Energy Efficiency and Reliability in Wireless Biomedical Implant Systems

0.4 MB·English

by  Jamshid Abouei

IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 1
Energy Efficiency and Reliability in
Wireless Biomedical Implant Systems
Jamshid Abouei, Member, IEEE, J. David Brown, Konstantinos N. Plataniotis, Senior Member, IEEE and
Subbarayan Pasupathy, LifeFellow, IEEE
TABLEI
Abstract—The use of wireless implant technology requires
COMPARINGSOMEMICSSPECIFICATIONSWITHTHEIEEE802.15.4
correct delivery of the vital physiological signs of the patient STANDARDUSEDFORWSNS
alongwiththeenergymanagementinpower-constraineddevices.
1
Towardthesegoals,wepresentanaugmentationprotocolforthe Parameter MICS IEEE 802.15.4
1
physical layer of the Medical Implant Communications Service Frequency Band 402−405MHz 868/915 MHz(Eur./US)
0
(MICS) with focus on the energy efficiency of deployed devices 2.4GHz(Worldwide)
2
over the MICS frequency band. The present protocol uses the Bandwidth 300KHz 62.5KHz(for2.4GHz)
n ratelesscodewiththeFrequencyShiftKeying(FSK)modulation DataRate supportmorethan 250kbps(for2.4GHz)
250kbps
a scheme to overcome the reliability and power cost concerns in
TransmitPower 25µw(−16dBm) 0.5mw(−3dBm)
J tiny implantable sensors due to the considerable attenuation
Operating Range Typically0−2m Typically0−10m
5 of propagated signals across the human body. In addition,
the protocol allows a fast start-up time for the transceiver

] circuitry.Themainadvantageofusingratelesscodesistoprovide
T an inherent adaptive duty-cycling for power management, due [4]. A wearable WBAN provides RF communications be-
I to the flexibility of the rateless code rate. Analytical results tween on-bodysensors and a central controllerfor the remote
s. demonstrate that an 80% energy saving is achievable with the monitoring of vital signs of a patient such as ECG, blood
c proposedprotocolwhencomparedtotheIEEE802.15.4physical
[ layer standard with the same structure used for wireless sensor pressureandtemperature.Suchawearablestructureisashort-
networks. Numerical results show that the optimized rateless range communication system based on the Wireless Medical
1
codedFSKismoreenergyefficientthanthatoftheuncodedFSK Telemetry Service (WMTS) officially adopted by the Federal
v
scheme for deep tissue (e.g., digestive endoscopy) applications, Communications Commission (FCC) [5]. On the other hand,
6
wheretheoptimizationisperformedovermodulationandcoding
0 in an implantable WBAN, a biosensor embedded within the
parameters.
9 body communicates with an apparatus sticking on the body
0 Index Terms—Energy efficiency, green modulation, medical or with an external monitoring device in the vicinity of the
. implants, wireless body area networks (WBANs)
1 human body. Examples of implantable devices are cardiac
0 pacemakers, insulin dispensers, neurostimulators and bladder
1 I. INTRODUCTION controllers [6]. Such a wireless implantable technology is a
1
: A. Background very short-range communication system used for monitoring
v health conditionsof patients such as brainwaves in paralyzed
Recent advances in wireless sensor technologies have
i
X persons, glucose levels in diabetic patients and patients with
opened up a new generation of ubiquitous body-centric sys-
gastrointestinal disorders. The medical implantable devices
r tems, namely Wireless Body Area Networks (WBANs), for
a operate according to the Medical Implant Communications
providing efficient health care services [1]. WBANs sup-
Service (MICS) rules established by FCC [7].
port a broad range of medical/non-medical applications in
The MICS standard is distinguished fromthe existing stan-
home/health care, medicine, sports, etc. Of interest is the use
dardsusedforWirelessSensorNetworks(WSNs)[8],[9]due
of WBANs for the continuous remote monitoring of the vital
to the size, power consumption and frequency band require-
physiological signs of patients, regardless of the constraints
ments for implantable devices (see Table I). For instance, the
on their locations and activities [2].
MICS protocol uses an 25 µW (-16 dBm) Effective Isotropic
Broadly, WBANs can be classified into two categories of
Radiated Power (EIRP) at the 402-405 MHz frequency band.
wearable (on-body) and implantable (in-body) systems [3],
This provides a low power transmitter with a small size
Manuscript received November 28, 2010; accepted January 4, 2011. The antenna resulting in a compact and lightweight implantable
workwassupportedinpartbyanOntarioResearchFund(ORF)projectenti- device, while, today’s low power radio devices which use
tled“Self-PoweredSensorNetworks”.TheworkofJ.Aboueiwasperformed
standards such as ZigBee and Bluetooth (IEEE 802.15.1)
when he was with the Department of Electrical and Computer Engineering,
University ofToronto,Toronto,ONM5S3G4,Canada. cannot meet these stringent requirements. Furthermore, the
J.AboueiiswiththeDepartmentofElectrical andComputerEngineering, limit -16 dBm EIRP in the MICS standard reduces the effect
YazdUniversity, Yazd,Iran(e-mail:[email protected]).
ofthe interferenceon theotherbiosensordevicesoperatingin
J. D. Brown is with the DRDC Ottawa, Ottawa, ON, Canada.
He completed his Ph.D. in ECE at University of Toronto (e-mail: the same frequencyband [7]. In addition, deployingultra low
david jw [email protected]). power devices minimizes the heat absorption and the temper-
K. N. Plataniotis, and S. Pasupathy are with the Edward S. Rogers Sr.
ature increase caused by implantable biosensors, reducingthe
Department of Electrical and Computer Engineering, University of Toronto,
Toronto,ONM5S3G4,Canada(e-mails:{kostas,pas}@comm.utoronto.ca). thermal effects on the body tissues [10].
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 2
Besides the simplicity and low power consumption charac- the above dynamic channel conditions is desirable. Existing
teristics,implantabledevicesmustberobustenoughtoprovide protocols use various methods to overcome the packet loss
the desired service for long-term applications. Furthermore, and decoding error concerns in classical wireless sensor net-
since biosensors frequently switch between the sleep mode works (e.g., see [17]). The Automatic Repeat Request (ARQ)
andthe activemode,theyshouldhavefast start-uptimes. The approach, for instance, requires many retransmissions in the
protocolusedforsuchimplantabledevicesmustsupportahigh case of poorchannelconditions,resulting in a high latencyin
degreeofreliabilityandenergyefficiencyinarealisticchannel thenetwork.Thisisincontrasttothetransmissionofpatient’s
model for the human body. Each of these characteristics real time vital signs which are sensitive to the latency. More
presents considerable design challenges for different layers recently,the attention of researchershas been directed toward
of implantable WBANs, including the MAC layer and the deploying rateless codes (e.g., Luby Transform (LT) code
physical layer [11]–[13]. Central to the study of the physical [18]) in conventionalwireless networks due to the significant
layer design of implantable WBANs is the modulation and advantages of these codes in erasure channels. However, to
coding. Deploying proper modulation schemes which mini- thebestofourknowledge,thereis noexistinganalysisonthe
mize the total energy consumption in both circuit and RF energy efficiency and reliability of rateless coded modulation
signaltransmissionprolongsthebiosensorslifetime.Enhanced in implant WBANs.
