This paper is recommended in Skizit Gesture’s video: Gangstalkers Use Smart Phones To Remotely Operate Implants
Skizit Gesture: “This type of Healthcare has been covertly implanted in Targeted Individuals for body modification and torture, not healthcare.”
Crilly, P, Muthukkumarasamy, V. (2010)
Conference Title
Proceedings of the 2010 6th International Conference on Intelligent Sensors, Sensor Networks
and Information Processing, ISSNIP 2010
t
© 2010 IEEE. Personal use of this material is permitted. However, permission to reprint/
republish this material for advertising or promotional purposes or for creating new collective
works for resale or redistribution to servers or lists, or to reuse any copyrighted component of
this work in other works must be obtained from the IEEE.
Abstract—Pervasive health care is regarded as a key driver in
reducing expenditure and enabling improvements in disease
management. Advances in wireless communication and sensor
technologies permit the real time acquisition, transmission and
processing of critical medical information. In this paper, we
examine different approaches of streaming physiological data
from body sensors over a wireless network. Modern mobile
phones provide sufficient storage and computational abilities and
provide a flexible programming environment, making them ideal
to process and store sensed data from multiple sources. We
compare the approach of using a central data server, against
using a smart phone, to store and process the medical data. The
competing requirements of minimization of energy consumption
versus the timely delivery of anomalous conditions are
investigated using a simulated body sensor network. The
measurements show that when a patient is mobile, a smart phone
is the device best suited to perform the initial processing of vital
signs and sending of medical alerts.
I. INTRODUCTION
In Western countries, the aging population and the rise of
chronic diseases are placing increasing pressure on the cost of
providing health care. For example, Australian Government
expenditure on health and aged care is expected to rise from
9% to 12.4% of GDP (i.e. $246 billion) by 2020 [1].
Furthermore it is estimated that 30 to 50 per cent of patients
with chronic disease are hospitalised because of inadequate
care or management [2]. The global shortage of qualified
health care professionals is also contributing to a higher
burden on those in the industry [3]. Technology has often
provided a platform for enhancing the provision of quality
healthcare and for driving down the health expenditure.
A wireless sensor network consists of a large number of
distributed nodes. The nodes are designed to be small, cheap
and autonomous. Each node consists of a microcontroller, a
transceiver, an energy source and a sensor. The nodes are
usually connected in an ad-hoc arrangement. Wireless sensor
networks are deployed in a variety of application areas, such
as environment and habitat monitoring, healthcare and traffic
control.
Body area networks can be considered as a special type of
wireless sensor networks with their own specific requirements.
They differ in nature and requirements from traditional wide-
area wireless sensor networks, with the most important ones
being the increased demand for reliability, energy efficiency
and mobility support. The sensor nodes are designed to be
worn on the body or implantable. IEEE has identified three
application domains where body sensor networks are useful;
disease management, health and fitness, and independent
living [4]. In the later case, a home health care network,
comprising body sensors, cameras and a home health
controller, can be used to monitor a patient’s well being in
their home environment [5].
The home health controller is expected to reside
permanently within the home, so when a patient leaves their
residence, another device must act as the control node for the
body sensor network. The mobile phone has already been
proposed to act as a control node [6]. It provides a far more
sophisticated human interface than the limited interfaces
found on a sensor node. Mobile phones are equipped with
several services capable of communicating with a centralised
server e.g. SMS, MMS, and GPRS. Modern phones, such as
the iPhone, are equipped with cameras, accelerometers and
GPS devices, which can be utilised to provide contextual data
about a patient. The phone itself can be used to enter other
information or to record a patient’s emotional or mental state.
As demonstrated by Liu et al. [7], this feature can be used to
enhance a patient’s compliance to medical program or
adherence to a prescribed drug regimen. Typically Bluetooth
has been used as the wireless technology to interconnect the
mobile phone and the sensors, as it generally found on
commercially available phones. In general, the mobile phone
penetration is very high both in developed and developing
countries. For example, there are more than 22 million phones
owned by Australians and about 98% of the population has
access to a mobile phone network.
