of this trend is significant; the healthcare system is struggling to meet the costs of the ever-increasing health services; with 40% of its budget exclusively dedicated to people over 60 [13]. Another major outcome is the increasing prevalence of chronic diseases typically appearing with age such as diabetes and cardiac vascular disorder [4]. Statistics in various countries showed that diabetes only accounts for between 2.5 and 15% of the total healthcare spending [5]. These challenges can be mitigated if the paradigm of care is shifted from hospitals to homes. To make this vision true, a wealth of research has begun giving birth to an exciting research discipline: Pervasive healthcare. Pervasive healthcare focuses on utilising pervasive technology to produce solutions for remote healthcare. Such systems will not only reduce workload in hospitals but will also allow patients to lead an independent lifestyle in their own home. Given such exciting vistas, this paper is dedicated to investigate the state of the art developments in this field. Firstly, the terminology used throughout the paper is defined. This is followed by a description of the generic platform architecture dominating pervasive healthcare systems. The paper focuses on mapping the existing technologies by discussing and evaluating the most prominent case studies. Finally, it concludes with an outlook of research thrusts that will steer developments in this field. II. PRELIMINARIES There does not appear to be an agreed definition of ubiquitous systems or pervasive computing as these terms have been used interchangeably throughout literature. This paper defines both ubiquitous and pervasive systems as the set of coherent technologies and devices that have the capabilities to be present everywhere at once. The vision of such technologies was first articulated by Mark Weiser in 1991 [49]. Ahead of his time, Weiser envisaged environments which are intensively equipped with pervasive devices with communication capabilities and yet gracefully amalgamated with human users. Such a vision could not turn into reality then due to inexistent enabling technologies. Today, the major advancements in wireless technologies, mobile and sensing devices have made the quest for Weiser s vision possible. One of the major applications of pervasive technology has been in the health domain resulting in pervasive health, also known as mobile health (m-health) or personalised health (phealth). Ubiquitous computing has made inroads into health through two avenues: wearable and environmental systems. Wearable health systems can be broadly defined as the infrastructure of mobile electronic devices and sensors that are in constant physical contact with the user in forms of nonobtrusive clothing or accessories. Vital physiological parameters such as Electrocardiogram (ECG) and blood pressure are gathered and communicated to remote health centres. Environmental systems consist of settings whereby the environment of the patient is equipped with a distributed network of sensors which collaborate to collect various data related to the patient s wellbeing or health. III. PLATFORM ARCHITECTURE FOR M-HEALTH SYSTEMS The wealth of research carried out in ubiquitous health has resulted in a plethora of architectures for m-health systems. The key features of these platforms are concluded in what follows. Typically, a pervasive health system relies on a sensor node consisting of three components: the sensor, the signal processing and the transmission modules (Fig. 1).

2 2 systems which funnel towards the physiological aspect. Fig.1. Generic platform architecture used in pervasive healthcare systems consisting of sensors, signal processing and transmission modules. Medical sensors are one of the up-and-coming technologies [35]. They constitute the link between the physical world, the patient, and the digital world, the systems interpreting the data. Two types of sensors are employed: wearable and environmental. Wearable sensors, also known as biosensors, capture physiological phenomenon, such as muscle activity and blood flow. Paper [33] reviews the recent advancements in biosensors. Environmental sensors measure parameters such as temperature and movement. In both sensors, the physical signal is transformed into an electrical signal, amplified and digitised and sent to the signal processing unit. The latter, typically implemented in a microprocessor or microcontroller, performs the high-level data processing which involves analysing the data to detect abnormal disease situations and possibly creating alerts. Processed data is sent to a remote health centre by the transmission module. This unit can either be implemented in PC with internet connection or a mobile device with wireless capabilities such as mobiles and PDAs. This architecture offers numerous advantages. Communication is the most power consuming part of the sensor node. By restricting the transmission of data to alerts, the lifetime of the sensor is prolonged and the workload of health professionals is significantly reduced. In an attempt to incorporate the three units in one platform, the University of Alabama designed a Wireless Intelligent SEnsor (WISE) [39]. WISE is based on Texas Instrument microcontroller and LINX wireless transceiver. However, the signal processing capabilities of WISE is limited to basic alerts. Many variations to this architecture exist. These will be highlighted throughout the paper along with various intelligent algorithms used for signal processing. IV. PERVASIVE HEALTHCARE: ENVIRONMENTAL SYSTEMS (CATEGORY I) The home was determined to be ideal for monitoring patients and elderly to enable these two categories to live independently and safely in their own homes. Research in the field of smart homes has been steered by two scientific facts. Firstly, elderly people are more prone to cognitive decline, and secondly, the prevalence of chronic diseases augments with age [45]. Therefore, this paper will map the existing environmental technologies into two streams: systems which ensure the wellbeing of elders by monitoring whether daily activities such as eating and washing are performed; and health-monitoring A. Environmental: Wellbeing Systems (Stream 1) An intuitive technology for home-based monitoring is image sensing. This paradigm was explored in two main models: the Digital Family Portrait [32] and the NetCare projects [9]. Information is collected using image sensors and transmitted via internet to a GPRS-enabled digital frame which presents details about the elder s activities (Fig. 2). The frame is either owned by a family member [32] or a caregiver [9]. While the Digital Family Portrait emphasises on the social aspect, CareNet provides a mechanism to the care network to coordinate their care-related efforts. Although case studies on the NetCare systems were positive [9], a survey conducted in [31] revealed that elderly people were anxious about disclosure of their privacy and generally demanded that data is only sent to one or two family members. Fig.2. Digital family portrait activity monitor. (Sourced from Mynatt [32], Figure 4). Another stream of research funneled towards maximising the residents privacy. The un-intrusive activity monitor suggested for the Ubiquitous Room (U-Room) employs magnetic switches attached to doors to monitor walking activity [7]. The voltage switching frequency (3V when door is opened and 0V when closed) is used to verify whether the elder is performing regular exercise. Data is sent to a Bluetooth access point and then to a PDA device and finally conveyed to a remote home server via WLAN. Switch systems do not provide conclusive indication of activity, as an elderly person suffering dementia may open the door once and forget to close it. A more accurate approach was suggested in the Anxious House where switches were employed with additional sensors to increase accuracy such as pressure mats, current and flow sensors, all communicating via X10 protocol [50]. Unlike [7], information is not continuously sent to the remote server. A probabilistic approach is used to learn the elder s activity pattern and infer normal and abnormal activities. Each device in the house is associated with a statistical model based on the frequency of use of the device by the resident. Therefore, caregivers are only alerted in abnormal situations: such as when the interval to check on a meal takes more than a specific time. The dogma of learning and inferring activities was also adopted in BT Ubiquitous House where

3 3 Bayesian networks and fuzzy logic techniques were used [11]. Another well-established technology for activity monitoring is Radio Frequency IDentification (RFID). In a typical RFID system, high-power antennas send a signal to battery-free devices called tags. The latter reply with a unique identification encoded in a modulated signal [43]. Unlike switches which are limited to doors, tags can be attached to various objects. For instance, in the Wireless Identification and Sensing Platform (WISP) project a variety of household objects were equipped with tags [38]. However, the WISP system can only provide discrete monitoring since information can only be obtained when the user makes contact with objects. A continuous, RFbased system was proposed in [15] for an early diagnosis of cognitive impairment. The system monitored the change in walking speed using only one motion sensor in a room, with additional contact sensors attached to the door. Both sensors send RF signal to a transceiver attached to a serial port of a computer in which the collected data is analysed using learning algorithms such as Kernel Density. Although RFID can be omnipresent in any object, it can not always guarantee accuracy. For instance, a RFID attached to a toilet seat can measure the frequency of toilet use but can not infer whether any toilet activities have actually occurred. This particular issue was addressed in [38] whereby a sound system based on high accuracy microphones was used to capture the sounds occurring during toilet activities. Microphones send data to a computer port, which processes information using Hidden Markov Models. Although the model was proposed for home use, it can easily be augmented with wireless capabilities for remote monitoring. Acoustic technology was also used in U- Room [7] to detect snoring via electrets condenser microphones. While the previous models investigated normal daily activities, others explored technologies for detecting home accidents such as falls; one of the most hazardous obstacles to independent living of elderly people. A home-based fall detection system was proposed in [46]. The system consists of Infra-Red Integrated SYStem (IRISYS) thermal imaging sensors mounted on a wall, a separate detector unit and a home PC. IRISYS sensors send information about the thermal target, such as velocity and location, to the detector unit where a dedicated processor applies neural network and perceptron algorithms to identify falls as well as long periods of inactivity. While this system can detect falls, it can not foresee them. Toshio et al proposed a system able to anticipate fall accidents in a nursing home. The detection is based on placing a badge consisting of an ultrasonic emitter and RF receiver on the wheelchair of the elder, and installing numerous ultrasonic receivers and an RF transmitter in the ceiling [18]. The RF transmitter sends an ID to all the RF receivers on the badges. Only the badge with the corresponding ID replies with ultrasonic pulses back to the ultrasonic receivers on the ceiling. The distance between the chair and the sensor is inferred from the difference between transmission and reception time. If the chair is located in areas where the elder is prone to fall off the chair such as bedside and toilet, the system alerts the caregivers. In a more recent and holistic laboratory-house, Gator House, the paradigm of monitoring elder s activity was taken a step further by incorporating systems capable of aiding elders to perform daily activities such as smart wardrobes that can suggest suitable clothes [16]. Gator uses a large sensor network integrated by a software-based platform installed in a home PC where each sensor is controlled by a Java program. Sensors communicate with the PC wirelessly such that whenever a sensor is activated, its corresponding monitoring Java code is started. Another ongoing similar research is the Aware Home laboratory-house of the Georgia Institute of Technology. The major assistive system introduced by the Aware Home is the medication monitor consisting of an RFID reader connected to a display screen and RFID tags attached to the medication. When a drug is touched, the RFID tag informs the RFID reader, and the display alerts the user in abnormal situations such as wrong type of medication [4]. B. Environmental: Health Monitoring Systems (Stream 2) Less research was performed in investigating environmental health surveillance systems. Ishima proposed an un-obtrusive method to measure ECG [20] and respiration [21] in bed. To this end, conductive electrodes are designed from conductive thread and woven into pillows and bed sheets at critical contact points such as the legs, the shoulders and the chest. ECG signal was extracted from the pillow and leg electrodes while respiratory activity was inferred from the capacitive difference between the shoulder and the chest electrodes. Bathtub ECG measurement was also obtained in [22] by placing silver and silver chloride electrodes on the wall of the bathtub near the contact areas of shoulders and toes. Respiratory rate measurement in bathtub was also suggested in [34]. The sensing mechanism consisted of placing a probe on the bottom surface of a bathtub to establish physical contact with the lower part of the body to obtain photoplethysmography (PPG) signal from which respiration rate is extracted. One of the earliest house models that initiated using these sensor developments was the Welfare Techno House developed in Takaoka, Toyama, Japan [26]. The house was equipped with bed and bathtub ECG sensors and a weight platform consisting of four load sensors installed near the toilet bowl for unconscious weight measurement. However, sensors data is collected and analysed in a wired LAN of home PCs. Wired systems are ergonomically impractical and may annoy patients. In a more recent model, U-Room [7], similar ECG and weight sensors are augmented with the Bluetooth wireless communication capability [7]. Load cells, installed in the four legs of the bed, are not only used for weight measurement but also for monitoring sleeping movements. A very recent attempt investigated the use of electrodes attached to the back of a chair to unobtrusively obtain ECG (Fig. 3) [27].