reliability,onthe otherhand,is achievedby addingredundant This work deals with the first in-depth analysis on the en-
information bits in the data packets in the Forward Error ergyefficiencyandreliabilityofawirelessbiomedicalimplant
Correction (FEC) coding stage at the cost of an extra power system which uses LT codeswith the FrequencyShiftKeying
consumption. Thus, there is generally a trade-off between a (FSK) modulation scheme. The present analysis is based on
higherreliabilityandalowerpowerconsumptioninusingFEC a realistic channel model for the human body as well as the
codes. requirementsstated in the MICS standard. The main outcome
of this work is to introduce an augmentation protocol for the
physicallayeroftheMICSstandardforimplantWBANs.The
B. Related Work and Contribution
protocol allows a fast start-up time for implantable devices
There are several works in the literature that study the along with the lower power consumption during active mode
physical layer design of implantable WBANs. Oh et al. [11] duration. It is demonstrated that an 80% energy saving is
proposea new type of phase shift keyingmodulationscheme, achievable compared to the IEEE 802.15.4 physical layer
namelyPhase Silence ShiftKeying(PSSK), forhighdata rate protocol with the same structure used for wireless sensor
implantablemedicaldeviceswhichismorebandwidthefficient networks. Furthermore, we show that an inherent adaptive
than orthogonalmodulation schemes such as OOK. Under an duty-cyclingfor power managementis providedwhich comes
Additive White Gaussian Noise (AWGN) channelmodelwith from the flexibility of the LT code rate. Numerical results for
pathlossforthehumanbody,reference[11]showsthattheBit deep tissue applications show a lower energy consumption in
Error Rate (BER) of an 8-PSSK is lower than that of various the optimized coded scheme compared to the uncoded case.
sinusoidal carrier-based modulation schemes. Reference [12] The rest of the paper is organized as follows. In Sections
proposes a simple and ultra low power hardware design II and III, the implant wireless system based on a realistic
for an FSK/MSK direct modulation transmitter in medical channel model for the human body is described. The energy
implant communications which supports the data rate and consumption of an uncoded FSK modulation scheme is an-
the EIRP requirements in the MICS standard. However, no alyzed in Section IV. Design of LT codes and the energy
channel coding scheme and energy efficiency analysis were efficiency of the LT coded FSK are presented in Section
considered in [11] and [12]. Reference [14] investigates the V. Section VI provides some numerical evaluations using
feasibility of applying an Ultra-Wideband (UWB) scheme realisticparameterstoconfirmouranalysis.Also,somedesign
to implantable applications. However, according to the FCC guidelinesforusingLTcodesinpracticalimplantWBANsare
regulations in [15], UWB technologies with a typically 3-10 presented. Finally in Section VII, an overview of the results
GHzfrequencybandareusedfortransmittinginformationover and conclusions are presented.
a wide bandwidthat least500MHz,which arewiderthan the
frequencybandandthebandwidthusedintheMICSstandard.
II. SYSTEMMODEL AND ASSUMPTIONS
Since, the human body is a lossy medium, and due to
the adverse circumstances in the patient’s body (e.g., variable In this work, we consider a wireless biomedical implant
body temperature due to the fever, changing the location of system depicted in Fig. 1 with a bidirectionalcommunication
the implant capsule in the digestive system), the RF signal between an implantable biosensor, denoted by IBS, and an
transmittedfromthein-bodybiosensorexperiencesadynamic external Central Control Unit (CCU), where the transmission
channel environment to reach to its corresponding on-body ineachdirectiontakesplaceinahalf-duplexmode.Wedefine
receiver. In addition, it is shown in [16] that the efficiency the link used for the transmission of signals from the IBS to
of wireless implantable devices is related to the human body theCCUasanuplink,whilethelinkfromtheCCUtotheIBS
structure (e.g., male or female, body size, obese or thin and is represented as a downlink. During the uplink transmission
tissue composition). Clearly, finding a proper FEC coding period, the IBS wakes up after receiving a command signal
scheme which overcomes the patient’s vital information loss, from the CCU, and transmits the processed raw physiological
and operates in a high degree of flexibility and accuracy over signals to the CCU. For this period, the CCU is set to the
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 3
(cid:4) (cid:4) (cid:4) (cid:4)
(cid:2)(cid:3)fi (cid:5)(cid:6) (cid:5)(cid:6) (cid:5)(cid:6)fi (cid:2)(cid:3) (cid:2)(cid:3)
(cid:4)
(cid:7)(cid:11)(cid:7)(cid:12)(cid:13)(cid:14)(cid:4)(cid:15)(cid:16)(cid:7)(cid:17)(cid:18)(cid:19)(cid:15)(cid:20)(cid:21)(cid:4)(cid:20)(cid:20)(cid:4)(cid:9)(cid:15)(cid:7)(cid:17)(cid:4)(cid:21)(cid:6)
(cid:4) (cid:11)(cid:7)(cid:22)(cid:9)(cid:23)(cid:15)(cid:14)(cid:4)(cid:15)(cid:16)(cid:7)(cid:17)(cid:18)(cid:19)(cid:15)(cid:20)(cid:21)(cid:4)(cid:20)(cid:20)(cid:4)(cid:9)(cid:15)(cid:7)(cid:17)(cid:4)(cid:21)(cid:6)
(cid:8)
(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)(cid:8)(cid:9)(cid:10)(cid:6) (cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)(cid:8)(cid:9)(cid:10)(cid:6)
(cid:1)
(cid:4) (cid:4)
(cid:8) (cid:7)
Fig.2. Apractical multi-modeoperation inareactive implantWBAN.
switching process which achieves a significant energy saving
in the uplink mode.
For the proposed communication system, the IBS and the
CCU mustsynchronizewith oneanotherandoperateina real
time based process known as duty-cycling as depicted in Fig
2. DuringtheactivemodeperiodT fortheuplinkdatapath,
ac
the weak raw signal sensed by the implantable biosensor is
first passed through the amplification and filtering processes
toincreasethesignalstrengthandtoremoveunwantedsignals
(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)(cid:5)(cid:8)(cid:4)(cid:9)(cid:10)
(cid:11)(cid:12)(cid:13)(cid:14)(cid:9)(cid:6)(cid:14)(cid:13)(cid:15)(cid:10) and noise. The filtered analog signal is then digitized by
(cid:16)(cid:1)(cid:11)(cid:17)(cid:18)(cid:10) an Analog-to-Digital Converter (ADC), and an L-bit binary
message sequence M = {m }L is generated, where L
L i i=1
(cid:19)(cid:9)(cid:6)(cid:7)(cid:15)(cid:5)(cid:4)(cid:10) is assumed to be fixed. The bit stream is then sent to the
(cid:19)(cid:13)(cid:6)(cid:7)(cid:15)(cid:13)(cid:4) FEC coding unit. The encoding process begins by dividing
(cid:20)(cid:6)(cid:12)(cid:7)(cid:10)(cid:16)(cid:19)(cid:19)(cid:20)(cid:18)
the message M into blocks of equal length denoted by
L
B =(m ,...,m ), j =1,...,L, where k is the length
j (j−1)k+1 jk k
of any particular B . Each block B is encoded by a rateless
j j
code (in a mannerto be describedin Section V) to generatea
Fig.1. Awireless biomedical implant system.