In this research, we examine the feasibility of using the
mobile phone as a control node to provide mobility and to
perform analysis and diagnosis of medical conditions, when
acting as the control node in a personal healthcare system.
Section II outlines some of the relevant research projects into
body sensor networks. The ability of various devices in a body
sensor to perform processing of physiological parameters is
examined. Section III proposes architecture for a personal
healthcare system and identifies a number of requirements the
system should fulfil. Section IV details a testbed for
measuring some metrics for the proposed system. Section V
presents and discussed the results of these measurements.
Finally conclusions are reported in Section VI.
978-1-4244-7177-5/10/$26.00 © 2010 IEEE 291
ISSNIP 2010
TABLE I: SAMPLE IMPLEMENTATIONS OF USING A MOBILE DEVICE FOR MONITORING
Platform Base Device Wireless Protocol Function
HealthGear Phone Bluetooth Wearable system for connecting sensors and mobile phones
CodeBlue PC, PDA 802.15.4 Provision of medical monitoring in a hospital environment
ALARM-NET PC, PDA,
Stargate
Bluetooth 802.11 Wireless sensor network for
assisted-living and residential monitoring.
DexterNet PC, PDA,
Phone
802.15.4 Body sensor network for indoor and outdoor monitoring.
II. RELATED WORK
A range of software and hardware architectures have been
proposed and implemented to evaluate and address the needs
of body sensor networks in health care applications. Table I
summarises some of the major research projects into body
sensors networks and monitoring [8-11].
The HealthGear project implemented a wearable real-time
monitoring solution for the detection of apnea events. A pulse
oximeter was connected to a mobile phone using the
Bluetooth protocol. The project demonstrated the feasibility of
storing and processing of physiological data on a mobile
phone. The Harvard University research project, CodeBlue,
developed a number of wireless sensors for monitoring vital
signs and examined the benefits to clinicians, nurses,
emergency personal in hospital environments and disaster
situations, of connecting the sensors in an ad-hoc arrangement
and relaying the data. The Alarm-Net project developed a
system architecture for assisted living and residential
monitoring. The Alarm-Net architecture has been used by
several other projects to investigate issues surrounding
wireless body sensors and healthcare. DexterNet is open
source platform that uses a three layer architecture to support
monitoring in indoor and outdoor environments. The layers
are the Body Sensing Layer (BSL) which is used to
interconnect sensing for monitoring the biological signals; the
Personal Network Layer (PSL), which co-ordinates the BSL;
and the Global Network Layer (GNL), which supports high
level applications such as remote health monitoring.
To evaluate where storage and analysis of data should take
place in a body sensor network, we categorise them into three
distinct approaches to data processing and detection of
abnormal conditions:
1. Continuously streaming data from the sensors to a
control node. The control node then acts as an aggregator of
the data, forwarding it on to a remote server where diagnosis
is performed.
2. The sensors are programmed to detect and raise alerts
when abnormal conditions are detected. These alerts are sent
to the control node which forwards them on to an emergency
centre or medical practitioner, so appropriate action can be
undertaken.
3. Data is continuously streamed to a control node,
where the analysis is performed. The control node processes
the data and raises alerts when abnormal conditions are
detected. These alerts are then sent to emergency centre etc.
We examine the research utilising body sensor networks to
aid in the detection of cardiac arrhythmias to explore the
relative merits of each approach. Cardiac arrhythmia is an
irregular heart beat or abnormal heart rhythm. Most
arrhythmias are harmless, but some can indicate a serious or
life threatening condition. Some arrhythmias occur
infrequently, making them hard to diagnose. An ECG, which
records the electrical activity of the heart is the most method
used to diagnose arrhythmias.
Yousef & Lars [12] demonstrated that an ECG sensor and
mobile phone could be integrated into a usable system for
remote cardiac monitoring. A remote server was used to
analyse the data. Their system had an uptime of 95.5% and
90% of ECGs were delivered reliably. As transmitting data
over the cellular network is a large drain on a phone’s battery,
the reliability was achieved by powering the phone from a
separate battery package and having the heart patient carry an
extra mobile phone. Continuous streaming of data is therefore
unsustainable in terms of a phone’s battery life.