4 4 Fig.3. Un-obtrusive, non-invasive ECG measurement in a chair. (Sourced from Lim [27], Figure 1). Similar to [7], the electrodes were coupled with very highimpedance amplifiers to compensate for the insulating effects of the clothes. While Choi et al in [7] proved that the use of amplifiers resulted in an accuracy comparable to the one obtained from fixed-on-body electrodes; Lim et al experiments showed that measurement was less accurate due to clothes properties. This difference may be due to the fact that in [7], the type of sleeping clothes was assumed to be fixed and hence amplifiers were designed accordingly, contrary to [27] in which the chair was designed for convenient everyday use and therefore all types of clothes were assumed. C. Evaluation of Environmental Systems Most studies on pervasive environmental systems do not report data about field trials. Therefore, this paper determines four crucial criteria against which the models are assessed: Privacy: measures the commitment of the system in protecting the personal, health and mental status of the patient. Practicality: measures the degree of difficulty or ease required to install, deploy, or use the system including ergonomic factors. Cost: the system s cost considering hardware and software. Reliability: measures the extent to which the system can deliver accurately and consistently. Table I evaluates the environmental systems. Based on the analysis presented in Table I, six predominant obstacles to environmental systems are identified and discussed as follows. 1) Privacy Privacy is a major matter of concern when monitoring within the confines of a private home. In RFID systems only the tags ID is transmitted. Compared to image sensors where detailed private information is transmitted, RFID and infra-red technologies are clearly less intrusive. Also field trials show that infra-red and RFID enjoy higher acceptability. This shows that there is a strong correlation between patience acceptance and privacy, proving that users are more comfortable when minimum data transmission is involved. 2) Measurement Coverage and Continuity: Switches are restricted to doors and can only provide discrete monitoring. RFID is capable of providing more continuous monitoring since various objects can be tagged. However, for RFID to be accurate, a few thousands object need to be tagged. Although the cost of a single RFID tag is only $0.5, a comprehensive installation may become cost prohibitive. On the other hand, Infra-red provides continuous location tracking but does not provide further information on what the person is doing when long inactivity periods are noticed such as when the person is watching TV. This information could be concluded, for example, if the remote controller was tagged with RFID. This demonstrates that continuous environmental monitoring requires the amalgamation of various technologies to reduce the discrete nature of measurement. However, this poses a trade-off between cost and accuracy. TABLE I EVALUATION OF ENVIRONMENTAL TECHNOLOGIES Model and Advantages Disadvantages Technology NetCare: Image sensors [9] U-Room: Switch [7] Fall Detection: Infra-red technology [46] Fall prediction system: Ultrasound technology [18] WISP platform: RFID system [38] Gator house: Various contact sensors [16] Bathtub ECG: Ag/AgCl electrodes [22] Bed ECG: Textile Electrodes [7] ECG in chair: Active electrodes [27] Practicality: intuitive and friendly user interface [10] Trial: Positive feedback Reliability: successful Cost: low-cost Cost: cost-effective Trial: Positive Privacy: does not report high-level of details Privacy: Only badge s ID is transmitted Reliability: acceptable with some reported errors Privacy: only objects IDs are transmitted Practicality: Batteryfree tags are practical for elders Cost: affordable Trial: on-going Reliability: quality of ECG is acceptable Reliability: Signals obtained 97% of sleeping time Practicality: One chair can be used for many patients (if the chair can distinguish between many users) Privacy: elders are concerned on who receives the data Reliability: Display downtime problems with the GPRS card Practicality: requires PDA installation in each room Reliability: Detection of 30% of falls only Practicality: The system relies on the assumption that all elders use wheelchairs Reliability: Inaccurate monitoring. For example, the use of a fork does not guarantee that eating has occurred as the fork could have been used for cooking only Cost: Combined software and hardware platforms are costly for extensive network Practicality: The person must sit according to the position of the sensors Reliability: Wear and tear reduces measurement accuracy. Sheets need to be regularly replaced Reliability: Inaccurate ECG obtained 3) Identification and Personalisation: While this is not an issue when direct imaging information is transmitted or a personal badge is attached to a wheelchair, systems using switches, RFID or infra-red clearly fail to deliver correct measurement when more than one person is at home, for example, when the elder receives a visitor.

