research article multiple-site hemodynamic analysis of ...research article multiple-site hemodynamic...

Research ArticleMultiple-Site Hemodynamic Analysis of DopplerUltrasound with an Adaptive Color Relation Classifier forArteriovenous Access Occlusion Evaluation

Jian-Xing Wu1 Yi-Chun Du2 Ming-Jui Wu3 Chien-Ming Li4

Chia-Hung Lin5 and Tainsong Chen1

1 Department of Biomedical Engineering National Cheng Kung University Tainan City 70101 Taiwan2Department of Electrical Engineering Southern Taiwan University of Science and Technology Tainan City 71005 Taiwan3Department of Internal Medicine Kaohsiung Veterans General Hospital Tainan Branch Tainan City 71051 Taiwan4Division of Infectious Diseases Department of Medicine of Chi Mei Medical Center Tainan City 71004 Taiwan5Department of Electrical Engineering Kao Yuan University Kaohsiung City 82151 Taiwan

Correspondence should be addressed to Tainsong Chen chentsmailnckuedutw

Received 28 February 2014 Accepted 24 March 2014 Published 30 April 2014

Academic Editors H-K Lam J Li and G Ouyang

Copyright copy 2014 Jian-Xing Wu et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

This study proposes multiple-site hemodynamic analysis of Doppler ultrasound with an adaptive color relation classifier forarteriovenous access occlusion evaluation in routine examinations The hemodynamic analysis is used to express the propertiesof blood flow through a vital access or a tube using dimensionless numbers An acoustic measurement is carried out to detectthe peak-systolic and peak-diastolic velocities of blood flow from the arterial anastomosis sites (A) to the venous anastomosissites (V) The ratio of the supracritical Reynolds (Resupra) number and the resistive (Res) index quantitates the degrees of stenosis(DOS) at multiple measurement sitesThen an adaptive color relation classifier is designed as a nonlinear estimate model to surveythe occlusion level in monthly examinations For 30 long-term follow-up patients the experimental results show the proposedscreening model efficiently evaluates access occlusion

1 Introduction

Chronic renal failure is an irreversible and progressive dis-ease In Taiwan more than 70 thousand people need toreceive hemodialysis treatment and this number is increas-ing year by year Arteriovenous access such as Brescia-Cimino arteriovenous fistulas (AVFs) or polytetrafluoroethy-lene grafts (AVGs) is a vital access for hemodialysis therapyMaintenance of proper function of intragraft blood flow(IBF) lt600mLmin for AVF and lt400ndash500mLmin forAVG is the most important issue for end-stage renal diseasepatients Due to repeated puncturing of this access everytwo days or long-term use access occlusion and failuresare caused by inadequate arterial inflow or venous outflowocclusionThis causes thrombosis resulting in intimal hyper-plasia chronic fibrin cellular deposits and aneurysm [1 2]Narrowing of the interior of the access increases the stress on

the vascular wall and produces vibration turbulent flow andhigh-pressure pulsatile flow

For early detection and homecare applications previousstudies have shown that the occlusion produces audible vas-cular murmurs resulting in increased new high-frequencycomponents in the power spectra [3ndash7] Stethoscope aus-cultation provides a noninvasive diagnostic technique todetect high-pitch murmurs with existing access occlusionHowever vascular murmurs are regarded as random andstochastic signals and are difficult to detect in the timedomain Frequency domain and time-frequency techniquessuch as the maximum entropy method (MEM) [4 5] Burgautoregressive method [6 7] and Fourier transform (FT)wavelet transform (WT) [6ndash10] are used to extract thespecific frequencies and the magnitude of the characteristicfrequencies Then the decision-making diagnosis systemswith artificial neural networks support vector machines and

Hindawi Publishing Corporatione Scientific World JournalVolume 2014 Article ID 203148 13 pageshttpdxdoiorg1011552014203148