Wireless data transmission is the major source of energy
consumption from the sensor. The approach adopted by
Rincon, et al. [13] was to implement ECG analysis in a sensor.
They showed that energy consumption of the sensor could be
reduced by 92.6% by detecting and sending pertinent points of
the ECG signal. They also showed that if complete diagnosis
is performed on the sensor, the reduction in transmission is
100% if no abnormalities are detected. The limitations of this
approach are that there is no data saved for verification or
auditing of the analysis and that the algorithm used for
detection is not easily updated.
Oresko, et al. [14] implemented a prototype system, where
ECG analysis and classification is performed by streaming
data from an ECG heart monitor to a smart phone. This
approach represents a compromise between conserving power
in the system and permitting data analysis and storage.
Figure 1: Proposed personal flexible health system for capturing vital signs
III. PROPOSED PERSONAL HEALTHCARE SYSTEM
In this section, we examine and identify the requirements in
terms of communication, storage, processing that need to be
satisfied for an acceptable healthcare system. The most
appropriate architecture is investigated and proposed.
The control node in the personal health system should be
able to perform the following functions
• Be able to analyse and process vital signs for
abnormal conditions using limits and techniques that are
adapted to a patient’s personal history.
• Be able to store data for remote queries, auditing and
comprehensive analysis.
• Be flexible. The system should have the ability to
have algorithms downloaded as required and permit the
detection metrics to be updated remotely.
• Be able to send alarms and associated vital signs in
timely manner, when medical alerts are detected. These vital
signs can undergo further analysis on a remote expert system
or by a clinician.
Figure 2 presents a proposed architecture for personal
health system that can be used for monitoring patient’s vital
signs. The system is used to add mobility to a home health
care server, so that a patient need not be confined to their
place of residence but instead be able to participate in their
usual day-to-day activities.
The personal health system depicted consists of the patient
with a number of implanted or attached body sensors, smart
phone with Internet connectivity to a remote server
1. A set of wireless sensors for measuring the vital
signs of blood pressure, heart rate, temperature and respiration
rate are either implanted or attached to the body. The
acceptable ranges of the physiological values vary depending
on the activity level of the patient, so an accelerometer is
included in this instance to provide some basic context
information about the activity level of the patient e.g. have
they had a fall, are they running or at rest.
2. A smart phone which acts as the control node in the
network. There is no clear industry definition of what
constitutes a smart phone. We use the term smart phone to
distinguish between a phone which offers basic features such
a making calls and a phone which offers a richer set of built-in
applications and provides Internet connectivity, the later being
defined as a smart phone. Smart phones are distinguished by a
more powerful processor, larger screen and more memory
than that of a basic mobile phone. With Bluetooth being
widely available on mobile phones, it is adopted as the low-
power communication protocol for connecting the sensors to
the mobile phone. The phone is responsible for discovering
and configuring the connections to the sensors.
The sensors are arranged in a star network, since they are
heterogeneous in nature and do not exchange information
between each other.
3. A server, which may be a doctor’s office, hospital or
a remote monitoring service. Abnormal events and alarms are
sent to the server when they are detected. Clinical data
pertaining to the event can either be retrieved from the phone
or sent to the server over the Internet using a 3G mobile
telephony protocol.
As a patient leaves their residence, the control node
performs network assembly. The sensors remain connected to
the mobile phone and transmitting physiological data for the
duration of the patient’s absence from their home. To preserve
energy whilst not communicating, one of the Bluetooth low
power states can be utilised. While the patient is mobile, the
phone is responsible for analysing the vital signs, storing them
to the phone, and detecting and sending alerts when abnormal
conditions are detected.
Communicating with the public interface from the phone is
expensive in terms of the energy and the network connection
is unreliable. Furthermore, receiving the raw data in real-time
is not the primary function of the system. The focus is
detecting when parameters exceed predefined bounds or
anomalies are detected. When an anomaly is detected, a
doctor or clinician can query the phone and request the phone
to send the data pertaining to this event. The delay between
requesting and receiving this window of information is
important, since time to respond to a medical emergency and
in many other situations can be critical in determining the
outcome.