5 5 4) Suitability for Chronic Disease Management Despite the considerable research efforts to incorporate unobtrusive biosensors in various objects (bathtubs, bed sheets, toilet seats and chairs) vital sign measurement remains fundamentally dependent on the proximity or the physical contact of the person with these objects. Such approach is impractical for individuals with chronic diseases, such as severe heart conditions, which necessitate continuous and accurate surveillance. 5) Interaction with Technology If the technology developed for elderly houses is to be successfully adopted, it needs to be low-complexity and lowmaintenance. For instance, the NetCare system included an option to control the propagation of information by either turning off the transmission or by limiting the sharing of an activity to specific users. Yet, trials conducted in [9] revealed that this option was not used by elders. This may be due to the elderly lack of understanding of how technology worked which resulted in the non-utilisation of the full system s functionality. 6) Signal Processing The accuracy of intelligent systems depends on the performance of algorithms, such as neural networks, which learn normal patterns and detect abnormal situations. However, these algorithms depend on the amount of training/history data provided to them [14]. Therefore, any errors in the algorithm judgment may result in false alarms leading to inappropriate intervention and an intrusion in the elder s privacy. Therefore, intensive verification is a precondition before such systems can be reliably deployed. Fig.4. Generic platform architecture employed in wearable systems. A. Wearable: Wellbeing Systems (Stream 1) These systems mainly focus on two applications: determining the person s location and recognising their activity. An activity recognition system, PROACT, which deduces daily activity from the interaction with objects tagged with a RFID technology was proposed in [37]. The user is required to wear a glove equipped with a RFID reader (Fig. 5). The touched object s ID is detected by the glove s RFID reader which then sends it to HPiPaq wearable computer. The latter stores and forwards the data to a home PC in which data is analysed by a probabilistic software. V. PERVASIVE HEALTHCARE: WEARABLE SYSTEMS (CATEGORY II) In a similar fashion to environmental systems, this paper classifies wearable solutions into wellbeing and health monitoring systems. The generic architecture of wearable systems is first outlined prior to discussing case studies. The various technologies adopted in wearable devices were generally assembled in a common model: Wireless Body Area Network (WBAN) shown in Fig. 4. The main component of the WBAN is the network of biosensors such as ECG sensors, Electromyogram (EMG) sensors for muscle activity and Electroencephalogram (EEG) sensors for brain activity. The location of the sensors depends on the end-user application. Sensors in the WBAN are augmented with processing power to enable them to process signals in real-time. Inter-network communication is implemented via various short-range communication protocols. A personal server; typically implemented on an off-shelf internet-enabled device; controls the network and secures external communication with a remote medical centre. Fig.5. Glove with RFID reader and battery. (Sourced from Philipose [37], Figure 1). The main flaw in PROACT is that lightweight tag sensors are placed in the environment while heavier systems (battery and antenna) are carried by the user. An opposite approach was adopted in the system proposed to track the elder s location in [24]. Only two lightweight ultrasonic tags (transmitters) of type Hexamite s Low Cost Positioning System (HLCPS) are attached on the person s shoulders, while ultrasonic receivers are positioned on the ceiling. Ross et al take the dogma of localisation a step further by designing an infrared tracking and guidance system for elderly with vision loss [42]. Location detection is performed by placing infrared transmitters (IR Locust, MIT Media Lab) along the house and incorporating receivers in the user s headphones. Messages received from the transmitter are converted in the headphones into walking instructions. B. Wearable: Health Monitoring Systems (Stream 2) In this paper, wearable health solutions are categorised into: systems providing generic monitoring for post-hospitalisation

6 6 patients and others allowing intensive monitoring for chronic disease management. 1) Wearable for Patients Monitoring A body area network was designed in [25] to monitor stress levels by measuring Heart Rate Variability (HRV) using various sensors as shown in Fig. 6. Sensors wirelessly transmit data to a wearable process server specifically designed for this system. A separate PDA device uploads processed data from the personal server using 900MHz RF protocol and sends it to internet using Bluetooth. Clearly, this system relies on the user s collaboration and expertise in placing various sensors in the most appropriate body locations. These assumptions may not hold for elderly people with forgetfulness or users who are not technology savvy. Fig.6. Stress monitoring system consisting of sensors, personal server and a mobile device. (Sourced from Jovanov [25], Figure 1). Other wearable solutions were designed in a form of shirts to provide a ready-made and easy to wear solution such as Vetement de Tele-Assistance Medicale (VTAM) [48] and the commercial Sensatex smart shirt [35]. In smart shirts, various sensors (ECG, temperature, etc ) are interwoven into fabric [48]. Data is collected via a bus embedded in the shirt and sent either directly to a health centre via a GSM module placed in the belt [48], or to an integrated controller for further high-level processing prior to transmission such as in Sensatex. 