Archive of SID

IJST, Transactions of Electrical Engineering, Vol. 38, No. E1, pp 1-20

Printed in The Islamic Republic of Iran, 2014
© Shiraz University

DESIGN AND IMPLEMENTATION OF PRECISE HARDWARE FOR

*
ELECTRICAL IMPEDANCE TOMOGRAPHY (EIT)

M. KHALIGHI1** , B. VOSOUGHI VAHDAT2, M. MORTAZAVI3 AND M. MIKAEILI4

1
School of Engineering and Science, Sharif University of Technology, International Campus, I. R. of Iran
Email: khalighi58@yahoo.com
2
School of Engineering and Science, Sharif University of Technology, Tehran, I. R. of Iran
3
School of Engineering, Islamic Azad University-Abhar Branch, I. R. of Iran
4
Dept. of Engineering, Biomedical Engineering Group, Shahed University, Tehran, I. R. of Iran

Abstract– Electrical Impedance Tomography (EIT), is one of the safest medical imaging
technologies and can be used in industrial process monitoring. In this method, image of electrical
conductivity (or electrical impedance) distribution of the inner part of a conductive subject can be
reconstructed. The image reconstruction process is done by injecting an accurate current into the
boundary of a volume conductor (Ω), measuring voltages around the boundary (∂Ω) and
transmitting them to a computer, and processing on acquired data with software (e.g. MATLAB).
The image would be reconstructed from the measured peripheral data by using an iterative
algorithm. A precise instrumentation (EIT hardware) plays a very important and vital role in the
quality of reconstructed images. In this paper, we have proposed a practical design of a low-cost
precise EIT hardware including, a high output impedance VCCS (Voltage-Controlled Current
Source) with pulse generation part, precise voltage demodulator and measuring parts, a high
performance multiplexer module, and a control unit. All the parts have been practically and
accurately tested with successful results, and finally the proposed design was assembled on PCB.
The quality of experimental results at the end of this paper, (reconstructed images by using the
implemented system), confirms the accuracy of the proposed EIT hardware.

Keywords– EIT, electrical impedance tomography, EIT hardware, EIT instrumentation, EIT current source

1. INTRODUCTION

Electrical Impedance Tomography (EIT) is a relatively new imaging technique. In this method, an image
of the inner part of a conductive domain (Ω) can be made with an array of external electrodes which are
located on the boundary of domain (∂Ω). In this imaging method, the image of electrical conductivity (or
impedance) distribution of the internal part of a typical conductive subject can be reconstructed [1]. EIT
procedure includes injecting an accurate current into the boundary of domain (∂Ω) via a pair of electrodes,
measuring the boundary voltages by means of other electrodes around the boundary and transmitting them
to a computer, and at the end, processing the acquired data with software (e.g. MATLAB) to reconstruct
the image. Human body tissues contain a wide range of conductivities, and hence the potential exists to
use EIT to carry out medical imaging using the conductivity as the parameter to be mapped [2].
As a matter of fact, EIT is a challenging problem. This technique has some advantages compared to
other methods, including: simplicity of application, no hazard to the patient (such as X-ray), low cost and
portable, and the high speed of data collection and image reconstruction [3]. Although EIT systems suffer


Received by the editors January 21, 2013; Accepted May 18, 2014.

Corresponding author

www.SID.ir
Archive of SID

2 M. Khalighi et al.
l.

from poor im mage resoluttion, the trannsfer impedannces must still be measurred with highh accuracy. T The effect
of unknownn skin–electrrode impedannce on measuurement erroors can also have h a signifficant effect on image
quality [4]. EIT still hass technical difficulty
d in terms
t of devveloping harddware for daata acquisitioon and the
algorithms tto reconstrucct the images. However, there are soome methodss to improvee imaging quuality. The
most imporrtant of them m is hardwarre improvem ments to acqquire a high--accuracy meeasured datuum [5, 6].
Reconstructted image quality
q mainnly depends on the bouundary data accuracy an nd the reconnstruction
algorithm [77].
This paaper is focussed on hardw ware implemmentation of EIT.
E First off all, the top view of our proposed
design is illlustrated andd explained, and each paart of the dessign will be described inn detail as well.w After
that, some experiments
e done by usiing the impleemented systtem and theiir results willl be presenteed. At the
end, final reeconstructed image of ouur EIT system m will be com
mpared with other works..

Fig. 1. The maain block diaggram of proposed EIT hardw

ware design

2. TO
OP VIEW OF
O PROPOS
SED DESIG
GN

In general, E EIT instrumentation com mprises digitaal and analogg circuits, ann array of eleectrodes whiich acts as
EIT-sensorss and a PC. The image is i finally recconstructed from
f the acqquired data (voltages) byy using an
iterative alggorithm. In our design, thhe analog secction of hardw
ware includees an accuratte current souurce and a
data acquisiition part (too measure vooltages with enough preccision). The digital part contains conntrol unit.
The precisioon of all uniits plays a veery importannt role in reconstructed im mage qualityy. The top viiew of the
proposed EIIT system is illustrated inn Fig. 1. Thee block diagrram of the prroposed EIT hardware (F Fig. 1) has
different paarts:
1- VCCS (Voltage-Coontrolled Cuurrent Sourcee) which includes a VCO O as a wavefoorm generatoor, a filter,
a VCCC (Voltage-too-current connverter) part, and a pulse generator.
2- Multipplexer module.
3- Controol unit.
4- Voltagge measurement part conttaining a volttage demoduulator and a voltmeter.
v

The coonstant and accurate

a currrent produceed by the VC CCS is injeccted into thee boundary ofo domain
(∂Ω) via thhe multiplexeer part (MU UX). The seccond task off the current source section is generrating two
types of puulse train sim w maximuum point of positive peaaks and zero points of thhe current
multaneous with
waveform. They can bee used for saampling of reeal and imagginary sectionns of measurred voltage in i voltage
demodulatioon portion. InI order to m measure the boundary vooltage, mean ning demoduulating and measuring
m
the electroddes voltage, “V-Demo.”
“ aand “V-Meteer” blocks shhould be appplied respectiively after multiplexer
m
block. Conttrol unit is appplied for coontrolling thee multiplexerr module and measurement parts, annd also for
communicaating with thee PC for dataa transaction.. The assembbled EIT harddware is shown in Fig. 2.