coded bit stream C = (a ,...,a ), j =1,...,L, with
j (j−1)n+1 jn k
blocklengthn,wherenisarandomvariableandisdetermined
based on the channel condition.
listen mode. For the downlink transmission interval, the IBS
is switched to the listen mode to receive control commands ThecodedstreamisthenmodulatedbyanFSK modulation
from the CCU. It is worth mentioning that uplink has much schemeandtransmittedtotheCCUacrosstissuesinthehuman
more data to transmit than downlink, meaning the downlink body. We will explain later the advantages of using FSK
transmissiontimeismuchsmallerthanthatoftheuplink.The modulation scheme in the proposed implant system. Finally,
mainadvantageofthishalf-duplexoperationisthatacommon the IBS returns to the sleep mode, and all the circuits of
carrier is shared between the uplink and the downlink (by the transceiver are powered off for the sleep mode duration
switchingthebandwidthresourceintime),resultinginfurther Tsl for energy saving. It should be noted that only the wake-
reductionoftheIBScomplexity.Inaddition,incontrasttothe up circuit in the IBS remains on during Tsl, and waits for a
full-duplex systems, where they use two different frequency wake-up command from the CCU. Since the wake-up circuit
bands for transmitting and receiving signals, using the half- isbattery-free(similartothatusedinpassiveRadioFrequency
duplex solution maximizes the data rate for each link over Identification(RFID)tagtechnologies),itspowerconsumption
the whole bandwidth. This arises from the Shannon capacity is very low [19]. We denote Ttr as the transient mode
formulaC =Blog (1+SNR), wherethechannelcapacityC duration consisting of the switching time from sleep mode
2
is proportional to the bandwidth B. The assumption of half- to active mode (i.e., Tsl ac) plus the switching time from
→
duplex operation is used in many studies on this topic (e.g., activemodetosleepmode(i.e.,Tac sl),whereTac slisshort
see [16] for implant WBANs). Longer transmission time for enough compared to Tsl ac to b→e negligible. Fu→rthermore,
→
the uplink compared to the downlink time interval makes the the amount of power consumed for starting up the IBS is
uplink to be considered as a critical path in wireless in-body more than the power consumption during Tac sl. Under the
communications from the power consumption point of view. above considerations, the IBS/CCU devices h→ave to process
More precisely, unlike the CCU which has a power source, oneentireL-bitmessageML during0≤Tac ≤TL−Tsl−Ttr
theIBShaspowerconstraint,anditneedsto reduceitspower in the uplinkmode beforea new sensed packetarrives,where
consumption for minimizing the tissue heating, extending the TL =Ttr+Tac+Tsl with Ttr ≈Tsl ac.
→
batterylifeintheIBSandmeetingthepowerrestrictioninthe Since an implant WBAN communicates over a very short-
MICSstandard.Forthispurpose,weintroduceasynchronized rangeacrossthehumanbody,thecircuitpowerconsumptionis
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 4
TABLEII
comparabletotheRFtransmitpowerconsumption.Wedenote
ELECTRICALPROPERTIESOFTHEMAINTISSUELAYERSOFANADULT
thetotalcircuitpowerconsumptionintheuplinkpathasPc = HUMANBODYAT403.5MHZ
P +P ,whereP andP representthepowerconsumption
ct cr ct cr
of the circuits embedded in the IBS (as the transmitter) and Tissue εr σ (S/m)
Skin(Dry) 46.706 0.6895
the CCU (asthereceiver),respectively.In addition,the power Fat 5.5783 0.0411
consumption of RF signal transmission in the IBS is denoted Muscle 57.10 0.7972
byP .Takingtheseintoaccount,thetotalenergyconsumption
t
TABLEIII
during the active mode period in uplink, denoted by E , is
ac PARAMETERSFORTHEPATHLOSSMODELOFAHUMANBODYAT403.5
given by Eac = (Pc+Pt)Tac. Since the downlink data path MHZ[23]
is onlyused for transmissionof commandsignals, we assume
thatthetotalenergyconsumption(inbothRFtransmissionand Tissue L0 (dB) η σχ (dB)
DeepTissue 47.14 4.26 7.85
circuits)duringthedownlinktransmissionperiodisnegligible
NearSurface 49.81 4.22 6.81
compared to that of the uplink case. The energy consumption
in the sleep mode period is given by E = P T , where
sl sl sl
Psl is the corresponding power consumption. The present the in-body channel model is extremely difficult. Reference
state-of-the art in implantable devices aims to keep a very [23] uses a sophisticated 3D virtual simulation platform to
low sleep mode leakage current (coming from the CMOS extract a simple statistical path loss model for the in-body
circuits in the IBS), resulting in Psl ≈ 0 when compared channel over the MICS frequency band. We denote Pt and
to the power consumption in the active mode duration. This P as the transmitted and the received signal powers for the
r
assumption is used in many studies on the energy efficiency uplink, respectively. In addition, it is assumed that the IBS
of wireless sensor networks (e.g., see [20], [21]). As a result, and the CCU are separated by distance d. In this case, the
the energy efficiency,referredto as the performancemetric of gain factor L for a ηth-power path loss channelis expressed
d
the proposed implant WBAN, can be measured by the total as
energy consumption during the uplink transmission interval
P d η
and corresponding to L-bit message ML as follows: Ld = Pt =L0(cid:18)d (cid:19) χr (2)
r 0
EL =(Pc+Pt)Tac+PtrTtr, (1) d
L (dB) = L (dB)+10ηlog +χ (dB), (3)
d 0 10(cid:18)d (cid:19) r
where P is the circuit power consumption during the tran- 0
tr
sient mode period. It is seen from (1) that the active mode whereL is the gainfactorat thereferencedistanced which
0 0
duration T is an influential factor in energy saving, as E is specified by the transmitter and the receiver antenna gains
ac L
is a monotonically decreasing function of T . and wavelength λ, η is the path loss exponent, and χ ∼
ac r
N(0,σ2) is a normal random variable which represents the
χ
deviation caused by differenttissue layers (e.g., skin, muscle,
III. IN-BODY CHANNELMODEL
fat) and the antenna gain in different directions [23]. Table
Animportantelementintheoptimumdesignofimplantable
III shows the parameters for the statistical path loss model
devices is to characterize a realistic channel model for the
corresponding to the channel between the IBS and the CCU
human body. For this purpose, we first study the tissue
for near-surface (e.g., cardiac pacemaker located in the left
characteristics and their effects on the signal propagationloss
pectoral region) and deep tissue (e.g., digestive endoscopy)
across the human body over the MICS frequency band.
applications.