2) Wearable for Chronic Heart Diseases A wealth of systems was developed for remote monitoring chronic heart diseases such as heart arrhythmias. An Arrhythmias Monitoring System (AMS) prototype was developed at NASA for real-time ECG surveillance [29]. In AMS, the patient is required to wear three-lead ECG electrodes connected to an ECG Holter (Fig. 7). The latter collects data and sends it to a local PDA server via short distance radio frequency network. Fig.7. The NASA ECG monitoring system. (Sourced from Liszka [29], Figure1). AMS, however, does not provide any feedback to the user as to their health state due to the rudimentary signal processing performed at the user-end. In the wireless heart rate system in [28], Lin et al used a pocket PC in which ECG signals are directly processed, and physiological parameters such as heart rate can be displayed. The user can also program alarms and data can be sent via WLAN to a remote health unit. In the Vital Sign Monitor (VST) system of [30], ECG measurement was displayed in a Graphical User Interface (GUI). First data is collected by a data acquisition card and then transmitted to a home PC on which a dedicated GUI-based software displays information to the user and sends it to a remote management unit. All systems in [30], [29] and [28] require careful placement of ECG leads by the patient which may not be feasible by elderly users. The AMON project [1] provides a more userfriendly solution whereby ECG electrodes and signal processing are all assembled in a wrist-worn device. Another approach was adopted in [19] whereby the heart rate is inferred from PPG signals detected by a finger ring sensor. The PPG signals are wirelessly transmitted to a watch like device which computes and displays heart rate. In some severe cases, prediction of cardiac deaths is vital. Medical devices capable of autonomously correcting the heart rhythm have been developed such as the Implantable Cardioverter Defibrillator (ICD) [40]. The critical functionality of these devices led to the necessity of monitoring them. In the recent commercial system Belos VR-T (Biotronik, Berlin, Germany [40]), the data stored in the ICD device is transmitted to a dedicated patient GSM-telephone which in turn encrypts it and sends it to a Home Monitoring Service. Also, in case of the ICD device failure, an SMS message is sent to the physician. 3) Wearable for Diabetes Diabetes requires intensive care and surveillance by both the patient and the doctor. Scientific research has shown that continuous monitoring of serum glucose is necessary to avoid long-term complications [47]. Such tight control may necessitate numerous blood samples a day; a painful process for the patient. Various glucose sensors were developed to address

7 7 this issue such as surgically implantable, subcutaneous, and transderamal sensors [44]. One of the earliest commercial diabetes monitoring devices is Minimed which is based on an implantable enzymatic electrochemical sensor attached to a portable pager-sizemonitoring unit (Fig. 8) [47]. MiniMed requires regular surgical implantation of the sensor due to limited sensor battery. Fig.8. Minimed device for independent monitoring of diabetes. (Adapted from Tamada [47], Figure 3). Glucose measurement through the skin is an alternative used by the wrist-watch Cygnus device of Fig. 9 [47]. The underside of the watch contains embedded sensors, special disks for glucose collection and electrodes for measurement. Every 20 minutes, sensors apply extraction current and glucose is collected and measured. Fig.9. The Cygnus transdermal monitor device. (Adapted from Tamada [47], Figure 3). Another non-invasive glucose measurement technique was proposed in [51] whereby an enzyme sensor is employed to measure saliva glucose from which blood glucose is inferred. Nevertheless, total self-monitoring by patients is notoriously inefficient, particularly for patients who are not fully aware of the complications of diabetes or may not be competent to operate self-monitoring devices. Numerous models were proposed for remote diabetes management. The M2DM project aims to provide remote care for patients with diabetes [2]. M2DM consists of a multiple access server connected to computer telephony server. Patients would plug their glucose meter to the phone to upload measurements onto the M2DM server for analysis by doctors. Although M2DM facilitates data sharing and access, it results in huge amount of data for analysis by doctors. In a more recent work [17], a multi-agent based architecture was proposed to interpret data in the M2DM server and raise alarms. With the advances of mobile phones, various mobile-based diabetes surveillance systems were developed. In the GLUCONET, a project developed by France Telecom and Toulouse University hospital, patients send glucose measurement to doctors in the form of phone messages [3]. Logging data is time consuming and error-prone, particularly when daily frequent measurements are required. Zou et al proposed a mobile station -Mobile Diabetes Management and Internetworking System (MDMIS) - in which data from glucose meter is transmitted to mobile phone via Bluetooth [52]. The system was implemented using Sony Ericsson P800 and MediSense Optium glucose meter. Currently, this system is under trial in St George's Hospital in London [23]. This review reveals a lack in pervasive systems that do not require full patient cooperation in monitoring diabetes. A recent and ongoing research -MobiCare- is aiming to integrate glucose sensor into WBAN [6]. While some vital signs such as ECG must be continuously monitored, blood glucose levels require periodic surveillance. The novel architecture of MobiCare allows health professionals to remotely configure, control and program sensors. For diabetes patients, this option can be used to program the frequency of glucose measurements based on the needs of the patient. 4) Wearable for Chronic Blood Pressure The commonly used cuff-based BP meter does not allow continuous measurement, and is not easily portable due to its weight. As part of the AMON project, described earlier, the wrist-watch autonomously performs blood pressure measurement without the patient s cooperation by means of an inflatable compression cuff. Although the wrist device is practical for elderly people who suffer from forgetfulness, it does not provide continuous monitoring. Hung et al developed a continuous cuffless blood pressure meter [19]. Blood pressure is inferred from the signals sent from wireless ECG sensors to a PDA. Data is processed and blood pressure is computed and displayed on the PDA. C. Evaluation of Wearable Systems Table II evaluates the wearable systems. Based on the analysis presented in Table II, three significant challenges are identified for wearable devices; and are outlined as follows. 1) Patient Cooperation All wearable systems require the end-user s cooperation. The degree of cooperation depends on the system s mechanism which can be classified into either fully-cooperative or semicooperative. Fully-cooperative systems require the end-user to have knowledge of systems he is not normally familiar with, such as placing ECG leads. Semi-cooperative systems incorporate the sensing components in objects of everyday use such as watches. Although the latter solutions are more