IJST, Transacctions of Electrrical Engineeriing, Volume 38

8, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 3

Fig. 2. Asssembled EIT hardware

h

3. VOLTA
AGE-CONTR
ROLLED C
CURRENT SOURCE
S

Electrical Im mpedance Tomography (EIT) system ms require acccurate curreent sources that
t work ovver a wide
frequency range
r and witth a large vaariation in loaad impedancce. The simplest techniquue to obtain a constant
current is too use a voltaage-controlleed current soource (VCCS S), which caan be definedd as a combination of
positive andd negative feeedback arouund a high gain operationnal or instrum mentation ammplifier [8]. A typical
VCCS usedd in the EIT systems,
s connsists of wavveform generrator and Voltage-to-Currrent Convertter (VCC)
parts. Desiggn of the VCC in EIT sysstems is espeecially imporrtant. The VC CC part requuires stabilityy and high
precision, wwhich means that it must have high ouutput impedaance. Generaally, the waveform used iin the EIT
systems in m most cases isi the sinusoiidal waveforrm which is produced byy the wavefoorm generation part in
digital or annalog form [9].
[ The mainn features foor a typical waveform
w geenerator are as
a follow; ann accurate
output waveeform (i.e. exxact sinusoiddal waveform m without anyy jitter), a low output imp pedance (simmilar to an
ideal voltagge source), wide
w operatioonal bandwiddth, and steaady amplitud de over all thhe frequency range. In
addition, thhe main charaacteristics foor the VCC part p to have an excellennt VCCS are based on hiigh output
impedance, linearity in converting voltage to current, c precision of outp put waveform, supportinng a wide
range of looad, and propper workingg in a broadd range of frrequency [100]. The high h-performancce current
source can aameliorate thhe image quaality to a certtain extent [111].
The cuurrent source in EIT systeems must be able to delivver the currennt over a freq quency rangee between
10KHz andd 1MHz and be able to support s the load betweeen 100Ω and d 10KΩ [9]. Its output im mpedance
must be in tthe range of more than 1000KΩ as weell [12], thereefore an exceellent currentt source design should
satisfy the aabove condittions. Figuree 3 shows th he flowchart of the propoosed VCCS. It consists of o a VCO
(Voltage-Coontrolled Osscillator) partt, a Butterwoorth band-paass filter (Buutterworth lo ow-pass and high-pass
filters in serrial connectioon), and a VC CC (Voltagee-to-Current Converter) part.p

Figg. 3. Flowcharrt of proposed VCCS designn

a) Waveforrm generatio
ion
Schemaatic circuit of
o the wavefform generattion part is sshown in Figg. 4. As it can
c be seen, XR-2206
(EXAR Inc.) has been used
u in the waveform
w osscillator partt as a VCO. An AD844 IC has been placed at
June 2014 IJST,
I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

4 M. Khalighi et al.
l.

the output oof the VCO,, to amplify the sinusoiddal waveform

m and also to
t create a low output im
mpedance
value at the output off waveform generation partp (measurred output impedance
i o the circuiit without
of
AD844, is aabout 690Ω and
a with thiss IC, is aboutt 14Ω).

Fig. 4. Voltagge-controlled oscillator (waaveform generration part)

b) Butterw
worth band-p
pass filter
In ordeer to have ann exact wavveform witho out any jitterr, noise and distortion, and
a also to create
c low
output impeedance for th he next stagee (VCC), a Butterworth
B band-pass fiilter Rangingg between 100KHz and
250KHz, haas been desiigned and pllaced betweeen the VCO and VCC parts p of curreent source. Hence
H the
operational frequency ranger of the designed EIIT system would be betw ween 10 and d 250KHz. Inn order to
design the filter
f for otheer frequenciees, the Butterrworth coeffficients have been cited in [13]. The schematic
s
circuit of thhe filter is shown in Fig. 5. As it is sh
hown, the filtter consists of
o a fourth-order Butterw worth low-
pass filter ((U1 and U2) with cutoff frequency of 250 KHz and a also a foourth-order Butterworth
B high-pass
filter (U4 annd U5) with a cutoff frequuency of 10 KHzK that are connected serially.
s

Fig. 5. Buutterworth Baand-pass filterr which is put between the VCO

V and VCC
C parts

IJST, Transacctions of Electrrical Engineeriing, Volume 38

8, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 5

c) Voltage--to-current converter
c
The ennd structure of
o mentionedd flowchart is
i Voltage-too-Current connverter (VCC
C), many expperiments
were done to improvee and modify its operaation. Severral essential tests were done on four f main
configuratioons of a VCC
CS:
 VCCS S based on AH
A (Advanceed Howland), [14, 15].
 VCCS S based on DOA
D (Doublee-Operationaal Amplifier), [16].
 VCCS S based on TOA1 (Triplee-Operational Amplifier form
f 1), [17]].
 VCCS S based on TOA2 (Triplee-Operational Amplifier form
f 2), [9].

These experimentss were done with identiccal VCOs, different d VCC C structures,, different coomponent
values suchh as differentt resistance values,
v different types of OP-AMP, and also diffferent rangees of load.
According tto results off the primaryy practical teests, the best structure of o the VCC part p was sellected and
modified. TThe main seelected criteeria in the primary
p testss were baseed on the ou utput impeddance, the
operational band-width h, and the minimum
m annd maximum m loads thaat each VCC CS can suppport. The
schematic ccircuits of different
d VCCSs which have been used u in pracctical tests and
a simulations (with
PSPICE), aand their resuults have beeen shown in Appendix A. A Proposed structure
s of the VCC parrt with its
componentss value, accoording to thee best resultt in experimeents is show wn in Fig. 6. For VCCS based on
TOA1 circuuit, (2) shouldd be satisfiedd and therefoore IL is calcuulated by (3)).