Humanbodyisaninhomogeneousandlossymediumwhich
For the above channel model and denoting s (t) as the
i
consistsofvarioustissue layerswith differentthicknessesand
RF transmitted signal with energy E , the received signal
t
their own unique electrical characteristics over the frequency at the CCU is given by r (t) = 1 s (t) + z (t), where
of interest. Each tissue layer absorbs the electromagnetic z (t) is the AWGN atthe CCiU with√twLod-siidedpowierspectral
energywhichresultsinaconsiderableattenuationofultralow i
densitygivenby N0.Itshouldbenotedthatthepowerspectral
power waves propagated from the IBS to the CCU. In this 2
density of the noise introduced by the receiver front-end at
work, we consider three different tissue layers composed of the CCU is calculated by N0 = κT010N10F (W/Hz), where
muscle, fat and skin, approximating the structure of a human
κ=1.3806503×10 23 is the Boltzmann constant, T is the
torsoformodelingthechannelbetweentheIBSandtheCCU. − 0
body temperature (in Kelvin) and NF is the noise figure of
This model is used in many recent published works (e.g., see
the receiver at the CCU. Under the above considerations, the
[4],[13]).TableIIdetailsthefrequency-dependentpermittivity
Signal-to-NoiseRatio (SNR), denoted by γ, can be computed
ε andconductivityσ oftheaforementionedtissueofanadult
r as γ = Et .
male at 403.5 MHz [4], [22]. The electrical properties of the LdN0
bodytissueshelpindesigningproperantennasforimplantable
devices,andthein-bodypathlosschannelmodel.Thephysical IV. UNCODED FSK MODULATION
radio propagation with implanted antennas through human A challenge that arises with implantable medicaldevices is
tissueshasbeenextensivelystudiedintheopenliteratureover to design a small size and ultra low power transceiver which
thepastfewyears(e.g.,see[16]anditsreferences),andisout operates efficiently over the MICS frequency band. With this
of scope of this work. Derivation of the exact expression for observation in mind, finding the energy efficient modulation
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 5
with low-complexity implementation is a crucial task in the IVsummarizestheparametersofan uncodedNC-MFSK over
design of tiny biosensors. Broadly, the transceivers can be an AWGN channel with path loss which is the same as the
dividedintotwocategoriesbasedonthefollowingmodulation channelmodelforthehumanbody.Therearesomeinteresting
techniques:i)sinusoidalcarrier-basedand ii)UWB schemes. points extracted from Table IV.
UWBmodulationschemessuchasOOKbenefitinverysimple i) It is observed that the active mode duration T is a
ac
design in transmitters along with low power consumption in monotonically increasing function of constellation size M,
circuits.However,accordingto theFCC regulationsforUWB when L and B are fixed. Since, we are interested in having
systems[15], UWB technologyis used solely fortransmitting T as small as possible for energy saving (according to (1)),
ac
information over a wide bandwidth at least 500 MHz, which as well as having a low complexity detector at the CCU, the
ismuchwiderthanthebandwidthusedintheMICSstandard. higherorderof constellationsize M for the aboveNC-MFSK
In addition, the 3-10 GHz frequency band used for UWB is not desirable. Another advantage of using a lower order
systems is far from the 402-405 MHz frequency range in of M is to improve the bandwidth efficiency which is the
implantable devices. On the other hand, sinusoidal carrier- main concern in MFSK scheme. Of course, a good spectral
based modulation schemes can meet the MICS specifications, efficiency is achieved using Minimum-Shift Keying (MSK)
in particular the frequency band 402-405 MHz and the data and in particular Gaussian Minimum-Shift Keying (GMSK)
rate of at least 250 kbps. Among various sinusoidal carrier- as a special case of FSK.
based modulation techniques, Frequency Shift Keying (FSK), ii) Using B = M = 300 KHz, the data rate is obtained
2Ts
known as the green modulation, has been found to provide as R= b =2Blog2M which must be greater than 250 kbps
a good compromise between high data rate, simple radio accordinTgstotheMIMCSspecificationsinTableI.Itisseenthat
architecture, low power consumption, and requirements on only M =2 and M =4 satisfy this requirementemphasizing
linearity of the modulation scheme [24], [25]. Note that more using the lower order of M for MFSK modulation scheme.
complex modulation schemes such as QAM which are often iii) Assuming M and L are fixed, it is revealed from
used in conventional RF communication applications are not T = ML thattheactivemodedurationinthe IBSwith
ac 2Blog M
easily amenable to the ultra low power communication de- B = 300 KH2z is approximately 20% of T for the wireless
ac
manded by implantable medical devices, due to higher power sensorapplicationswith the same NC-MFSK structure,where
consumption. they use the bandwidth B = 62.5 KHz according to Table
An FSK scheme is a very simple and low power struc- I. Thus, an 80% energy saving is achievable for implantable
ture which is widely utilized in energy-constrained wireless devices with the MICS standard as compared to the IEEE
applications [12], [25] and the IEEE 802.15.6 WPAN Work 802.15.4physical layer protocolwith the same structure used
Group (WG) (e.g., see [19]). In an M-ary FSK modulation for sensor networking applications. This interesting result
scheme, M orthogonal carriers are mapped into b = log2M justifies the reason of using a wider bandwidth in the MICS
bits. The main advantage of this orthogonal signaling is that standard than that of the IEEE 802.15.4 protocol for WSNs.