Analysing and storing sensor data and on the phone
represents a trade off between conserving energy and timely
detection and reporting of abnormal conditions. Section IV
discusses a test setup and measurements used to analyse this
trade off.
IV. EXPERIMENTS
To assess the relative merits of continuously streaming data
to a remote sever against analysing and storing data on the
mobile phone, we examined two test scenarios:
1. An ECG sensor is used to continuously stream
samples to a remote server via a mobile phone
connected to the Internet.
2. Samples are sent from an ECG sensor to a mobile
phone where they are stored. After a number a
samples are sent an alert is simulated by sending
the stored samples to a remote server.
We are not concerned with the actual analysis of the sensed
data. It is assumed that same algorithm for processing of the
data or appropriately modified version can be implemented
equivalently on either the remote server or on the mobile
phone, and the difference in the compute times between the
two platforms is negligible relative to packet transmission
times and other delays.
A. Testbed
The testbed show in the figure below is used to perform the
measurements. The testbed consists of a simulated wireless
ECG sensor connected to mobile phone via Bluetooth. A PC
which is connected to the Internet acts as the remote server
Figure 2: Test Setup: simulated ECG sensor, mobile phone and remote server
This setup is used to measure the time taken to transmit the
data and the power consumption in both the scenarios. For the
first case – continuously streaming samples, time taken to
send the sensed data from the sensors to the mobile phone and
then from the phone to the server are recorded. For the second
case, the time taken to connect and transmit stored data from
the phone to the server is measured. Power consumption by
the mobile phone is also measured.
1) Wireless ECG Sensor
To simulate a wireless ECG sensor, a C application is
developed and used. The MIT-BIH Arrhythmia Database is
utilised to provide ECG data. The database contains 48 half-
hour excerpts of two-channel ambulatory ECG recordings,
obtained from 47 subjects [16].
The application reads samples using a single record from
the database. To permit timing calculations to be performed,
the application sends a packet containing the current time and
the sampled value every 2.7 ms, as the ECG recordings were
digitized at 360 Hz.
Open ECG record.
Read n samples.
Connect to phone via Bluetooth.
For n samples
Sleep sampling freq.
Send packet.
Figure 3: Pseudo code for simulated ECG sensor
2) Mobile Phone
The timing measurements were performed using a Nokia
E65 mobile phone. To perform power consumption
measurements, a Nokia N95 mobile phone running the Nokia
Energy Profiler MIDlet was used. Internet connectivity was
established using a commercial 3G network offering 7.2 Mbps
HSDPA and 2 Mbps HUSPA services.
For each of the test scenarios a MIDlet written in J2ME is
used on the phone. The pseudo code for the MIDlets used to
implement the test scenarios of continuous streaming of
sensor data and storing data to the phone is shown in figure
and figure below.
Listen for Bluetooth connection.
Connect to remote server.
While Bluetooth connected
Read packet from sensor.
Read system time.
Add to packet.
Send packet to server.
Figure 4: Pseudo code for MIDlet which handles continuos streaming of
sensor data.
Listen for Bluetooth connection.
Receive n packets from sensor.
Connect to remote server.
Send packet with samples to server.
Figure 5: Pseudo code for MIDlet which stores sensor data and sends an alert
3) Server
A C application on the server listens for incoming
connections from the Internet. For the data streaming test
case, it reads a fixed number of packets and performs timing
calculations on them. In the second scenario, the application
waits for the connection from the mobile phone simulating an
alert. The application then receives the packet containing the
alert data and calculates the time it took to send.
4) Synchronising Time
The client program which simulates the wireless ECG
sensor and the server application are run the same PC to
ensure that the timing calculations are accurate. The PC also
runs an NTP server daemon. Before each test, the
FreeTimeBox [17] application is run on the mobile phone to
synchronise time with the PC.
At the start of each set of timing measurements, the time
delta between the phone and the server is determined by
sending 20 packets from the server to the mobile phone, and
have the packets echoed back with the time on the phone
included in them.