IL = (1)

Here, if: – =0, or R1 . R4 = R2 . R3 (2)

We can get: IL .
.Vi (3)

Fig. 6. Proposed
P struccture of the VC w suggested components values
CC section, with v

The ouutput impedaance Zout of a typical currrent source can

c be calcullated by (4). As it is show wn in Fig.
7, V1 repressents the meeasured outpput voltage of o R1, when the switch S is opened and V2 repreesents the
measured output
o voltagge of R1, when S is closeed. RP and R1 are conneected at the output of VC CCS with
optional vallue [2]. In orrder to measuure the outpuut impedancee and also loaad voltage off a typical VCCS
V with
an oscillosccope, the inpu ut impedancee of the voltm
meter (oscillooscope) shouuld be considdered.

∙ (4)
–

June 2014 IJST,

I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

6 M. Khalighi et al.
l.

Fig. 7. A method
m for meaasuring VCCS
S output impeddance [2]

Figure 8A shows thhe results off other practiccal tests to determine

d thee best value for R1 to R4 according
to the circuuit of Fig. 6. In the curvee, Maximum m allowable load
l represennts the maxim mum value oof load in
which the V VCCS can suupport in a sppecific frequuency. The output
o wavefoform of VCC CS (load voltage) must
be exact sinnusoid and the
t relationshhip of the innput voltage of VCC parrt and load current
c mustt be linear
during the ttest. The amp plitude of looad current was
w adjusted on 1mA in all frequenciies. In fact, tthe output
Impedance of a currentt source depends on inveerse of frequuency [9]. The T maximum m allowable load in a
particular frrequency is also
a restricted by the valuue of output impedance. As a result according
a to the curve
in Fig. 8A. The best value for R1 too R4 was seleected as 1KΩ Ω. This is duue to the VCC structure with
w 1KΩ
resistor, whhich creates higher
h outputt impedance compared too others.

Fig. 8. (A) Thhe curve repreesents the rannges of maximmum allowablee load which can
c be supporrted by the immplemented
VCCS, in different frequeencies and diffferent values of R1 to R4 according
a t circuit in Fig. 6. (B) Open circuit
to the
output voltagge of implemeented VCCS in i different freequencies. (C) Output Impeedance of pro
oposed VCCS,, simulated
with PSPICE E in different frequencies
f
IJST, Transacctions of Electrrical Engineeriing, Volume 38
8, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 7

With tthis resistancce value the maximum allowable

a loaad in frequeencies less thhan 100KHzz is 12KΩ
which decreeases to abou ut 8KΩ in 2250KHz as sh hown in the upper curvee of Fig. 8A. The circuitt in Fig. 6
was tested with resistorrs less than 1KΩ (~0.5K KΩ), howevver, the waveeform was distorted
d andd the OP-
AMPs weree heated up. The T result off the other prractical test for
f the propo osed VCCS iss shown in FFig. 8B. In
this part thee output volttage of the VCCS
V was measured
m whhen there waasn’t any loaad at the outpput of the
circuit (openn circuit volttage), in opeerational freqquency rangee. Part C of Fig.
F 8 shows the output im mpedance
of simulatedd VCCS in PSPICE.
P
d) Pulse geenerator
A pulse generationn module hass been attach hed to the VCCS
V circuit design in order
o to gennerate two
types of synnchronous puulse train wiith the maxim mum point ofo positive peeaks and zerro points of the t output
current waaveform (peak detectionn and zero detection of the currrent).The pu ulses were used for
demodulatinng the bounddary voltagess in order to measure theem. Figure 9 shows the pulse p generattor circuit
diagram. Thhe input of thhe pulse geneerator is conn
nected to thee output of thhe filter. We may shift thhe position
of the pulsees with R2. As
A shown in Fig.
F 10, sectiions A and B depict the worthwhile
w u
usage of thiss part. The
pulse widthh can be variied by increaasing C1 from m 1nF to 10nF. For instaance, by reggulating R2, the t output
pulse train and also thee signal of Point B may be used for sampling annd extracting g real part of the load
voltage in thhe demodulaator section.

F 9. Pulse geeneration scheematic circuit

Fig.

Fig. 10. The outpuut generated puulses. (A) Peakk detection (B)) Zero detectioon

4 VOLTAG
4. GE MEASUR
REMENT

mit the overaall accuracy [18]. Errors arise from both

The data accquisition system can lim b current drive and
voltage acqquiring partss of the EIIT system. Related
R to the topologyy that is appplied for measuring
m
June 2014 IJST,
I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

8 M. Khalighi et al.
l.

(measuring pattern) in each

e momennt, only one voltage can be measured
d via two paarticular elecctrodes. In
addition to avoiding meeasurement error becausee of commonn mode effectts, the voltagge of electroddes which
are connecteed to the currrent source is
i not measurred.

a) Demodu
ulation of eleectrodes volt
ltage
In ordeer to measuree the boundaary voltage, at
a first the electrodes volttage must bee demodulateed. At this
stage, the vooltage demodulator sampples the voltaage waveform m in specificc points and prepares
p sam
mpled data
for measureement. As ann example, in i order to measure
m the real portionn of the loadd voltage, thee samples
should be ttaken from thet load volttage when th he injecting current wou uld be on thhe maximum m value of
positive peaak. On the mentioned
m pooints, the volltage of capaacitance commponents of the subject iss zero due
to the 90 degree
d phase shift. The method
m of modulation
m is based on sample-and-hold approaach that is
known as ppulse-sample demodulatioon. The sugggested method does not require mulltiplier or ouutput filter
[19].
As it iis shown in Fig. 11, both electrode voltages affter passing through the high pass filters f and
buffers are entered to thhe high speeed CMOS-loogic analog multiplexers
m s (74HC4053 3). The contrrol pulses
entered via ports C and D, come froom the pulsee generator part p of the cu urrent sourcee. To measurre the real
part of the lload voltagee, port C must be conneccted to the positive peak detector (po oint B of Figg. 11) and
port D mustt be connecteed to the maxximum pointt detector (too the output of o the pulse generator
g parrt, when it
works as thhe peak deteector). Hence the input voltages of AD625 (proogrammable gain instrum mentation
amplifier) are
a the demoodulated electrode voltages and the output voltage of the IC is i differentiaal value of
those mentiioned voltagees. The gain value determ mined by R7, R8 and R9 is optional an nd is calculatted by (5)
[20]. To meeasure the im maginary paart of the loaad voltage thhe pulse of point
p B and zero detectiion output
pulse shouldd be applied as well.
R7 = R8 Gainn = +1 (5)