2 The Scientific World Journal

fuzzy petri nets are carried out to perform the classificationtasks However phonography techniques are affected by themeasurement sites the sizes of the sampling window andthe parametric mathematical models for phonoangiography(PCG) signal analysis In addition Doppler sonographyintravascular ultrasound image colour duplex ultrasoundand X-ray angiography [11ndash13] are also used to detect theocclusion in these vascular accesses in clinical examinationsStationary instruments provide reliable techniques and highaccuracy in clinical environments However they need toappreciate the operational principles and require extra learn-ing due to the limitations of the patients themselves

The ultrasound flowmeter is also a reliable and highlyaccurate technique and provides a good resolution in detect-ing blood properties [14ndash16] However the Doppler powerand velocity across an access under a pulsatile blood flow andthe ultrasonic backscattering signal from pulsatile flowingblood are affected by the shear rate as well as the accelerationof the flow velocity and the turbulences To improve ultra-sonic signal quality and the SNR at a specific depth Dopplerultrasound with a bipolar pulse drive can achieve the bestresolutions in high- and low-frequency ranges of 20MHz to50MHz and 8MHz to 20MHz respectively which is enoughto characterize the properties of the blood flow throughsurface and deep vessels It is well developed compact fastand free of irradiation [15ndash17] The purpose of this study isto use the Doppler ultrasound device for measuring bloodflow such as peak-systolic velocity peak-diastolic velocityend-diastolic velocity and heart rate Then the dimension-less numbers supracritical Reynolds (Resupra) number andresistive index (Res) are computed using flow velocitiesand the quantitative degrees are applied to screen the DOSColor relational analysis (CRA) [18] as a classifier performsthe recognition task for assessment of arteriovenous accessocclusion using a hue-saturation-value color model [19] Ithas a flexible pattern mechanism with add-in and delete-offtraining data and does not demand strict statistical methodsand inference rules in real time screening applications [18]An adaptive modeling system with an adjusting recognitioncoefficient to enhance the patency rate using a particle swarmoptimization (PSO) algorithm was utilized to increase theaccuracy [20 21] For 30 follow-up patients the results showthe proposed screening system is more efficient for occlusionevaluation

The rest of this paper is organized as follows Section 2addresses themeasurement technological support and exper-imental setup and Section 3 describes the adaptive color rela-tion classifier In Sections 4 and 5 experimental results andconclusion are provided and the efficiency of the proposedscreening method is demonstrated

2 Measurement Technological Support andExperimental Setup

21 Doppler Ultrasound Measurement In clinical examina-tions auscultation and ultrasound examinations are usedto detect the arteriovenous shunt (AVS) occlusion Ultra-sound examinations such as Doppler ultrasound imaging

intravascular ultrasound imaging and ultrasound flowmeter[22 23] are reliable and highly accurate clinical techniquesIt can operate in A-scan and B-scan instrument using low-frequency and high-frequency sound waves which can beprocessed to create quantitative images formarking the vesselaxis and diameter of the 2D ultrasound image [24 25] andalso provides a good resolution in detecting blood propertiessuch as flow velocities aggregation and coagulation [14ndash16] as shown in Figure 1 In a follow-up examination itcan also be used to determine whether an abnormality isstable or a treatment is working Among these the ultrasoundflowmeter is noninvasive well developed compact andfree of irradiation measurement technique The purposeof this study is to use the Doppler ultrasound device formeasurements of physiologic blood flow in AVS access suchas Doppler ultrasound velocity flow peak-systolic velocitypeak-diastolic velocity end-diastolic velocity and heart rate

Figure 1 shows the Doppler angles correct with flowvelocities that can be estimated from the Doppler shift Themaximum frequency 119891max and mode frequency of Dopplerspectra are less changed along with the change in the samplevolume lengthwhen setting the sample volumeposition at thecenter of the blood vesselThe flow velocity within the samplevolume is measured as [26]

119881 =

119891119889119888

21198910 cos 120579

(1)

where 119888 is the speed of sound in the blood the nominalvelocity of ultrasound in tissue is approximately 1540msec119891119889 is the Doppler shift frequency 1198910 is the center frequencyof transmitted ultrasound and 120579 is the Doppler angle 50sim80degrees between the acoustic beam and blood flow