the received signals at the CCU do not interfere with one Now, we are ready to derive the total energy consumption
another in the process of detection at the receiver. An MFSK ofanNC-MFSKbasedontheresultsinTableIV.Recallfrom
modulator benefits in using the Direct Digital Modulation E =P T ≈2L N ln M , we have
(DDM) approach, meaning that modulation is implemented t t s d 0 4Pb
T L N L M
digitallyinsidethefrequencysynthesizer.Thispropertymakes P T =E ac =2 d 0 ln . (4)
MFSKconsumeverylittlepowerandhasafasterstart-uptime t ac t Ts log2M 4Pb
than other sinusoidal carrier-based modulations. The output Substituting (4), T = ML and P = αP in (1),
ac 2Blog M Amp t
of the frequency synthesizer can be frequency modulated and the total energy consumption2of an NC-MFSK scheme for
controlled simply by b bits in the input of a “digital control” transmittingLbitsduringtheuplinktransmissionintervaland
unit. The modulatedsignalis then filtered again,amplified by for a given P is obtained as
b
the Power Amplifier (PA), and finally transmitted across the
L M
humanbodychannel.Wedenotethepowerconsumptionofthe E = 2(1+α)L N ln +
L d 0log M 4P
frequency synthesizer, filters and the power amplifier as P , 2 b
Sy
ML
cPoFnisltumanpdtioPnAomfpth,ereIsBpSecitsivgeilvye.nInbythPicstc=asPe,Styh+ePcFiriclut+itPpAowmepr, (Pc−PAmp)2Blog2M +1.75PSyTtr. (5)
where P = αP and the value of α is determined based In transmitting the vital physiologicalsigns of patients, the
Amp t
on type of the power amplifier. Also, denoting P , P , BER should be as small as possible. Since, the BER is a
LNA Filr
P , P and P as the power consumption of Low- function of the received SNR and to maintain the BER in
ED IFA ADC
NoiseAmplifier(LNA),filters,envelopedetector,IFamplifier a specific limit, one would increase the transmit signal power
andADC,respectively,thepowerconsumptionoftheCCUcir- whichis notdesirablein ultra low powerimplantablemedical
cuitrywiththe Non-CoherentM-aryFSK (NC-MFSK)canbe devices with EIRP=-16 dBm. One practical solution to avoid
obtainedasP =P +M×(P +P )+P +P an increase in the transmit power is to use channel coding
cr LNA Filr ED IFA ADC
[25]. In addition, it is shown that the energy consumption schemes. For energy optimal designs, however, the impact of
during T is obtained as P T =1.75P T [25]. channelcodingontheenergyefficiencycomputedin (5)must
tr tr tr Sy tr
The energy and bandwidth efficiency of FSK have been be considered as well. It is a well known fact that channel
extensivelystudiedintheliterature(e.g.,see[20],[25]).Table coding is a fundamental approach used to improve the link
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 6
TABLEIV
PARAMETERSFORANUNCODEDNC-MFSKOVERAWGNCHANNELWITHPATHLOSS
Parameter Formula
TransmittedSignal si(t)=q2TEst cos(2π(f0+i∆f)t), i=0,1,...,M−1
SymbolDuration Ts
FirstCarrierFrequency f0
MinimumCarrierSeparation ∆f = 1
2Ts
Channel Bandwidth(Hz) B≈M×∆f = M
2Ts
DataRate(b/s) R= b
Ts
Bandwidth Efficiency(b/s/Hz) Beff = BR = 2loMg2M
Active ModeDuration Tac= LbTs= 2BMlogL2M
BitErrorRate Pb≤ M4 e−γ2 = M4 e−2NE0tLd
TransmitEnergyPerSymbol Et=PtTs≈2LdN0ln4MPb
PowerConsumption ofIBSCircuitry Pct=PSy+PFilt+PAmp
PowerConsumption ofCCUCircuitry Pcr =PLNA+M×(PFilr+PED)+PIFA+PADC
reliability using redundant information bits along with the this may be done by adaptively adjusting the transmitted
transmitter energy saving due to the providingof coding gain powerbasedonthechannelcondition.Suchanadaptivepower
[26].However,theenergysavingcomesatthecostofextraen- control is not feasible for the IBS due to the complexity
ergyconsumedintransmittingtheredundantbitsincodewords concerns for implantable medical devices. LT codes, on the
as well as the additional energy consumption in the process other hand, can adapt to different channel realizations via
ofencoding/decoding.Foracertaintransmissiondistanced,if changing the code rate (instead of the power control) to
theseextraenergyconsumptionsoutweighthetransmitenergy achieve a certain BER. In addition, they have the capability
savingduetothechannelcoding,thecodedsystemwouldnot ofhighdegreeofreliabilityincorrectdeliveryofdatapackets
be energy efficient compared with an uncoded system. In the over a lossy channel.
subsequent sections, we will argue the above issue for LT
iii) Simplicity: The LT encoding process is extremely
codes and show that the LT coded NC-MFSK surpasses the
simplewhencomparedwithsometraditionalblockcodessuch
distance constraint (or equivalently the dynamic environment
as Reed Solomon (RS) codes and Low-Density-Parity-Check
between the IBS and the CCU) achieving a given BER in the
(LDPC) codes, and has a tight power budget. In addition,
proposed system.
most of the complexity of an LT code (i.e., message passing
decoder)ispushedtotheCCUwhichhaseffectivelyunlimited
V. ENERGY EFFICIENCY ANDRELIABILITY OFLT CODES power and computational resource.
TheabovecapabilitiesofLTcodeshaveremovedthecritical
The task of the IBS is to communicate vital signs of the
obstaclesofusingtheclassicalfixed-ratecodes(e.g.,blockand
patient as accurately as possible to the CCU, due to the
convolutional codes), where the transferred data experience
potentially life-threatening situations in patient monitoring.
different packet loss or error rates in different SNRs due
This is called data reliability which depends strongly on
to the dynamic channel conditions for the human body. In
the channel conditions. Reliability in a wireless network, in
this section, we briefly introduce some basic concepts and
general,is evaluatedin termsof the probabilityof packetloss
definitions for the LT codes. Then, we analyze the energy
or the probability of a data packet being delivered correctly
efficiency and the reliability of LT coded NC-MFSK for the
to the receiver. Enhanced reliability is achieved by adding
proposed implant WBAN. To get more insight into how LT
redundant bits in data packets in the FEC coding stage at
codes affect the circuit and RF signal energy consumptions
the cost of an extra power consumption. Thus, there exists
in the system, we modify the energy concepts in Section IV,
a tradeoff between a higher reliability and a lower power
in particular, the total energy consumption expression in (5)
consumption in using FEC codes. Finding a proper FEC
based on the LT coding gain and code rate.
codingschemeforbalancingenergyconsumption,complexity
and data reliability is a challenging task in designing implant Following the notation of [27], an LT code is specified by
WBANs; in particular, when data transfer across tissues is the number of input bits k and the output-node degree dis-
susceptible to loss and transmission errors. The emerging LT tribution Ω(x) = k Ω xi, where Ω , i = 1,...,k, denotes
i=1 i i
codes (as the first practical rateless codes) have exhibited the probability thaPt an output node has degree i. Without loss
strongcapabilitiesinreachingtheabovetargetsoverthelossy of generality and for ease of our analysis, we assume that
channelmodel for the human body for the following reasons: a finite k-bit message B = {m }k ∈ M is encoded to
1 i i=1 L
i) Optimality: It is shown in [18] that LT codes are near the codeword C = {a }n with Ω(x) : {m }k → a ,
1 j j=1 i i=1 j
optimalerasurecorrectingcodeswithouttheknowledgeofthe j = 1,...,n. More precisely, each single coded bit a is
j
channel erasure rate at the transmitter. generated based on the encoding protocol proposed in [18]:
ii) Code-Rate Flexibility: To ensure reliable communica- i) randomlychoose a degree 1≤D≤k from a priori known
tion across tissues, a sufficient amount of energy must arrive degree distribution Ω(x), ii) using a uniform distribution,
at the CCU. For fixed-rate codes (e.g., linear block codes), randomly choose D distinct input bits, and calculate the
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 7
(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:2)(cid:6)(cid:7)(cid:8)(cid:8)(cid:9)(cid:10)(cid:7)(cid:2)(cid:11)(cid:7)(cid:12)(cid:13)(cid:7)(cid:14)(cid:15)(cid:7)
nearly identical manner to the “Algorithm E” decoder in [30]
for Low-Density Parity-Check (LDPC) codes. Description of
the ternary message passing decoding is out of scope of this
work, and the readeris referredto Chapter 4 in [28] for more
details. It should be noted that the degree distribution Ω(x)
in (6) was optimized for a ternary decoder in a BSC and we
areawareofnobetterΩ(x) fortheternarydecoderinAWGN
channels.