Fig. 11. Proposed design for

fo demodulatiing of electrodde voltages

IJST, Transacctions of Electrrical Engineeriing, Volume 38

8, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 9

b) Voltage measuremeent

At thiss stage the deemodulated voltage

v is meeasured, hennce it must be converted from analogg to digital
value. The digital valuue is then trransmitted to t the microocontroller (ATmega128
( 8) of control unit. In
measuremennt part, AD11674 with 122-bit resoluttion is applieed as an AD DC. ATmegaa128 has a nnumber of
ADCs by itself,
i but itss ADCs cannnot be used d in measureement. Accoording to thee data-sheet [21], the
conversion time (i.e. req quired time for
f conversio on of an anaalog value to valid digitall data) of its ADCs, is
between 13 to 260μs whichw is not an exact am
mount of tim me. So it wouuld not be possible to caapture the
analog volttage signal at a specifiic point in the range of o micro seccond. However, in AD1674, the
conversion or sampling time is exacctly 10μs [22 2], hence it iss very interessting to captuure the analoog voltage
whenever itt is requiredd. This is due to the 12-bit resolutioon of the IC,, for amplitu ude between ±5V, the
mentioned ADC
A can sennse 2.44 mV changing on n the analog voltage (onee bit is used for
f sign).

5. MU
ULTIPLEX
XER

A high-speeed multiplexxer module with w 32 outpputs can be addeda to the proposed innstrumentatioon design.
As it is seeen in Fig. 122, the schemmatic circuit ofo the multipplexer moduule is illustraated. ADG5006AKN is
used as the analog multiiplexer in ouur design.
For a 32-electrode
3 EIT systemm, eight ICs of o the mentiioned multipplexer are reqquired. Four of which
are applied for injectionn and sink poorts of curren a used for voltage meaasurement.
nt source andd the others are
The multipllexer is appliied for sharinng the currennt source andd voltmeter between
b mulltiple electrodes. For a
32-electrodee system, 5 address
a liness are needed;; hence four address lines of ADG50 06AKN plus its enable
pin can be uused. Each laatch is applieed for a pair of
o ADG506A A (for lower and higher thhan 16 electrrodes).

Fig. 12.. The proposed

d design for multiplexer
m mo
odule

June 2014 IJST,

I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

10 M. Khalighi et al.
l.

6. CO
ONTROL UN
NIT

The conntrol unit con

ntains the following taskks:
 Controolling the muultiplexer paart to determiine the state oof each electtrode on eachh moment.
 Commmunicating with
w a PC andd matching with w MATLA AB for data trransaction viia RS-232.
 Conneecting to LCD D (if requireed), to displayy the voltagees and also number of meeasurements..
 Conneecting to AD DC and controolling that ass a slave withh handshakinng and interru upting signals.
 Commmunicating with
w AVR proogrammer foor programming via ISP mode. m

As it ccan be seen in Fig. 13, ATmega128

A is applied as
a the microocontroller off control uniit. All the
output portss should be buffered
b to aamplify the port
p output cuurrent and allso to prevennt the loadingg effect of
the next stagges.

Figg. 13. The propposed design oof control unit

nsaction via RS-232

a) Data tran

Serial communicattion is one of o the protoocols that aree supported by many ty ypes of commputer and
therefore coonnecting thee computer and
a microcoontroller, RS--232 commuunication pro otocol is ofteen applied
[23]. After completion of the voltaage measurem ment, the measured
m volttage stored in
i the microocontroller
register muust be transsmitted to the computter. In the proposed EIT E system,, the microocontroller
(ATmega1228) is connected to the computer viia MAX232,, which convverts the TT TL voltage to RS-232
voltage andd vice versa.. Hence the data transacction is donee by RS-232 mmunication via 9-pin
2 serial com
COM port.. In this deesign based on the stanndard frequeencies and baud-rates (bps) table [23], the
microcontrooller frequenncy is assigneed on 14.74556MHz. In thhe mentioned d frequency for all standdard baud-
rates such aas 2.4, 9.6 and 115.2 kbpss, the data traansaction errror is zero.

IJST, Transacctions of Electrrical Engineeriing, Volume 38

8, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 11

7 STIMULA
7. ATION AND
D MEASUR
REMENT PA
ATTERNS

To show thhe experimeental results of the prop posed EIT system,

s diffeerent topologgies can be used for
stimulation (current injeection) and voltage meaasurement, suuch as: adjaccent pattern,, opposite paattern and
cross patternn. In this papper the adjaccent pattern that
t is the moost common pair-drive prrotocol [24] is applied
as the stimuulation and measurement
m topologies.
In the adjacent (neeighboring) pattern,
p showwn in Fig. 144, the currennt is applied through twoo adjacent
electrodes aand the voltaages are measured from successive
s paairs of adjacent electrodees. The curreent is then
applied throough the nexxt pair of elecctrodes, and the voltage measuremen
m nts will be reppeated for otthers. The
procedure w will be continnued until eaach possible pair of adjaccent electroddes is used too inject curreent. In this
pattern, it iss common too use all N × (N − 3) meeasurements in most reco onstruction algorithms,
a w
where N is
the numberr of electrodes. ATmegaa128 in the control c unit has been prrogrammed based b on thee adjacent
pattern as thhe system stiimulation annd measurem ment protocols. In this patttern the currrent is injecteed mainly
in the outerr region of th he imaged object. The cuurrent densitty is highestt between thee injecting electrodes,
e
and decreasses rapidly ass a function ofo distance.