In a narrowed access turbulence occurswhen blood flowsthrough a stenosis cross-sectional area increasing blood vis-cosity and flow resistance In previous studies [27] Dopplershift usually encounters a velocity range of 010msecsim

140msec in clinical measurement and represents a fre-quency shift depending on the transducer frequency Thesingle transducer was excited by a pulse repetition frequency(PRF) using bipolar pulses [25 27] as shown in Figure 2Then a 12 bit high-resolution digitizer with 200MHz sam-pling frequency was used to record backscattering and orig-inal signals by PRF trigger synchronize control with FPGAcontrol (DE2-70 Altera San Jose CA USA) It has goodresolution within a focusing zone of 100mm times 100mmand attenuation from 250 dB to 200 dB Thus the Dopplerfrequency shift can be calculated by those two signals inan embedded system One advantage of the bipolar pulsesdesignwith FPGA control is that it drives a higher voltage andshorter pulse width than the traditional ultrasound systemThose features can provide better SNR of the signal and aresuitable to detect blood changes in a hemodialysis access withthe bipolar pulse control in the ranges of 75MHz to 10MHzThe bipolar pulse design is also a low-cost design that can beassembled easily into a handheld system

22 Hemodynamic Analysis A flow with periodic variationsis a so-called pulsatile flow in small blood vessels or large

The Scientific World Journal 3

Arterial anastomosis site (A site)Venous anastomosis site (V site)

Loop site (L site)

(a)

Venous transducer

DBlood flow

Turbulent flow

120579

Skin Fixed

120579

Fixed

Low-frequencytransducer

High-frequency

Doppler ultrasound velocity flow (DUVF)

Peak-systolicvelocity

Peak-diastolicvelocity

Arterialanastomosissite (A site)

anastomosissite (V site)Loop site (L site)

P-P intervalVp

Vm

(b)

Figure 1 (a) Arteriovenous shunt and itsmeasurement sites (b)Doppler ultrasoundmeasurements for ultrasound images and flow velocities

arteries In physical phenomena the characteristics of theDoppler ultrasound velocity flow (DUFV) waveform can beused to analyze the pressure and propagation velocity of thephysiologic conditions of the blood flow According to theinfluence of dimensionless numbers such as the Reynolds(Re) number Womersley (120572) number and Strouhal (St)number they can be quantified to describe the instability flowfield pulsatile flow and oscillating flowmechanismsThey aredefined as follows [28ndash31]

(i) Reynolds (Re) number is the ratio of inertial forcesviscous forces Turbulent flow will occur at a highReynolds numbers and is dominated by inertialforces which tend to produce chaotic eddies vorticesand instability flows

(ii) Womersley (120572) number is the ratio of unsteady forcesviscous forces and is a dimensionless expression ofthe pulsatile flow frequency in relation to viscous

effectsThe vessel diameter decreases as the 120572 numberdecreases in a stenotic cross-section

(iii) Strouhal (St) number is the ratio of oscillatory inertialforcesconvective inertial forces and is a dimension-less number describing oscillating flow mechanismsIts value also represents the dimensionless strokevolume

The resistive (Res) index is also a parameter of pulsatileblood flow reflecting both vascular compliance (changein flow volume with a change in pressure) and vascularresistance in in vivo or in vitro studies The Res index isdefined as [26]

Re 119904 =

119881119901 minus 119881119898

119881119901

(2)

where 119881119901 is the peak-systolic velocity (PSV) and 119881119898 is thepeak-diastolic velocity (PDV) through the Doppler ultra-sound measurements In the literatures [32] the cylindrical


Dataacquisition

FPGA systemBipolarsystem

Blood fow

Expander

Limiter

θ

Timercountersystem

AD converter

Band passfilter

Preamplifier

75MHzsim10MHztransducer

Dopplerangle

(a)

Width (mm) 0 10 30

Dep

th (m

m)

(dB)

Bipolar pulse

Beam profilep

5-cycle

100V

25

20

15

10

5

20

15

10

5

0

(b)