ToanalyzetheenergyefficiencyoftheLTcodedNC-MFSK
(cid:2)(cid:16)(cid:14)(cid:15)(cid:17)(cid:18)(cid:7)(cid:18)(cid:2)(cid:3)(cid:4)(cid:5)(cid:2)(cid:11)(cid:5)(cid:19)(cid:7)(cid:9)(cid:20) scheme, we note that the number of transmitted bits during
the uplink transmission period is increased from the L-bit
Fig.3. FactorgraphofanLTcodewithk=3andn=4.
uncoded message to L = Ln bits coded one. In order to
Rc k
keep the bandwidth of the coded system the same as that of
encoded bit a as the XOR-sum of these D bits. The above the uncodedcase, we must keep the informationtransmission
j
encoding process defines a factor graph connecting encoded rate constant, i.e., the symbol duration Ts of uncoded and
nodes to input nodes (see, e.g., Fig. 3). It is seen that the LT coded NC-MFSK would be the same. According to Table IV,
encoding process is extremely simple and the encoder has a however,the active mode duration increases from Tac = LbTs
in the uncoded system to
much lower computational burden than the block codes. For
instance,thecomplexityoftheLTcodesisoforderO(nlogn) L T ML
comparedtoO(n2)forRScodesusedinWiMAXintheIEEE Tac,c = bR Ts = Rac = 2R Blog M, (7)
c c c 2
802.16 standard. This low complexity comes from exploiting
XOR operationsin LT codes (unlike field operations used for for the LT coded case, where we use Tac = 2BMlogLM in
Table IV. An interesting point raised from (7) is that2 T
RS codes). Thus, one can with a good approximation assume ac,c
is a function of the random variable R which results in an
that the energy consumption of an LT encoder is negligible c
inherentadaptiveduty-cyclingforpowermanagementin each
compared to the other circuit components in the IBS. As a
channelcondition.Tocomputethetotalenergyconsumptionof
result, the energy cost of the IBS circuitry with the LT coded
codedscheme,we usethefactthatchannelcodingcanreduce
NC-MFSK scheme is approximately the same as that of the
the required SNR value to achieve a given BER. Denoting
uncoded one.
codOenweoirndhebrleonctkprloepnegrtthysotfoLmTactocdhesaiswtihdaettrhaenygceanofvaproyssthibelier γsch=emLEedNst,0reasnpdecγtciv=elyL,EdttNh,ce0 caosdtihnegSgNaRinoGfcun(ecxopdreedssaenddincoddBed)
is defined as the difference between the values of γ and γ
channel conditions. In fact, for a specific value of SNR, c
an ∈ C1 is the last bit generated at the output of the LT trheiqsuiinretodatoccaocuhnite,vteheapcreorptaoisnedBEimRp,lawnhtesryestEemt,cw=ithEGctL.TTcaokdinegs
encoderbeforereceivingtheacknowledgementsignalfromthe
CCU in the uplink path indicating terminationof a successful benefits in transmission energysaving specified by Gc. Tables
V and VI (also see Fig. 4) give the LT code rates and the
decoding process. This is in contrast to classical linear block
corresponding coding gains of LT coded NC-MFSK scheme
and convolutionalcodes, in which the codeword block length
is fixed. As a result, the LT code rate, denoted by R = k, using Ω(x) in (6) and for M = 2 and 4 and different values
c n of BER. It is observed that the LT code is able to provide a
displays a random variable indicating the rateless behavior of
the LT code. largevalue of codinggain Gc givenBER, butthis gain comes
at the expense of a very low code rate, which means many
We now turn our attention to the LT code design for the
additional code bits need to be sent. This results in higher
augmentation protocol. As seen in the LT encoding process,
energy consumption per information bit. An interesting point
theoutput-nodedegreedistributionΩ(x)isthemostinfluential
extractedfromFig.4istheflexibilityoftheLT codetoadjust
factor in the complexityof the LT code. Typically,optimizing
its rate (and its corresponding coding gain) to suit channel
Ω(x) for a specific wireless channel is a crucial task in
conditions in implant WBANs. For instance, in the case of
designingLTcodes.Inthiswork,weusethefollowingoutput-
favorablechannelconditions,theLT codedNC-MFSK is able
nodedegreedistributionwhichwasoptimizedforaBSCusing
toachieveacoderateapproximatelyequaltoonewithG ≈0
a hard-decision decoder [28], [29]: c
dBin the BER=10 3, whichis similar to the case of uncoded
−
Ω(x) = 0.00466x+0.55545x2+0.09743x3+ NC-MFSK. The effect of LT code rate flexibility on the total
0.17506x5+0.03774x8+0.08202x14+ energyconsumption is also observed in the simulation results
0.01775x33+0.02989x100. (6) in the subsequent section.
To getmore insightinto the resultsof Tables V and VI, we
The LT decoder at the CCU is able to reconstruct the illustrate the LT code rates versus SNR for several different
entire k-bit message B with a high degree of reliability after valuesofBER forthe caseswhereM =2 andM =4 inFig.
1
receivingany(1+ǫ)k bitsinitsbuffer,whereǫdependsupon 5. These results were obtained by evaluating the performance
the LT code design and channel condition [27]. In this work, of the LT code using density evolution, as described in [28].
we assume that the CCU recoversa k-bitmessage B using a An interesting phenomenon exhibited in the figure is that the
1
simple hard-decision “ternary message passing” decoder in a curves tend to converge for a given M as SNR decreases.
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 8
TABLEV
CODERATEANDCODINGGAIN(dB)OFLTCODEDNC-MFSKOVER 12 M=2
AWGNFORDIFFERENTVALUESOFPbANDM=2. Pb=10−3
Pb=10−3 Pb=10−4 Pb=10−5 10 PPbb==1100−−45
SNR(dB) Rc Gc Rc Gc Rc Gc 8
3 0.1624 7.94 0.1624 9.31 0.1624 10.35
4 0.2346 6.94 0.2346 8.31 0.2346 9.35 6
5678 0000....3456234540318483 5432....99994444 0000....3456234540318483 7654....33331111 0000....3345186408951998 8765....33335555 Coding Gain (dB) 24
9 0.7378 1.94 0.7378 3.31 0.6075 4.35
10 0.8003 0.94 0.8003 2.31 0.6479 3.35
0
11 0.8650 -0.06 0.8419 1.31 0.6671 2.35
12 0.9229 -1.06 0.8488 0.31 0.6732 1.35
−2
13 0.9553 -2.06 0.8498 -0.69 0.6748 0.35
14 0.9685 -3.06 0.8508 -1.69 0.6754 -0.65
−4
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
LT Code Rate
TABLEVI
(a)
LTCODERATEANDCODINGGAIN(dB)OFLTCODEDNC-MFSKOVER
AWGNFORDIFFERENTVALUESOFPbANDM=4.