Fig.. 14. Adjacentt pattern as thee stimulation and

a measurem
ment protocolss of the propossed EIT system
m

8. EXPERIM
MENTAL RESULTS
R

To evaluatee performancce of the sysstem to recon nstruct images, some expperiments were
w carried out
o on the
cylindrical pphantom witth a few testt objects. Thhe images recconstructed by using thee implementeed system
will be show wn and desccribed, and soome elementts that can afffect the quaality of imagges will be inntroduced.
In all experriments, the saline
s used in
i phantoms is prepared withw solutionn of NaCl annd de-ionizedd distilled
water. A cuurrent signal (~1mA) witth 20 KHz frequency
fr is injected to th
he boundaryy of phantomms via two
electrodes. Adjacent pattern
p is used as the stimulation and voltagge measurem ment protoccol in all
experimentss. All imag ges are recoonstructed by b using thee Eidors MATLABM paackage [25, 26]. The
cylindrical pphantom is made
m of Plexxiglas with a radius of 15ccm.

a) Experim
ment 1

ment two piecces of a plasstic (Teflon) shaft have been

In the first experim b placed in
i front of ellectrode 1
of a 16-elecctrode cylind drical phantoom. The aimm of experimment is to evaaluate the quuality of recoonstructed
image of suubjects having less condductivity thaan saline. Thhe reconstru ucted imagess of this phaantom are
shown in Fig. 15. Partss B, E and F show the fine-model
fi im
mages of phaantom, and thet coarse m models are
illustrated inn parts C and
d D. In partss B, D, and F the image of
o phantom, isi reconstruccted onto noddes, but in
parts C and E the imagee is reconstruucted onto eleements.

June 2014 IJST,

I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

12 M. Khalighi et al.
l.

Fig. 15. (A) Phantom withh two pieces of plastic shaaft. (B) Fine model
m reconstrructed onto nodes.
n (C) Coaarse model
reconstructedd onto elemeents. (D) Coaarse model reconstructed onto nodes. (E) Fine mo odel reconstruucted onto
elements. (F)) Fine model reconstructed
r onto nodes

b) Experiment 2

In this experiment, two pieces of o a metallicc (Aluminumm) shaft have been placedd in front of electrodes
e
4 and 14 off a 16-electroode cylindriccal phantom with 30 cm diameter. Th he aim of thiis test is to understand
u
the quality of reconstru ucted image of o the subjeccts having ass much cond
ductivity as saline.
s Parts B to F of
Fig. 16 illusstrate the finnal reconstruucted images of the phanntom (shown in part A) byb using diffferent fine
and coarse mmesh generaation models.

Fig. 16. (A) Phantom withh two pieces ofo metallic shhaft. (B) Fine model
m reconsttructed onto nodes.
n (C) Coaarse model
reconstructedd onto elemeents. (D) Coaarse model reconstructed onto nodes. (E) Fine mo odel reconstruucted onto
elements. (F)) Fine model reconstructed
r onto nodes

c) Experim
ment 3
In this experiment, four piecess of the mettallic and plaastic shafts have
h been pu ut in the 16--electrode
cylindrical phantom. Thhe experiment is conduccted to evaluuate the quallity of reconnstructed imaage of the
subjects in different con
nductivities locating in composite seetting. The final
f reconsttructed imagges in fine
and coarse mmodels are shhown in partts B to F of Fig.
F 17.
IJST, Transacctions of Electrrical Engineeriing, Volume 38
8, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 13

Fig. 17. (A)) Phantom witth four piecess of metallic and

a plastic shaafts. (B) Fine model reconsstructed onto nodes. (C)
Coarse moddel reconstruccted onto ellements. (D) Coarse moddel reconstruucted onto nodes. n (E) Fiine model
reconstructedd onto elemennts. (F) Fine model
m reconstruucted onto noodes

ment 4
d) Experim
In the following, th
hree plastic and
a metallic subjects havve been placced in a 16-eelectrode thoorax-shape
phantom. The goal of thhis test is evaaluation of quality
q of thee images recoonstructed by
y using a 16--electrode
thorax-shappe phantom. AsA it can be seen in Fig. 18B, the im mage is reconstructed ontoo elements, but
b in part
C (of Fig. 18) the imagee is reconstruucted onto noodes.

Fig. 18. (A) A 16-electrodde thorax-shappe phantom with

w some subjjects. (B) Fine model recon
nstructed ontoo elements.
(C) Fine moddel reconstruccted onto nodees

e) Experim
ment 5
m of conduccting this expperiment is too show the effects
The aim e of hyp
per-parameteer (λ) in recoonstructed
images, andd evaluate the t quality of the final images reeconstructed with different values oof hyper-
I fact hyper-parameter (λ) is a scallar that contrrols the amount of regularization. Thhe goal of
parameter. In
hyper-parammeter selection is to prodduce the besst reconstrucction with high quality anda resolution. Hyper-
parameter sselection shoould producee solutions thhat preserve as much off the measureed data as possible
p to
obtain a usseful reconstruction [244]. Practicallly in image reconstructtion, insufficcient hyper-pparameter
causes the iimage to be dominated by b noise andd as it is incrreased, noise is filtered thhrough the smoothing
s
action. As it can be seeen in Fig. 19, few reconstructed im mages of thhe phantom with differeent hyper-
parameter vvalues are shhown. Obviously the reco onstructed immage with λ = 0.01, reprresents the best
b image
in terms of pprecision andd quality.

June 2014 IJST,

I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

14 M. Khalighi et al.
l.

me images whiich were reconnstructed with

Fig. 19. Som h different λ values
v (hyper--parameter) byy using the 166-electrode
EIT system. The image wiith λ=0.01 reppresents the beest image in teerms of precision and qualitty

f) Experim
ment 6
In the current experiment, efffect of num mber of voltage frame or numberr of scanninng in the
reconstructeed image quality will bee analyzed. In n fact, in adj
djacent patterrn the numbeer of voltagees in each
frame acquiired by scannning an N-electrode phaantom is calcculated by N×(N N – 3). Hence
H after scanning a
16-electrodee phantom eaach frame off voltage con ntains 208 vooltages.
In expeeriment 6 some metallic and plastic subjects
s are llocated in diifferent posittions of a 16--electrode
cylindrical phantom
p as shown in Figg. 20.A. Thee image in paart B is reconnstructed by using only one o frame
of voltage wwhich meanss that the imaage is reconsstructed afterr one time sccanning, therrefore some effects of
noise can bbe seen on this image (near electroodes 4, 8, 9, 9 and 10). However
H the image in part E is
reconstructeed after scan
nning ten tim
mes, which means image reconstructio
r on is done ussing the average of ten
frames of vvoltages. So the effects ofo noise are improved
i inn the mentionned image. TheT images in i parts C
and F of Figg. 20 are the 3D images of o the phantoom of part A,
A which are reconstructed by differennt number
of voltage frames.
f The image in parrt F reconstrructed after scanning
s ten times is smoother than the t image
of part C whhich is reconnstructed afteer scanning only
o one timee. This resultt can be extraacted so thatt, by using
the averagee of voltage frames insteead of one frame,
fr in imaage reconstrruction proceess, the quallity of the
final reconsstructed imagges would bee improved.