Figure 2 (a) The diagram of the proposed ultrasound device (b) a 5-cycle 75MHz bipolar pulse with amplitude 100Vpp

models with different diameters viscosity stroke displace-ment and frequency were used to analyze flow instabilitiesThe critical peak Reynolds (Repeak) number was derived at apower law to indicate the flow instabilities for in vitro studiesIt is calculated based on both the Womersley (120572) numbersand Strouhal (St) numbers and indicates the transition flowto turbulence flow which can be defined as [30]

Repeak = 169120572083Stminus027 (3)

The supracritical Reynolds number Resupra = |Re minusRepeak|is correlated with body weight femalemale vessel diame-ter pulsatility index and cardiac output Therefore Resupradimensionless can be used to define critical values alongthe cross-sectional AVS access in an in vivo study Inaddition the cross-section for every analysis plane is thenoncircular duct in a blood vessel The height and width arecomparable and the dimension for internal flow is taken to

be the hydraulic diameter 119863119867 which can be considered as[31]

119863119867 =

4119860

119875

119863119867 = 119863 for a circular duct

119875 =

119873119897

sum

119897=0

119871 119897 119897 = 1 2 3 119873119897

(4)

where 119875 is the wetted perimeter 119871 119897 is the length of eachsurface in contact with the aqueous body and the spline inter-polation is used for noncircular surface interpolation with 119873119897

fractions In noncircular ducts the hydraulic diameter 119863119867needs to be substituted for the diameter of a circular ductin the Reynolds number Womersley number and Strouhalnumber

23 Experimental Setup We recruited patients who con-sented to undergo follow-up examinations This study wasapproved by the Institutional Review Board (IRB) of Kaoh-siung Veterans General Hospital Tainan Branch under


contract number VGHKS13-CT12-11 A total of 30 patientswere enrolled The participants comprised 9 females and 21males aged between 42 and 89 with amean age of 694 plusmn 1313years and all participants had anAVFThe participantsrsquomeandialysis duration was 1sim16 years Arteriovenous shunt (AVS)is a pathological physiology created on a patientrsquos forearmand upper arm to facilitate hemodialysis treatment Due torepeated puncturing of the AVS accesses and long-term usethe interior of the accesses can exhibit pathologic changesresulting in the formation of a thrombus intimal hyperplasiaand changes in the aneurysmal deformability of the access

Maintenance of proper function is the most importantissue for hemodialysis patients Therefore it is necessary toevaluate AVS functions in routine screening by confirmingthe specific degrees of stenosis (DOS) from X-ray imagesangiographic images and ultrasound images In clinicalresearch vessel stenosis is defined as DOS gt 050 reductionin luminal diameter judged by comparison with either theadjacent vessel or graft and can be defined as [7 10]

DOS = 1 minus (

119889119867

119863119867

)

2

(5)

where 119863119867 is the hydraulic diameter of the normal graftor vessel in the direction of the blood flow and 119889119867 is thehydraulic diameter of the stenosis lesion

Preliminary screening results by physicians confirmedthe specific degrees including Class I DOS lt 030 ClassII 030 lt DOS lt 050 and Class III DOS gt 050 whereDOS = 100 is total occlusion Then the proposed Dopplerultrasound device was carried out to measure the multisitesrsquoblood velocity as following the direction of blood flow froman arterial anastomosis site (A site) to a venous anastomosissite (V site) as shown in Figure 1

3 Adaptive Color Relation Classifier

31 Quantitative Analysis with the Dimensionless NumbersMultisite measurements were used to detect the blood veloc-ity Two dimensionless numbers were calculated and used toquantify the DOS which can be expressed as follows

(i) Ratios of Re119904119906119901119903119886 Divide an AVS access into two segmentsand suppose the same volume of blood through those theratios of Resupra at the A site loop site (L site) and V site aredefined as

Ratio(A) =

100381610038161003816100381610038161003816100381610038161003816

Re(A) minus Resupra(A)Re(L) minus Resupra(L)