M=4
Pb=10−3 Pb=10−4 Pb=10−5 10 Pb=10−3
SNR(dB) Rc Gc Rc Gc Rc Gc Pb=10−4
7 0.1613 7.37 0.1613 8.64 0.1613 9.62 8 Pb=10−5
8 0.2380 6.37 0.2380 7.64 0.2380 8.62
9 0.3365 5.37 0.3365 6.64 0.3190 7.62 6
10 0.4525 4.37 0.4525 5.64 0.4051 6.62
11111234 0000....5678786358844513 3210....33337777 0000....5678786358844513 4321....66664444 0000....4566972620805995 5432....66662222 Coding Gain (dB) 24
15 0.9032 -0.63 0.8468 0.64 0.6720 1.62
16 0.9486 -1.63 0.8498 -0.36 0.6746 0.62 0
17 0.9654 -2.63 0.8503 -1.36 0.6748 -0.38
−2
Thisresultcomesfromthefactthatforlargeblocklengths,LT
−4
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
codesexhibitadecodingthresholdinmuchthesamewaythat LT Co(deb R)ate
LDPC codes do. Essentially, this means that for a particular Fig.4. LTcodinggainversusLTcoderatefordifferentvaluesofBERand
SNR there exists a threshold rate above which the LT code fora)M =2,b)M =4.
will not successfully decode. At or below this threshold rate,
the LT code will decode and the decoded message will have
NC-MFSK scheme, denoted by E , and for a given P is
a particular BER. As the rate is further decreased, the BER L,c b
obtained as
will decrease in a continuous manner. Thus, there exists a
L N L M
discontinuityin the BER at the decodingthreshold,where the E = 2(1+α) d 0 ln +
BER of an LT decoded message drops from being very high L,c Gc Rclog2M 4Pb
(i.e., many errors since the LT decoder failed to converge to ML
(P −P ) +1.75P T +
a solution for rates above the threshold) to very low (i.e., the c Amp 2R Blog M Sy tr
c 2
LT code converged at a successful solution for rates below E +E
L enc dec, (9)
thethreshold).Thisdiscontinuityismorepronouncedatlower R
c
SNR, meaning that the BER values of 10 3, 10 4, and 10 5
− − − where the goal is to minimize E in each distance d, in
L,c
all appear to occur at the same points on the rate-SNR curve.
termsof modulationand codingparameters.Although,higher
We denoteE andE asthecomputationenergyofthe
enc dec reliability (corresponding to the lower BER) results in an
encoder and decoder for each information bit, respectively.
increase in the total energy consumption E , the effect of
L,c
Thus, the total computation energy cost of the coding com-
the coding gain G and the code rate R on E for a given
ponents for L bits is obtained as LEenc+Edec. Substituting c c L,c
sEigtn=al2eLndeNrg0yRlcnco4nMPsbuminptEiotn,cd=uriEGnctgaancdtivuesimRngcod(e7)d,utrhaetitoontaTl RF Pb should be considered as well.
ac,c VI. NUMERICAL EVALUATION
for the coded case is given by
This section presents some numerical evaluations using
T L N L M
P T =E ac,c =2 d 0 ln . (8) realistic parameters from the MICS standard and state-of-
t,c ac,c t,c T G R log M 4P
s c c 2 b the art technology to confirm the reliability and energy effi-
Substituting (7) and (8) in (1), and using the fact that ciencyanalysisofuncodedandLTcodedNC-MFSK schemes
P = αP , the total energy consumption of transmitting discussed in Sections IV and V. We assume that the NC-
Amp t,c
L bitsduringtheuplinktransmissionperiodforanLTcoded MFSK modulation scheme operates in the carrier frequency
Rc
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 9
optimized uncoded NC-MFSK for the cases P = 10 3
b −
and P = 10 5, and for the proposed three different tissue
b −
layers with the aforementioned thicknesses. The optimization
1 is performedover M and the parametersof codingscheme in
0.9 MM==22,, PPbb==1100−−34 Tables V and VI. For the case of Pb = 10−3, it is revealed
M=2, Pb=10−5 from Fig. 6-a that for distance d less than the threshold
0.8 MM==44,, PPbb==1100−−34 level dT ≈ 70 mm (near surface scenario), the total energy
0.7 M=4, Pb=10−5 consumptionofoptimizeduncodedNC-MFSKislessthanthat
Rate0.6 of the coded NC-MFSK schemes. However, the energy gap
LT Code 0.5 basetwexepenectthede.LFTocrodded>andd unwcohdicehd acpopvreoraschthees idseneepgltiigsisbulee
T
0.4 scenario, the LT coded NC-MFSK scheme is more energy
efficient than the uncoded one, in particular when d grows.
0.3
This result comes from the high coding gain capability of
0.2
LT codes. An interesting phenomenon exhibited in Fig. 6-
0.1 a is that the optimized LT coded NC-MFSK surpasses the
4 6 8 10 12 14 16
SNR (dB)
distanceconstraintinimplantWBANapplications.Inaddition,
the flexibility of the LT code rate (and the corresponding LT
Fig.5. LTcoderateversusSNRfordifferent values ofBERandM=2,4.
coding gain) allows the system to adjust its duty-cycling for
powermanagementandto suitto differentchannelconditions
TABLEVII
SYSTEMEVALUATIONPARAMETERS for the human body for any distance d.
In the case of P = 10 5 as depicted in Fig. 6-b,
b −
B=300KHz Ttr =5µs PADC =7mw the uncoded NC-MFSK scheme is preferable to use for
d0=30mm PSy =10mw PLNA=9mw
N0=−110.91dBm PFilt=2.5mw PED =3mw d < dT ≈ 120 mm which corresponds to the near-surface
TL=1.4sec PFilr =2.5mw PIFA=3mw applications.This result followsfrom the maximumcode rate
of R ≈ 0.675 for both M = 2 and M = 4 in P = 10 5.
c b −
However,forthedeeptissueapplicationswhered>d ≈120
T
f0 =403.5MHz accordingto the frequencybandrequirement mm,theoptimizedLTcodedNC-MFSKismuchmoreenergy
in the MICS standard. As shown in Section IV, we only efficient than the uncoded case.
consider the constellation sizes M = 2,4. We assume that Itis concludedfromthe aboveobservationsthatthe thresh-
in each period T , the sensed data frame size L = 1024 old level d for a certain BER is a fundamentalparameter in
L T
bytes (or equivalently L = 8192 bits) is generated, where thephysicallayerdesignof implantWBANs, asit determines
T is assumed to be 1.4 seconds. The channel bandwidth is “when the LT codes are energy efficient” in the proposed
L
assumed to be B = M = 300 KHz, according to Table I. wireless biomedical implant system. The parameter d is
2Ts T
As a result, the symbol duration T for M = 2 and M = 4 obtained when the total energy consumptions of coded and
s
η
is obtained as 3.33 µs and 6.67 µs, respectively. Also, the uncoded systems become equal. Using L = L d χ ,
data rate R = Tbs for both M = 2,4 will be 300 kbps andtheequalitybetween(5)and(9)foruncdodedan0d(cid:16)LdT0(cid:17)coderd
whichsatisfies the datarate requirementin Table I.Assuming
NC-MFSK, we have
T =310 K forthe bodytemperatureand NF=8 dB, the total
0 ◦ 1
noise power at the bandwidth B = 300 KHz is obtained as
N =κBT 10NF/10 =−110.91dBm. TableVII summarizes M(P −P ) G (1−R ) η
the0systemp0arametersforsimulation.TheresultsinTablesIII- dT =d04(1+α)N cBL χAmlnp M GccRc−1c  .