Fig. 20. Somme reconstruccted images off one phantom m, by using thhe average off n voltage fraames. (A) Phaantom with
four pieces of ( 3D image (3D surface plot)
o metallic and plastic shaffts. (B) n=1. (C) p of phanttom in part A with n=1.
(D) n=5. (E) n=10. (F) 3D image of phaantom in part A with n=10

IJST, Transacctions of Electrrical Engineeriing, Volume 38

8, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 15

g) Experim
ment 7
The obbjective of exxperiment 7 isi to show thhe effects of number
n of ellectrodes in the
t final recoonstructed
image qualiity. To evalu uate the imagge quality, thhree positionns of a phanttom surface would be coonsidered:
near the bouundary, in thhe central parrt and the whhole phantomm (both centrral part and near
n the bounndary). In
this test, thhree phantomms are appllied with different num mbers of elecctrodes whicch are 8, 16 and 32
electrodes. The dimensiions of phanntoms and innternal salinee and subjeccts are identiical. First poosition for
evaluation iis near the boundary.
b Ass it can be seen in Fig. 21,
2 the distaance betweenn subjects to the inner
wall of the phantoms iss about 2~3 cm, but in the t reconstruucted imagess by 8-electrrode and 16--electrode
phantoms, iin some placces show thaat the dark partsp (smudgges) are conn nected to the boundary of image,
whereas in reconstructed images byy 32-electrode phantom as a it can be seen, the darkk parts are coompletely
separated frrom the boun ndary of phanntoms.

Fig. 21. Recconstructed im

mages of subjeects which are located near tthe boundary of phantoms with
w various numbers
n of
electrodes to show their acccuracy in finaal image reconnstruction quaality

The neext position is the centrral part of the t phantom m. As is illusstrated in Fiig. 22, the quality
q of
reconstructeed image off subjects puut in the cenntral part is poor. This is i because thhe pattern which
w was
applied as the stimulattion and meeasurement topologies
t in all imagee reconstructtions in thiss paper is
Adjacent paattern, and iti is very sensitive to coonductive paarts near thee boundary anda insensitiive to the
central partts. It is senssitive to dissturbances in
n the bounddary shape of o the objectt, the positioon of the
electrodes, measuremen
m nt error and noise
n as well [24, 27].

Fig. 22. Shoowing qualityy of reconstructed images of

o the subjectt, which is loccated in centrral part of phaantoms, by
means of varrious numberss of electrodess

June 2014 IJST,

I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

16 M. Khalighi et al.
l.

The laast position is i the wholee phantom. As A shown inn Fig. 23, thhe reconstruucted image of the 8-
electrode phhantom is veery crude, annd it does nott have accepptable accuraacy to show the t correct pposition of
the subjectss which are loocated in thee phantom. ButB despite thhe bad qualitty and showiing inexact position
p of
the subjectss, there is a dark
d smudgee in the centrral part of thhis image, as the positionn of the centrral plastic
shaft. But itt cannot be seen
s in the reeconstructedd image of 322-electrode phantom.
p The reconstructed image
of the 16-electrode phaantom showss the correctt position off the subjectts located neear the bounndary, but
instead of thhe central plaastic shaft it shows a veryy light smuddge.

Fig. 23. Reeconstructed images

i of thrree phantoms with differennt numbers of electrodes. Subjects are located in
different possitions of the phantom (cenntral part and near the bounndary), and th
hese positionss were identiccal in three
reconstructioon processes

The reeconstructed image of thhe 32-electroode phantom m also exacttly shows thhe correct poosition of
subjects loccated near the boundary, but it does not
n show anyything, even a light smuddge, in its ceentral part
instead of pplastic shaft!
Hence the result thhat can be achheived from this experim ment is, for an EIT systemm that workss based on
the adjacennt pattern, whhen the num mber of electtrodes of phaantom increaases, the sysstem sensitivvity to the
parts near tthe boundaryy will be inccreased, but its sensitiviity to the cenntral part off the phantom
m will be
decreased sspecifically when some subjects haave been loccated near thhe boundaryy. The reasoon of this
phenomenoon is that in 8-electrode phantom
p thee distance beetween electrrodes is morre than the others
o (16
and 32-elecctrode phanto oms), and it causes moree current dennsity in the central
c part, which resultts in more
sensitivity. Therefore in image reconstructio
r on accuracyy, related to o measurem ment and sttimulation
topologies, the position of subjects is
i important as
a well, evenn when using g 32-electrodde system.