100381610038161003816100381610038161003816100381610038161003816

=

Resupra(A)Resupra(L)

Ratio(L) =

Resupra(L)Resupra(L)

= 1

Ratio(V) =

100381610038161003816100381610038161003816100381610038161003816

Re(V) minus Resupra(V)Resupra(L)

100381610038161003816100381610038161003816100381610038161003816

=


(6)

where each sitersquos quantity is the value of the quantity inthe loop site the so-called the value of per unit Thus theone advantage can be checked rapidly for gross changes at

each measurement site Another advantage of quantities isthat they can be correlated with body weight femalemalevessel diameter and hemodynamic factors for any patientAVFAVG and cross-sectional analysis

(ii) Resistive (Res) Index This index is an indicator of vesselresistance its value increasing as the end-diastolic velocitydecreases The normal range for the common vessel wave-form is shown in 050sim065 Higher values indicate diseaseor stenosis [26 27]

For 30 subjects Table 1 shows the results of velocitymeasurements at the arterial anastomosis site (A) loop site(L) and venous anastomosis site (V) respectively Becausethe AVS is a pulsatile access the blood flow will mix high-velocity blood passing the lesion vessel and low-velocityblood downstream from the lesion vesselThe ratios of Resupraand resistive (Res) indexes can be calculated with the keyvariables 119881119901 (PSV) 119881119898 (PDV) hear rate and hydraulicdiameter The Res indexes increase as the ratios of the119881119901119881119898 (gt30) increase at each measurement site as shown inFigures 3(a) and 3(b) The ratios of Resupra indicate the valuesapproach to 1 for the same Resupra through an AVS accessas the values gradually decrease for the degree 03 lt DOS lt

05 and for values greater than 1 for the degree DOS gt 05as shown in Figure 3(b) The first three ratios are used toidentify the laminar flow and flow instabilities from inflowto outflow and the last three indexes are used to identifythe intravascular resistances Six parameters have specificranges and can combine trend patterns to separate threedegrees According to these trend patterns of 30 subjects wecan systematically create training data for the adaptive colorrelation classifier

32 Adaptive Color RelationClassifier Training Classificationis the task of recognizing patterns into a few classes orcategories Based on similarity and dissimilarity a mecha-nism with a learning algorithm can be designed to recognizepatterns or features in diagnosis or screening applications[18] In this study a color relation analysis (CRA) method isproposed to develop a classifier which possesses a flexiblepattern mechanism with add-in and delete-off trainingdata and it does not require strict statistical methods orinference rules as shown in Figure 4 Assuming Φ119903 =

[Ratio(A)(0) Ratio(L)(0) Ratio(V)(0) Res(A)(0) Res(L)(0)

Res(V)(0)] = [1206011(0) 1206012(0) 1206013(0) 1206014(0) 1206015(0) 1206016(0)] is areference pattern where 119894 = 1 2 3 6 (119899 = 6) andcomparative patterns Φ119888(119896) = [1206011(119896) 1206012(119896) 1206013(119896) 1206014(119896)

1206015(119896) 1206016(119896)] are comparative patterns where 119896 =

1 2 3 119870 as described in training data and representedbelow [18]

[[[[[[[[[

[

Φ119888 (1)

Φ119888 (2)

Φ119888 (119896)

Φ119888 (119870)

]]]]]]]]]

]

=

[[[[[[[[[

[

1206011 (1) 1206012 (1) sdot sdot sdot 120601119894 (1) sdot sdot sdot 1206016 (1)


d

d


d

d


]]]]]]]]]

]

(7)


Table 1 The results of velocity measurements

DOSaverage velocity Measurement siteArterial anastomosis site (A) Loop site (L) Venous anastomosis site (V)

lt030 (12)119881119901(cmsec) 10320 plusmn 1835 6460 plusmn 2805 8256 plusmn 2530119881119898 (cmsec) 3700 plusmn 1080 2883 plusmn 1661 3543 plusmn 1737Δ119875 (mmHg) 426 plusmn 013 167 plusmn 031 272 plusmn 026