V. RESULTS AND DISCUSSION
In this section we summarise and discuss the results
obtained from our experiments using the testbed in Figure 2
and the two test scenarios outlined in section IV.
A. Comparison of Energy Consumption.
Figure 6(a) and 6(b) show the power consumption graphs
obtained when continuously streaming data and when storing
the same samples on the phone respectively.
(a) (b)
Figure 6: Power consumption graphs
Figure 6(a) shows that power consumption is about 0.3
Watts while waiting for a Bluetooth connection. This rises to
2 Watts as the sensor connects and starts streaming samples.
About 1.4 Watts is consumed for continuous streaming of
sensor data. The value in top right corner (4:41 hours) shows
the life of a fully charged battery at the average power
consumption, which in this case is 0.81 Watt.
Figure 6(b) shows power consumption of about 0.9 Watts
for receiving streaming samples over Bluetooth and rising to
about 1.3 Watts when sending the stored data over the 3G
network to remote server. The average power consumption
shows that battery would last 7.26 hours. Considering the
maximum amount of time permit a home-monitored may need
to be away from their residence, this value seems reasonably
adequate.
The results of the power consumption measurements are
summarised in Table II. By storing samples on the phone
around 39% energy saving is achieved, thus obtaining an
almost 65% increase in battery lifetime.
TABLE II: SUMMARY OF POWER CONSUMPTION MEASUREMENTS
Action Average Power (W)
Listening for Bluetooth connections 0.30
Sending samples over Bluetooth 0.9
Continuously streaming samples 1.4
Sending alert data to remote server 1.3
B. Continuous Streaming of Sensor Data
Table III below shows the results obtained when
continuously streaming data from the wireless ECG sensor to
the remote server via the mobile phone. Time taken to send
the data to the server vary widely as indicated by the standard
deviation. Although it is not very clear why this is the case,
we assume that it could be due to the fact that some other
activity is simultaneously taking place on the phone which
affects transmission over the 3G network.
TABLE III: TIMES TO SEND ECG SAMPLES WHEN STREAMING
min (ms) max (ms) avg (ms) stdv (ms)
Bluetooth 50 122 62.59 9.31
3G 92 1485 343.76 314.87
Total 145 1543 406.35 316.29
C. Storing Samples on the Phone
Table IV below shows the time it takes to connect and
forward 300 and 600 data samples to the server respectively.
As the sampling rate of ECG data was 360 samples per second,
the values of 300 and 600 were chosen to approximate one
and two heart beats. In the absence of a heart rate calculated
from the sensor data, this represents a patient with a heart rate
of 72 beats per minute, inside the normal pulse rate range of
60 to 80 beats per minute for a healthy adult.
TABLE IV: RESPONSE TIMES TO AN ANOMALY
# of samples stored 300 600
Response time 10686 ms 19727 ms
Although the response time is 10 – 20 seconds, depending
on the sample data size, given the savings in energy
consumption, the increased response times to abnormal
conditions, justify storing sensed data on the phone rather than
continuously sending the samples to a remote server.
D. Flexibility
Programming the phone did not require any special tools
and updates to the MIDlet were easily applied by connecting
the phone via USB to a PC. This demonstrates the flexibility
of adopting the mobile phone as the platform for acting as the
control node in a body sensor network.
The sensors are by their nature designed as an embedded
systems and therefore don’t provide a framework for updating
by the user. Smart phones easily support installation and
update of applications. Vendors such as Apple, Nokia and
Google already provide infrastructure via the AppStore, Ovi
and AppsMarketPlace respectively to download and install
applications for their products. Therefore new or additional
programs or algorithms for detecting medical conditions could
be relatively easily installed onto the phone in response to
specific events being detected or as the medical needs dictate.
VI. CONCLUSIONS
To provide mobility in a personal health system for
monitoring vital signs, the mobile phone is a convenient and
suitable device to act as the control node. The phone has
become the device ubiquitous for personal communication
and data access. It is rich in functionality and can be utilised
to provide additional sensed data, either from on board sensors
or from user input. Furthermore, the new technological trends
towards developing networks to satisfy human-to-human
(H2H) and machine-to-machine (M2M) communication will
only enhance the suitability of the use of phone in the
pervasive healthcare system.