9. VISU
UAL COMPA
ARISON ON
N RESOLU
UTION

Although thhe EIT has been

b developped substantially over reccent years, thhere are manny challengess (such as
resolution) that still neeed to be oveercome to make
m it a clinnically appliccable imaginng [28]. EIT has been
extensively researched in i clinical diiagnosis [29, 30], but duue to poor Siignal to Noisse Ratio (SN NR) of the
boundary vooltage data and
a poor spattial resolutioon of image, the t EIT systeems have noot yet been acccepted as
regular meddical imaging g devices.
In thiss part, the quuality of recconstructed image
i usingg the proposeed EIT systeem will be compared
visually witth quality of images prodduced by usinng other impplemented EIIT systems. The T other EIT T systems
include an EIT system proposed by Bera and Nagaraju inn [31] and a DSP basedd multi-frequuency EIT
system desiigned by Goh harian et al in
i [32]. In faact in both syystems, imag ge is reconsttructed by ussing a 16-
electrode phhantom and from this pooint of view thoset system
ms are similarr to the propposed system m. The test
objects locaated in phantooms are madde from simillar materials (plastic and metal) as weell.
As a mmatter of fact, the size off all phantom ms and test objects are different
d in comparison
c w each
with
other, but thhe size and position
p of suubjects in theeir own phanntoms are exaactly plotted in the middlle column
of Fig. 24 aaccording to their paper information [31-32]. Thhe mapped im mage of phaantoms and their
t inner

IJST, Transacctions of Electrrical Engineeriing, Volume 38

8, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 17

objects (in middle

m colum
mn of Fig. 244), have beenn drawn withh the exact prroportion com mpared to diimensions
of real phanntoms and theeir containerrs applied in image reconnstruction. Ass it can be seeen in Fig. 244, the first
column shoows three 166-electrode phantoms
p rellated to diffferent EIT syystems contaaining salinee and test
objects. Thhe third coluumn shows the t reconstruucted images made by three t differeent EIT systtems. The
quality or reesolution of reconstructe
r d image seenn in Fig. 24.CC is obtainedd by using onnly 16 electroodes from
32 electroddes of the im mplemented EIT system.. Of course, the quality and resoluttion of imagge will be
increased iff the number of electrodes increases.

Fig. 24. Vissual comparisoon on quality of final imagges made by applying

a threee different impplemented EIT systems.
(A) reconstruucted image by
b using EIT system propo osed by Bera and Nagarajuu in [31]. (B) reconstructedd image by
using a DSP based multi-ffrequency EITT system propo osed by Gohaarian et al in [332]. (C) recon
nstructed imagge by using
our EIT systeem

9. CO
ONCLUSIO
ON

In this papeer a practicaal low-cost precise

p desig
gn of EIT instrumentatio on was propoosed and desscribed in
detail. Systeem performaance has beenn checked with
w a saline tank t and som
me experimeental results have
h been
shown. Thee novelty off this designn is its preciision which was achieveed without any a complexx circuits.
Simplicity oof design maade it possibble to have a low-cost innstrumentatio on which coould pave thee way for
EIT researcchers for desiigning and im mplementation of EIT haardware. Som me hard worrk has been ddone from
the practicaal point of vieew to improvve the precission of the prroposed systeem such as trrying to maxximize the
output impeedance of thhe current soource, whichh was finallyy increased to more thaan 5 MΩ suuch that it
wasn’t posssible to be measured
m w
with oscillosccope. Higherr quality of the reconstrructed imagees by our
system withh 16 electro odes comparred to the others
o is thee best reaso
on to verify the accuraccy of the
June 2014 IJST,
I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

18 M. Khalighi et al.

implemented EIT hardware. Future research includes improvement of the current EIT hardware in terms
of accuracy and precision to become appropriate for clinical applications and development of EIT
application in industrial and medical fields.

REFERENCES

1. Horesh, L. (2006). Some novel aproches in modelling and image reconstruction for multi-frequency electrical
impedance tomography of the human brain. PhD thesis.
2. Denyer, C. W. L. (1996). Electronics for real-time and three-dimensional electrical impedance tomography. PhD
thesis.
3. Yan, W., Hong, S. & Chaoshi, R. (2006). Optimum design of electrode structure and parameters in electrical
impedance tomography. Physiol. Meas., Vol. 27, pp. 291-306.
4. Boone, K. G. & Holder, D. S. (1996). Current approaches to analogue instrumentation design in electrical
impedance tomography. Physiol. Meas.17(suppl 1A) A229–A247.
5. Murphy, S. C., Chin, R. K. Y. & York, T. A. (2008). Design of an impeller-mounted electrode array for EIT
imaging. Meas. Sci. Technol., Vol. 19, pp. 1-12.
6. Goharian, M., Soleimani, M., Jegatheesan, A., Chin, K. & Moran, G. R. (2008). A DSP based multi-frequency
3D electrical impedance tomography system. Ann. Biomed. Eng., Vol. 36, pp. 1594-1603.
7. Yorkey, T. J. (1986). Comparing reconstruction methods for electrical impedance tomography. PhD thesis,
University of Wisconsin at Madison, Madison, WI 53706.
8. Filho, B., Brown, B. H. & Wilson, A. J. (2000). A comparison of modified Howland circuits as current
generators with current mirror type circuits. Physiol.Meas., Vol. 21.
9. Holder, D. S. (2005). Electrical impedance tomography: methods, history and applications (Series in Medical
Physics and Biomedical Engineering). Institute of Physics Publishing Ltd.
10. Khalighi, M., Vosoughi Vahdat, B., Mortazavi, M. & Soleimani, M. (2012). Practical design of low-cost
instrumentation for industrial Electrical Impedance Tomography (EIT). IEEE I2MTC, Graz.
11. Zhao, D. (2011). High output-impedance current source for electrical impedance tomography. IEEE 4th
International Conference on Biomedical Engineering and Informatics (BMEI).
12. Denyer, C., Lidgey, F. & Zhu, Q. (1994). A high output impedance current source. Physiol. Meas., Vol. 15, pp.
A79-A82.
13. Kugelstadt, T. (2009). Active filter design techniques (chapter 20), Excerpted from OP AMPs for everyone.
literature number: SLOD006A, published by Texas Instruments, third edition.
14. Cusick, G., Holder, D. S., Birkett, A. & Boone, K. (1994). A system for impedance imaging epilepsy in
ambulatory human subjects. Innov. Tech. Biol. Med.15(suppl 1A) A33–A39.
15. Jossinet, J., Tourtel, G. & Jarry, R. (1994). Performance and operation of a set of wideband current generators
for EIT. Innov. Tech. Biol. Med. 15(suppl 1A) A40–A46.
16. Zhang, Y. L., Zhui, X., Chaoshi, R., Wei, W., Dechun, Z. & Huiquan, Z. (2010). Study of voltage controlled
current source in electrical impedance tomography system. IEEE, iCBBE, 4th Int. Conference.
17. Shuai, Z., Guizhi, X., Huanli, W. & Duyan, G. (2006). Multi-frequency EIT hardware system based on DSP.
IEEE EMBS, 28th Annual international conference, China.
18. Smith, R. W. M. (1990). Design of a real-time impedance imaging system for medical applications”, PhD
Thesis, University of Sheffield.
19. Soleimani, M. (1998). Designing and making the electrical impedance tomography systems. Sharif University of
Technology, Msc thesis.
20. Data-sheet, (2000). AD625, programmable gain instrumentation amplifier. Analog Devices Inc., USA.
21. Data-sheet, (2006). ATmega128, 8-bit AVR microcontroller with 128K bytes programmable flash. Atmel Inc.