030sim050 (11)119881119901 (cmsec) 10670 plusmn 2880 5491 plusmn 2154 7779 plusmn 1854119881119898 (cmsec) 3549 plusmn 1996 1661 plusmn 869 2987 plusmn 1581Δ119875 (mmHg) 455 plusmn 033 121 plusmn 019 242 plusmn 014

gt050 (7)119881119901 (cmsec) 7221 plusmn 4058 2561 plusmn 962 6603 plusmn 4212119881119898(cmsec) 1784 plusmn 87 786 plusmn 384 1492 plusmn 1419

Δ119875 (mmHg) 209 plusmn 066 026 plusmn 004 174 plusmn 071Note the pressure drop Δ119875 (mmHg) can be related to the velocity119881119901 (msec) and is defined by Δ119875 = 4(119881119901)

2 [26]

0

1

2

3

4

5

A site L site V site

Rat

ioV

pV

m

DOS gt 0503 lt DOS lt 05

DOS lt 03

(a)

0

02

04

06

08

1

12

14

Tren

d pa

ttern


DOS lt 03

Ratio(A) Ratio(L) Ratio(M) Res(L)Res(A) Res(V)

(b)

Figure 3 (a) The trends of ratios 119881119901119881119898 (b) the trend patterns of six dimensionless numbers

where119870 is the number of the comparative pattern Comparedwith the reference pattern Φ119903 and 119896th comparative patternΦ119888(119896) the Euclidean distance ED(119896) can be expressedas a similarity degree between reference pattern Φ119903 andcomparative pattern Φ119888(119896) as

ED (119896) = radic

6

sum

119894=1

(Δ120601119894 (119896))2 Δ120601119894 (119896) =

1003816100381610038161003816120601119894 (0) minus 120601119894 (119896)

1003816100381610038161003816 (8)

If pattern Φ119903 is similar to any pattern Φ119888(119896) the ED(119896) willbe a small value and EDmin = minED(119896) le ED(119896) le

EDmax = maxED(119896) These indices can be used for patternrelation analysis The overall indices ED(119896) are converted togray grade 120588(119896) by nonlinear transformation as [18]

120588 (119896) = 120585 exp [minus120585ED (119896)] (9)

where 120585 is the recognition coefficient with interval (0 infin)Intensity adjustment is used to enhance the contrast bymapping the original intensity value to a new specific rangeFor AVS stenosis evaluation these average gray grades can bestratified as three groups Class I (normal) Class II and ClassIII represented as

120588Iave =

sum119873119868

119905=1120588I(119905)

119873I

120588IIave =

sum119873II119905=1

120588II

(119905)

119873II

120588IIIave =

sum119873III119905=1

120588III

(119905)

119873III

(10)


Template 1

Compare

Template 2

Compare

Template K

Compare

ED(1)

ED(2)

ED(K)

120588(1)

120588(2)

Template 10

Compare ED(10) 120588(10)

Template k

Compare ED(k) 120588(k)

120588(K)

Euclidean distance

Gray grade

DOS gt 05

03 lt DOD lt 05

DOS lt 03

++

minus

Parameterestimation

Error

r

r

b

g

g

Primary color grade

Ratio(A)

Ratio(L)

Ratio(V)

Res(L)

Res(A)

Res(V)

Hue anglered

Saturationvalue S

Hue angleblue

Hue anglegreen

Class III

Class I

Class II

Desiredangle

Refe

renc

e pa

ttern

Tran

sfer

RG

B co

lor s

pace

to H

SV c

olor

spac

eFigure 4 Structure of the adaptive color relation classifier

where 119873I 119873II 119873III are the number of the respective groupsMinimum and maximum average grades are obtained asfollows

120588mim = min [120588Iave 120588

IIave 120588

IIIave]

120588max = max [120588Iave 120588

IIave 120588

IIIave]