In this paper, we have shown that by storing samples on the
mobile phone instead of continuously streaming data to a
remote server results in a significant saving in battery life with
little impact on the timely delivery of abnormal events and
their associated data.
No consideration has been given to the utility nature of the
device. Besides making calls, mobile phones are also used for
taking photos, playing MP3s, playing games, etc. Further
research should examine the impact these activities may have
on the performance of the device when acting as a control
node in a body sensor network.
REFERENCES
[1] Minister for Health and Aging, The Honorable Nicola Roxon MP,
“Speech Opening Address,” in Health e-Nation Conference, National
Convention Centre, Canberra, 2009.
[2] B. K. Armstrong, et al., “Challenges in health and health care for
Australia,” Med J Aust, vol. 187, pp. 485-9, Nov 5 2007.
[3] T. Evans, editor, The World Health Report 2006. Working together for
health: World Health Organization, 2006.
[4] R. Carroll, et al., “Continua: An interoperable personal healthcare
ecosystem,” Pervasive Computing, IEEE, vol. 6, pp. 90-94, 2007.
[5] K. Singh and V. Muthukkumarasamy, “Authenticated key
establishment protocols for a home health care system “, Proceedings
of the Third International Conference on Intelligent Sensors, Sensor
Networks and Information, Melbourne, Australia, 2007, pp. 353-358.
[6] F. Fitzek and S. Rein, “Convergence of Mobile Devices and Wireless
Sensor Networks,” ed, 2007, pp. 375-380.
[7] W. T. Liu, et al., “Efficacy of a cell phone-based exercise programme
for COPD,” Eur Respir J, vol. 32, pp. 651-9, Sep 2008.
[8] P. Kuryloski, et al., “DexterNet: An open platform for heterogeneous
body sensor networks and its applications”, Sixth International
Workshop on Wearable and Implantable Body Sensor Networks, 2009,
pp. 92-97.
[9] D. Malan, et al., “CodeBlue: An ad hoc sensor network infrastructure
for emergency medical care,” Workshop on Wearable and Implantable
Body Sensor Networks, 2004.
[10] N. Oliver and F. Flores-Mangas, “HealthGear: A real-time wearable
system for monitoring and analyzing physiological signals,” in
Wearable and Implantable Body Sensor Networks, 2006. BSN 2006.
International Workshop on, 2006, pp. 4 pp.-64.
[11] A. Wood, et al., “ALARM-NET: Wireless sensor networks for
assisted-living and residential monitoring,” Department of Computer
Science, University of Virginia, 2006.
[12] J. Yousef and A. N. Lars, “Validation of a real-time wireless
telemedicine system, using bluetooth protocol and a mobile phone, for
remote monitoring patient in medical practice,” Eur J Med Res, vol. 10,
pp. 254-62, Jun 22 2005.
[13] F. J. Rincon, et al., “Implementation of an automated ECG-based
diagnosis for a wireless body sensor platform,” presented at the
International Conference on Biomedical Electronics and Devices
(BIODEVICES 2009), Porto, Portugal, 2009.
[14] J. J. Oresko, et al., “A wearable smartphone-based platform for real-
time cardiovascular disease detection via electrocardiogram
processing,” IEEE Transactions on Information Technology in
Biomedicine, vol. 14, pp. 734-740, 2010.
[15] W. R. Mower, et al., “Pulse oximetry as a fifth pediatric vital sign,”
Pediatrics, vol. v99, p. p681(6), 1997.
[16] A. L. Goldberger, et al., “PhysioBank, PhysioToolkit, and PhysioNet :
Components of a new research resource for complex physiologic
signals,” Circulation, vol. 101, pp. 215-220,
June 13, 2000.
[17] FreeTimeBox. (2010, 30 August). [Online]. Available:
http://handheld.softpedia.com/get/Business/Clock/FreeTimeBox-
64856.shtml