8 8 practical, they are less accurate due to constraints imposed on the sensors locations and the difficulty involved in embedding powerful algorithms in one platform. Fully-cooperative solutions are more accurate as signal processing is disseminated over a distributed platform (sensors, personal server, and PDA device). Another flaw is the lack of consideration of elderly with cognitive impairment. To put this into context, the PROACT system aims to ensure the wellbeing of elderly suffering from forgetfulness. Yet, the user needs to remember to wear the glove on a daily basis. TABLE II EVALUATION OF WEARABLE TECHNOLOGIES Model Advantages Disadvantages PROACT: RFID technology [37] Location Tracking: Ultrasonic technology [24] Guidance: Infrared technology [42] Stress monitoring system: Various sensors [25] Shirt: Integrated sensors and controller [36] Remote ICD: ICD device + GSM phone [41] Heart monitoring : Electrodes + PDA [40] [29] AMON: ECG by wrist device [1] MinMed: Implantable sensor [47] Cygnus: Wrist watch (transdermic technology) [47] Diabetes Management : Phones [3] MDMIS project: Glucose meter +PDA [51] [52] Reliability: 88% correctness in inferring activities Practicality: Only two lightweight sensors required, power-saving Cost: low-cost Trial: Highly rated thanks to system s fast response Trial: Very good correlation between the HRV and stress Trial: comfortable Practicality: shirt is washable and robust Trial: 97% acceptance among aged users Reliability: acceptable Practicality: ergonomic device Practicality: Lightweight and ergonomic device Practicality: Nonobtrusive device Reliability: acceptable Practicality: Manual logging of data is error-prone Trial: On-going Practicality: It avoids manual data logging Practicality: The glove is bulky, impractical for handwashing and unhygienic for eating activity Reliability: Ultrasonic technology is sensitive to obstructions in the house Practicality: System becomes of little use once the user gains familiarity with the environment Cost: Customised designs are typically costly Cost: Long-term usability requires redundancy of sensors bringing the cost up Trial: Few patients rejected the system or were not able to use GSM mobile Practicality: patient required to install ECG leads Reliability: Inaccurate due to noise and incompetence of embedded algorithms Trial: system still immature Practicality: Requires regular surgical implantation - Does not alert the user Trial: The device causes skin reactions and hence impractical for frequent use Trial: 95 % of 537 elders found the technology easy to use [8]. Practicability: Patient is required to be familiar with the PDA 2) System s Mobility At first glance, wearable systems seem to offer a flexible surveillance solution. Nonetheless, the degree of mobility depends on the system s architecture. Solutions such as the smart shirt and the portable stress monitor allow full mobility but depend on the lifetime of the sensor s battery. On the other hand, the tracking and guidance systems of [34], [24], [42] and [35] adopt an architecture distributed over the user and his/her environment. Therefore surveillance is confined to a particular environment; reducing the system s efficiency in some cases. For instance, the guidance system for individuals with vision loss becomes ineffective once the user has gained familiarity with the environment in which the system was installed. 3) Communication Factor One of the major barriers in the implementation of WBANs is the communication cost. Experiments carried in [6] show that continuous monitoring produces data in the range of 10 MB per day per user. Clearly, with the current communication cost models, deploying a WBAN is cost prohibitive. VI. COMBINED PERVASIVE HEALTH SYSTEMS (CATEGORY III) The predominant advantage of environmental over wearable sensing lies in its non-invasive and non-obtrusive nature. This is coupled to the lenient power requirement on sensors since the latter can plug to power sources available at home. Nevertheless, environmental systems are not as accurate as their wearable counterpart due to personalisation issues, the discrete nature of their measurement and the restricted access to internal physiological parameters. Until today, no solution has been worked for environmental blood glucose measurement. Wearable devices are more accurate thanks to continuous surveillance and accessibility to internal fluids and offer more mobility to patients, although this depends on the battery lifetime of sensors. Another caveat is the obtrusive nature of wearable devices which may cause stigma to its users. Besides, the responsibility of wearing medical devices falls upon the patient, hence ease of use and user acceptance become unavoidable requirements. In practice, the two systems should not be segregated since wearable devices operate within an environment by which they are inevitably influenced. For instance, temperature sensors in WBAN are influenced by the environment temperature leading to erroneous body temperature measurement. Therefore, both approaches are complementary and if combined, maximum accuracy can be achieved. The Elite Care Retirement Home, currently operating in Portland Oregon Metro area, is an example of such systems [45]. Residents wear infra-red radio frequency badges to prevent them from wandering off and sensors are attached to beds for un-obtrusive weight and sleep monitoring. VII. FUTURE TRENDS AND RESEARCH THRUSTS Based on this review, the paper identifies three major challenges that will propel future research; these are listed in order of importance as follows.