VI are also used to comparethe energyefficiency of uncoded  0 0 r (cid:16)4Pb(cid:17) 
and LT coded NC-MFSK schemes. In addition, the current It can be seen from the above equation that a decrease in
simulation is based on the values in Table III for the two the P results in an increase in the d as observed in Fig.
b T
followingscenarios:i)nearsurfacewiththedistancerangeof 6. In addition, the growth rate of the energy consumption of
d0 =30mmto100mmandii)deeptissuewherethedistance the LT codes for deep tissue applications (i.e., d > dT) is
d varies between 100 mm and 300 mm. For this purpose, it muchsmallerthanthatoftheuncodedonewhenP decreases
b
is assumed thatthe skin varies in thicknessfrom0.5 mm to 6 (equivalent to the higher reliability).
mm[31]andthemusclethicknessvariesintherangeof13-40 Remark: As discussed previously, the proposed LT coded
mmaccordingtotheMRImeasurementsin[32].Furthermore, scheme benefits in adjusting the coding parameters in each
we assume that the fat thickness ranges from 0-250 mm for channelrealizationorequivalentlyeachdistancedtominimize
thin or obese patients. In order to estimate the computation thetotalenergyconsumption.Forthiscase,aslongastheCCU
energy of the channel coding, we use the ARM7TDMI core collects a sufficient number of bits, it will be able to decode
which is the industry’s most widely used 32-bit embedded theoriginalmessage.Indeed,onlytheCCUneedstobeaware
RISC microprocessor for an accurate power simulation [33]. of the channelconditions.The IBS, as the transmitter, always
Fig. 6 shows the total energy consumption versus distance sends an LT encoded stream using constant modulation and
d for the optimized LT coded NC-MFSK compared to the power, and the receiver alone determines how many bits to
IEEETRANSACTIONSONINFORMATIONTECHNOLOGYINBIOMEDICINE 10
10−1 Pb=10−3 codes.
Uncoded
Optimized Coded
VII. CONCLUSION
In this paper, we presented an augmentation protocol for
Total Energy Consumption (Joule)10−2 d < dT d > dT tespfthhocnreehoreeprppgmLoohyTsewyeedsoecfirvcfioseacdcmrliheeatelnhanmrceyaayeegtMerea.pmInorCAodfeSvnnttithahdflereweyestqhrMieuciaclanIeihClanbicSncriyeolhissmbettuyraaleenntsonsddfta.fdrraLIodetdTmmawwpocattionsihtvdhseseterhdfaflodotewuecNxdutnyCisb--ttcihhMolyiaantcFttylSittnhhaoKngeef
80% energy saving is achievable with the proposed protocol
when compared to the IEEE 802.15.4 physical layer protocol
with the same structure used for wireless sensor networks.
10−30 50 dT=70 100 Transmission1 D50istance d (mm) 200 250 300 Numerical results have shown a lower energy consumption
keeping at the same time a higher degree of reliability in
(a)
the LT coded NC-MFSK scheme for deep tissue scenarios
whencomparedtotheuncodedcase.Inaddition,itwasshown
10−1 Pb=10−5 that the optimized LT coded NC-MFSK is capable to surpass
Uncoded
Optimized Coded the distance constraint in implantable medical devices for the
BER=10 3.Infact,sincetheefficiencyofimplantablemedical
−
devices is strongly related to the human body structure (e.g.,
mption (Joule) d < dT d > dT ooovbpeeersrcaeoteomsreinsthtaihnehaipgnahdtideteinsgts’ruseeevioctofalmacipncofuosriratmicoyanto)io,vnethrleoasnpsyroadpnyodnsaeemrdroiscrcsch,heaamnnde-
Total Energy Consu10−2 ndueenlsiqicguoneneiddnitwiimointphslafsnoitrmWtphlBeicAhituNymaaapnnpdbliflocdaextyii.obTnilshi,teyasiniinttrohmdasiuncbdee,deanpnredoxtposlhciocoiultllidys
be usable in current wireless medical networking hardware.
ACKNOWLEDGMENT
10−30 50 100 TdrTa=n1s2m0ission1 D(50ibsta)nce d (mm) 200 250 300 The authors would like to thank Ali Tawfiq (University of
Fig.6. TotalenergyconsumptionofuncodedandoptimizedLTcoded Toronto) for graciously editing this paper.
NC-MFSK versus d for a) P =10−3 b) P =10−5.
b b
REFERENCES
[1] R. Schmidt, T. Norgall, J. Mrsdorf, J. Bernhard, and T. von der Grn,
collect, their reliability, and ultimately decodes the message. “Body area network BAN: a key infrastructure element for patient-
Thus, the transmitter does not need to change coding rate, centered medical applications,” Biomed Tech (Berl), vol. 47, pp. 365–
368,Jan.2002.
because the rate is determined entirely by how many bits it
[2] U. Varshney and S. Sneha, “Patient monitoring using ad hoc wireless
collects based on the channel quality. There is no channel networks: reliability and power management,” IEEE Communications
information required at the transmitter which reduces the Magazine, vol.44,no.4,pp.49–55,April2006.
[3] E. Reusens, W. Joseph, B. Latr, B. Braem, G. Vermeeren, E. Tanghe,
complexity and cost in implementation.
L. Martens, I. Moerman, and C. Blondia, “Characterization of on-
IncontrasttotheARQscheme,theonlyfeedbackintheLT body communication channel and energy efficient topology design for
coded system is a single message from transmitter to receiver wireless bodyareanetworks,” IEEETrans.onInformation Technology
inBiomedicine, vol.13,no.6,pp.933–945,Nov.2009.
(attheendofalongblock)indicatingthatasufficientnumber
[4] K. Y. Yazdandoost and R. Kohno, “Body implanted medical device
ofbitshavebeencollectedanditistimetomoveontothenext communications,” IEICE Trans. on Commun., vol. E92-B, no. 2, pp.
data block.Thisis notan ARQ scheme, butit is more similar 410–417,Feb.2009.
[5] FCC, “Wireless Medical Telemetry,” Federal Communications Com-
to an incremental redundancy scheme whereby a message is
mission,Jan.2000, onlineavailable athttp://wireless.fcc.gov.
encoded using a rate-compatible punctured code. [6] R.A.M.Receveur,F.W.Lindemans,andN.F.D.Rooij, “Microsystem
technologies forimplantable applications,” Journal ofMicromechanics
In addition, a nice feature about the optimized LT code
andMicroengineering, vol.17,no.5,pp.R50–R80,May2007.
as opposed to the punctured codes is that there is truly no [7] “Federal Communications Commission (FCC), code of federal regula-
“minimumrate”atwhichthecodecanoperate.Moreprecisely, tions(CFR),title47part95,MICSbandplan,” FCC,Washington,DC,
March2003, onlineavailable athttp://www.fcc.gov.
we can choose a code that allows us to drop the rate as
[8] IEEE Standards, “Part 15.4: Wireless Medium Access control (MAC)
low as we need. With punctured LDPC, it is reported in and Physical Layer (PHY) Specifications for Low-Rate Wireless Per-
[34] that depending upon the operating range, rate, etc. the sonal Area Networks (WPANs),” in IEEE 802.15.4 Standards, Sept.
2006.
LDPC performance drops off dramatically beyond a certain
[9] “Range extension for IEEE 802.15.4 and ZigBee applications,”
rate threshold–a phenomenon not observed by the rateless FreeScaleSemiconductor, Application Note, Feb.2007.