IJST, Transactions of Electrical Engineering, Volume 38, Number E1 June 2014

www.SID.ir
Archive of SID

Design
n and implemen
ntation of preciise hardware for…
fo 19

22. Data-sheeet, (2000). AD

D1674, 12-bitt 100 kSPS A/D/D converter. Analog
A Devicces Inc., USA.
23. Rahafroooz, A. (2009).. AVR Microcontrollers andd their applicaations. Nas puublishing.
24. Graham,, B. M. (20077). Enhancemeents in Electriical Impedancce Tomographhy (EIT) imagge reconstructtion for 3D
lung imaaging. Universsity of Ottawaa. PhD thesis, Degree in eleectrical and coomputing engiineering.
25. Vauhkonnen, M., Lionnheart, W. R. B., Heikkinenn, L. M., Vauhkonen, P. J. & Kaipio, J. P. (2001). A MATLAB
package for the EIDO wo dimensionnal EIT images. Physiol. Meeas., Vol. 22.
ORS project too reconstruct tw
26. Adler, A.
A & Lionheaart, W. R. B. (2006). Uses and abuses of o EIDORS: an extensible software basse for EIT.
Physiol. Meas., Vol. 27.
2
27. Dickin, F. & Wang, M. (1996). Electrical reesistance tom mography for process applications. IOP P Science,
Measureement Sciencee and Technoloogy, Vol. 7, No.
N 3, pp. 247--260.
28. Bayfordd, R. H. (20006). Bioimpeddance tomogrraphy (Electrrical impedannce tomographhy). Annual Review of
Biomediical Engineeriing, Vol. 8, ppp. 63-91.
29. Brown, B. H. (2001). Medical imppedance tomoggraphy and prrocess impedaance tomograpphy: a brief reeview. IOP
Science, Measuremennt Science andd Technology, Vol. 12, No. 8,
8 pp. 991–996.
30. Hoetink, A. E., Faes, T. J. C., Marrcus, J. T., Keerkkamp, H. J.. J. & Heethaaar, R. M. (20002). Imaging of thoracic
blood voolume changes during the heart
h cycle witth electrical im
mpedance usinng a linear spo
ot-electrode array.
a IEEE
Transaction on Mediccal Imaging, Vol.
V 21, Issue 6.
31. Bera, TT. K. & Naagaraju, J. (22011). Resistivity imaginng of a reconfigurable phantom p witth circular
inhomoggeneities in 2D
D-electrical im
mpedance tom mography. Elseevier, Measureement, Vol. 44 4, pp. 518-5266.
32. Gohariann, M., Soleimmani, M., Jegaatheesan, A., Chin,
C K. & Moran,
M G. R. (2008). A DSP P based multi-frequency
3D electtrical impedannce tomographhy system. Annals of Biomeedical Engineeering, Vol. 366, No. 9.
33. Khalighii, M., Vosoughi Vahdat, B. B & Mortazzavi, M. (2011). Designingg a practical high output impedance
current ssource for EIT RTET, Vol. 6, No. 2, DOI: 01.IJRTET.06
T systems. IJR 0 .02. 239.

APPENDIX A. SC
CHEMATICC CIRCUITS
S OF MENT
TIONED VC
CCSS
A
AND THE REESULTS OF
F PRIMARYY PRACTIC
CAL TESTS

Fig. A1. (A) VCCS basedd on Advancedd Howland [114, 15]. (B) VCCS
V based onn Double-Opeerational Ampplifier [16].
(C) VCCS bbased on Triple-Operationaal Amplifier fform 1 [17]. ((D) VCCS baased on Triplee-Operational Amplifier
form 2 [9]

June 2014 IJST,

I Transacti
tions of Electriccal Engineeringg, Volume 38, Number
N E1

www.SID.ir
Archive of SID

20 M. Khalighi et al.

Figure A1 shows different schematic circuits of the VCCS which have been used in practical tests. As it can be
seen in Table A1, the results of several experiments are shown briefly. These tests were done on the breadboard with
different resistance values in condition of 20KHz frequency and 1mA load current. It can be concluded (from Table
A1) that the VCCS based on TOA1 (fourth row) is the best choice to have an efficient voltage-controlled current
source. It can support a load in range of 10Ω to 10KΩ linearly and also has more than 5MΩ output impedance which
calculated with (4).

Table A1. The results of VCCS primary physical tests [33]

VCCS R1 R2 R3 R4 R5 Allowable Measured PSPICE

Type (KΩ) (KΩ) (KΩ) (KΩ) (KΩ) Load (Ω) (Ω) (Ω)
AH 1 2 1 1 1 10 ~ 1K 100K 420K
DOA 1 1 1 1 5 10 ~ 5K 3M 5.3M
TOA1 1 1 1 1 5 10 ~ 10K >5 M 7.8M
TOA1 2 2 2 2 5 10 ~ 10K 1M 3.9M
TOA1 5 5 5 5 5 10 ~ 10K 500K 1.6M
TOA1 10 10 10 10 5 10 ~ 10K 500K 770K
TOA1 22 22 22 22 5 10 ~ 10K 150K 340K
TOA1 27 27 27 27 5 10 ~ 10K 100K 280K
TOA2 1 1 1 1 5 10 ~ 10K 600K 2.5M

IJST, Transactions of Electrical Engineering, Volume 38, Number E1 June 2014

www.SID.ir