9 9 1) Hybrid Platforms A subtle gap can exist between activities reported by sensors and what the monitored person is actually doing. A sensor attached to the tap reports that the tap is opened but does not imply that the person washed his/her hands. This gap can be bridged if for each monitored activity, the most appropriate technologies are employed and fused. A more accurate system for hand-washing activity, for example, can be obtained if a RFID tag reports a tap is opened, an infra-red system records that the person is standing near the lavatory, and an acoustic system detects the associated sounds. While hybrid platforms may be the way forward to bridge the gap, they are harder to integrate. Fusion techniques that require minimal cost and complexity are to be investigated. 2) Sensor Design Maximum power-saving necessitates that the sensor node restricts transmission to abnormal situations and alarms. This requires high-level signal processing typically performed in a home PC or a personal server. This additional paraphernalia drives the cost and the weight of the system up; two limiting factors for WBAN. A potential revolutionary sensor design would integrate sensing, high-level signal processing and transmission capabilities in a single platform. An integrated control loop would decide on the most appropriate type of transmission (Bluetooth, zigbee, etc...) depending on the urgency of medical situation and status of the sensor s battery; as depicted in Fig. 10. Such a sensor would require a powerful, low-power microprocessor and would procure power using scavenging techniques; a newly-born and active area of research. VIII. CONCLUSIONS Although m-health has recently witnessed significant advances, a wealth of research is still seeking for cost-effective, well-placed and practical pervasive health solutions. Environmental health systems have been discussed and appear to be more successful and accurate for activity monitoring and ensuring wellbeing than for health surveillance. On the other hand, analysis of wearable technologies has revealed those systems to be more accurate and practical for health monitoring and particularly for chronic disease management. If m-health dogma is to take over the current health systems, it should consider social, ethical and legal aspects as well as address various technical, deployment and marketing challenges; as follows: Social and ethical aspect: It should combine reliability, affordability, privacy and ease of use in order to gain the acceptance and the trust of society. Legal aspect: M-health systems should comply with the legal system governing the current healthcare such as health insurance policies. Technical challenges: Data fusion and integration techniques in hybrid platforms need to be explored. Possible designs for lightweight, intelligent and lowpower sensors for WBAN are to be investigated. Deployment: More on-field trials need to be conducted on a large scale to raise awareness of m-health systems. Marketing challenges: New business models need to be employed, particularly in the cost policies related to mobile and communication industry. Pervasive health offers considerable opportunities for innovation and hence opens new vistas not only for research, but for the industry and the market as well. Fig.10. Proposed architecture for an integrated and intelligent sensor with a single platform including a control loop. 3) Predictive Pervasive Healthcare This review revealed that no research has aimed at combining environmental health and activity monitoring systems. Yet, a medical research has established that change in daily activities can be a very efficient way for an early detection of a medical condition [4]. Therefore, a combined health-activity home would be capable of automatically sensing change in activities, making conclusions as to the possible medical conditions relating to this change, immediately checking vital signs to verify any existing correlations and finally transmitting early diagnosis to health professionals. Such a house system would be a breakthrough in the field of remote predictive healthcare. REFERENCES [1] Anliker, U.; Ward, J. A.; Lukowicz, P.; Troster, G.; Dolveck, F.; Baer, M.; Keita, F.; Schenker, E. B.; Catarsi, F.; Coluccini, L.; Belardinelli, A.; Shklarski, D.; Alon, M.; Hirt, E.; Schmid, R.; Vuskovic, M., AMON: a wearable multiparameter medical monitoring and alert system, Information Technology in Biomedicine, IEEE Transactions on, vol. 8, no. 4, pp , Dec [2] Bellazzi, R.; Carson, E. R.; Cobelli, C.; Hernando, E.; Gomez, E. J.; Nabih-Kamel-Boulos, M.; Rendschmidt, T.; Roudsari, A.V., Merging telemedicine with knowledge management: the M2DM project, Engineering in Medicine and Biology Society, Proceedings of the 23rd Annual International Conference of the IEEE, vol. 4, pp , [3] Black, L. A.; McTear, M.; Black, N.; Harper, R., Intelligent computer consultants - interfacing the future, Computer-Based Medical Systems, CBMS 2004 Proceedings 17th IEEE Symposium on, vol. 24, no. 25, pp , Jun [4] Boutayeb, A.; Boutayeb, S.," The burden of non communicable diseases in developing countries," International Journal for Equity in Health, vol. 4, no. 1, pp.2, Jan [5] Boutayeb, A.; Twizell, E. H.; Achouyab, K.; Chetouani, A., A mathematical model for the burden of diabetes and its complications, BioMedical Engineering Online Journal, vol. 3, no. 1, pp. 20, Jun [6] Chakravorty, R., A programmable service architecture for mobile medical care, Pervasive Computing and Communications Workshops, PerCom Workshops Fourth Annual IEEE International Conference on, vol.13, no. 17, pp. 5, Mar

10 10 [7] Choi, J. M.; Choi, B. H.; Seo, J. W.; Sohn, R. H.; Ryu, M. S.; Yi, W.; Park, K. S., A System for Ubiquitous Health Monitoring in the Bedroom via a Bluetooth Network and Wireless LAN, Engineering in Medicine and Biology Society, IEMBS '04. 26th Annual International Conference of the IEEE, vol. 2, no. 1-5, pp , Sep [8] Chumbler, A. N. R.; Neugaard, A. B.; Ryan, A. P.; Qin, A. H.; Yongsung, A. J., Evaluation of a Care Coordination/Home-Telehealth Program for Veterans with Diabetes: Health Services Utilization and Health-Related Quality of Life, Journal of Evaluation and Health Profession, vol. 28, no. 4, pp , [9] Consolvo, S. D.; Roessler, P.; Shelton, B.E., Digital family portraits: supporting peace of mind for extended family members, Proceedings of the 6th Int'l Conference on Ubiquitous Computing: UbiCom04, pp. 1-17, Sep [10] Consolvo, S.; Towle, J., Evaluating an ambient display for the home, In CHI '05 Extended Abstracts on Human Factors in Computing Systems. ACM Press, pp , July [11] Cuddihy, P.; Hinman, R.T.; Avestruz, A.; Lupton, E. C.; Livshin, G.; Rodriguez, J. I.; Leeb, S. B.; Clark, C. M.; Horvath, K.J.; Volicer, L.; Landfeldt, B.; Kay, J.; Kummerfeld, R.; Quigley, A.; West, D.; Apted, T.; Sinclair, G.; Haniff, D. J.; Kalawsky, R.; Atkins, D.; Lewin, M.; Brown, S. J.; Shahmehri, N.; Aberg, J.; Maciuszek, D.; Chisalita, I., Successful Aging, Pervasive Computing, IEE, vol. 3, no. 2, pp , Apr-Jun [12] Department of Health, Building Telecare in England, 2005, Available at: [13] Global Watch Mission Report, Intelligent Housing: new and emerging technologies for home-based healthcare delivery-mission to USA, pp , Oct-Nov [14] Hammerstrom, D., Neural networks at work, Spectrum, IEEE, vol. 30, no. 6, pp , July [15] Hayes, T. L.; Pavel, M.; Kaye, J. A., An unobtrusive in-home monitoring system for detection of key motor changes preceding cognitive decline, Engineering in Medicine and Biology Society IEMBS '04. 26th Annual International Conference of the IEEE, vol. 4, no. 1-5, pp , Sep [16] Helal, S.; Mann, W.; El-Zabadani, H.; King, J.; Kaddoura, Y.; Jansen, E., The Gator Tech Smart House: a programmable pervasive space, Pervasive Computing, IEEE, vol. 38, no. 3, pp , March [17] Hernando, M. E.; García, G.;, Gómez, E. J.; del Pozo, F., Intelligent alarms integrated in a multi-agent architecture for diabetes management, Transactions of the Institute of Measurement and Control, vol. 26, no. 3, pp , [18] Hori, T.; Nishida, Y.; Aizawa, H.; Murakami, S.; Mizoguchi, H., Sensor network for supporting elderly care home, Proceedings of IEEE, vol. 2, no , pp , Oct [19] Hung, K.; Zhang, Y. T.; Tai, B., Wearable medical devices for telehome healthcare, Engineering in Medicine and Biology Society, IEMBS '04. 26th Annual International Conference of the IEEE, vol. 7, no. 1-5, pp , Sep [20] Ishijima, M., Monitoring of electrocardiograms in bed without utilizing body surface electrodes, Biomedical Engineering, IEEE Transactions on, vol. 40, no. 6, pp , Jun [21] Ishijima, M., Long-term cardiopulmonary monitoring in bed without subject awareness, Proceedings of the 18th Annual International Conference of the IEEE, vol. 1, no. 31, pp , Nov [22] Ishijima, M.; Togawa, T., Observation of electrocardiogram through tap water, Clin Phys PhysiolMeas, vol. 10, pp , [23] stepanian, R. S. H., Academic Talk: Emerging Personalised Mobile Healthcare Systems, March Available at: [24] Jovanov, E.; Milenkovic, A.; OttO, C.; De Groen, P.; Johnson, B.; Warren, S.; Taibi, G., Enabling location-aware pervasive computing applications for the elderly, Pervasive Computing and Communications, (PerCom 2003). Proceedings of the First IEEE International Conference on, vol. 23, no. 26, pp , Mar [25] Jovanov, E.; O'Donnell Lords, A.; Raskovic, D.; Cox, P.G.; Adhami, R.; Andrasik, F., Stress monitoring using a distributed wireless intelligent sensor system, Engineering in Medicine and Biology Magazine, IEEE, vol. 22, no. 3, pp , May-Jun [26] Kawarada, A.; Takagi, T.; Tsukada, A.; Sasaki, K.; Ishijima, M.; Tamura, T.; Togawa, T.; Yamakoshi, K, Evaluation of automated health monitoring system at the Welfare Techno House, Engineering in Medicine and Biology Society, Proceedings of the 20th Annual International Conference of the IEEE, vol. 4, no. 29, pp , Nov [27] Lim, Y. G.; Kim, K. K.; Park S., ECG measurement on a chair without conductive contact, Biomedical Engineering, IEEE Transactions on, vol. 53, no. 5, pp , May [28] Lin, Y.; Jan, I.; Ko, P. C. I.; Chen.Y.; Wong, J.; Jan, G, A wireless PDA-based physiological monitoring system for patient transport, Information Technology in Biomedicine, IEEE Transactions on, vol. 8, no. 4, pp , Dec [29] Liszka, K. J.; Mackin, M. A.; Lichter, M.J.; York, D. W.; Dilip Pillai; Rosenbaum, D. S., Keeping a beat on the heart, Pervasive Computing, IEEE, vol. 3, no. 4, pp , Oct-Dec [30] Mendoza, P.; Gonzalez, P.; Villanueva, B.; Haltiwanger, E.; Nazeran, H, A Web-based vital sign telemonitor and recorder for telemedicine applications, Engineering in Medicine and Biology Society, IEMBS '04. 26th Annual International Conference of the IEEE, vol. 3, no. 1-5, pp , Sep [31] Mynatt, E. D.; Melenhorst, A. S.; Fisk, A. D.; Rogers, W. A., Aware technologies for aging in place: understanding user needs and attitudes, Pervasive Computing, IEE, vol. 3, no. 2, pp , Apr- Jun [32] Mynatt, E. D.; Rowan, J.; Craighill, S.; Jacobs, A., Digital family portraits: supporting peace of mind for extended family members, In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ACM Press, pp , [33] Nakamura, H.; Karube, I., A Current researchch activity in biosensors, ICMB Analytical and Bioanalytical Chemistry Journal, vol. 337, no. 3, pp , Oct [34] Ogawa, M.; Tamura, T., Monitoring of heart and respiratory signals with PPG in bathing, BMES/EMBS Conference, vol. 2, pp. 778, Oct [35] Ooi, P.; Culjak, G.; Lawrence, E., A Wireless and wearable overview: stages of growth theory in medical technology applications, ICMB 2005 International Conference on, pp , Aug [36] Park W. S.; Jayaraman, S., Enhancing the quality of life through wearable technology, Engineering in Medicine and Biology Magazine, IEEE, vol. 22, no. 3, pp , May-Jun [37] Philipose, M.; Fishkin, K. P.; Perkowitz, M.; Patterson, D. J.; Fox, D.; Kautz, H.; Hahnel, D., Inferring activities from interactions with objects, Pervasive Computing, IEEE, vol. 3, no. 4, pp , Oct- Dec [38] Philipose, M.; Smith, J. R.; Jiang, B.; Mamishev, A.; Roy, S.; Sundara-Rajan, K., Battery-free wireless identification and sensing, Pervasive Computing, IEE, vol. 4, no. 1, pp , Jan.-March [39] Price, J, An implementation of wireless network of intelligent sensors WISE, M.S. thesis, Electrical and Computer Engineering Department, Univ of Alabama, [40] Res, J. C. J.; Theuns, D. A. M. J.; Jordaens, L., The role of remote monitoring in the reduction of inappropriate implantable cardioverter defibrillator therapy, Journal of Clinical Research in Cardiology, Mar [41] Res, J. C. J.; Theuns, D. A. M. J.; Jordaens, L., Home monitoring in ICD therapy: future perspectives, Europace Journal, vol. 5, pp [42] Ross, D. A., Cyber crumbs for successful aging with vision loss, Pervasive Computing, IEEE, vol. 3, no. 2 pp , Apr-Jun [43] Royal Society, Digital healthcare: the impact of information and communication technologies on health and healthcare, 2006, available at: [44] Shlomit, A. S; Moshe, A.P., Closing the loop: combining insulin pumps and glucose sensors in children with type 1 diabetes mellitus, Journal of Pediatric Diabetes, vol. 7, no. 4, pp [45] Stanford, V., Using pervasive computing to deliver elder care, Pervasive Computing, IEEE, vol. 1, no. 1, pp , [46] Sixsmith, A.; Johnson, N, A smart sensor to detect the falls of the elderly, Pervasive Computing, IEEE, vol. 3, no , pp , April-June