(11)

where 120588min = 120588max According to the hue-saturation-value(HSV) color model [18 19] CRA is a mathematical model bytransformations between the RGB primary color space andthe HSV color space Referring to the hue angle 119867 isin [0 360]

as

119867 =

[60∘

times (

119892 minus 119887

Δ120588

) + 360∘] mod 360

∘ if 120588max = 120588

IIIave

60∘

times (

119887 minus 119903

Δ120588

) + 120∘ if 120588max = 120588

Iave

60∘

times (

119903 minus 119892

Δ120588

) + 240∘ if 120588max = 120588

IIave

(12)

where Δ120588 = 120588max minus 120588min and primary color grades 119903 (red) 119892

(green) and 119887 (blue) are respectively formulated as

119903 =

120588max minus 120588IIIave

Δ120588

119892 =

120588max minus 120588Iave

Δ120588

119887 =

120588max minus 120588IIave

Δ120588

(13)

The saturation 119878 and value 120574 are defined as

119878 =

120574 minus 120588min120574

120588min = 120588max

120574 = 120588max 120588max = 0

(14)

The value of 119867 is generally normalized to lie between 0∘ and360∘ where the centers of color distribution are 120∘ 240∘and 360∘ for three degrees respectively and hue 119867 has nogeometric meaning when 120588min = 120588max and saturation 119878is zero The parameter 119867119862 is utilized to identify the threeclasses as

119867119862 = [

119867

360∘] 119867119862 isin [0 1] (15)


where the critical decision 119867119862 = [Class I Class II Class III]= [23 13 1] The concept of the proposed CRA is derivedfrom the HSV color model to describe perceptual colorrelationships for AVS stenosis estimation The saturation 119878varies from 00 to 10 and the corresponding colors varyfrom unsaturated (value 0) to fully saturated (value 1) Inthis study parameter 119867119862 is used to identify the possibledegree however the choice of recognition coefficient 120585 isa limitation As its value gradually increased 120585 ≫ 1 graygrades were distinctly separated into three degrees and thedecision-making might become increasingly nonlinear for ahigh-dimensional classification space which will affect thescreening accuracy

Therefore the CRA classifier with an optimal recognitioncoefficient 120585 can enhance the accuracy Consider the meansquared error function (MSEF) the objective is intended tominimize the MSEF as

min (MSEF) = min(

1

119870

119870

sum

119896=1

[119879 (119896) minus 119867119862 (119896)]2) le 120576 (16)

where 119879(119896) is the desired degree for the 119896th training dataHowever 119867119862 is a nonlinear function and its partial differ-ential equation is difficult to obtain using the conventionalgradient method

This study proposes the adaptive color relation classifierwith adaptability to adjust the recognition coefficient toenhance screening accuracy with add-in and delete-off train-ing data For an adaptive classifier design the particle swarmoptimization (PSO) algorithm is an evolutionary methodto solve optimization problems for time-varyingdynamicsearch spaces The property of a PSO algorithm searchesusing multiple particles that modify their search positionsaround amultidimensional search space until the unchangedpositions have been achieved Each position is adjustedbased on the past action experiences and the dynamicallyaltering the velocity of each particle [20 21] Let 120585119892119901 bethe current position of the 119892th agent at iteration number119901 agent 119892 = 1 2 3 119866 where 119866 is the population sizeMultiple particles form a population and are represented bya 119866-dimensional vector as 120585

119901= [120585119901

1 120585119901

2 120585

119901

119892 120585

119901

119866] The

modification of position 120585119901

119892 can be represented by velocity

Δ120585119901

119892 Δ120585119901

119892= [Δ120585

119901

1 Δ120585119901

2 Δ120585

119901

119892 Δ120585

119901

119866] The mathematical

representation is given by [20] the followingVelocity

Δ120585119901+1

119892= 120596Δ120585

119901

119892+ 1198881rand1 (120585best119892 minus 120585

119901

119892)

+ 1198882rand2 (120585best minus 120585119901

119892)

1198881 = (1198871 minus 1198861)

119901

119901max+ 1198861 1198882 = (1198872 minus 1198862)

119901

119901max+ 1198862

(17)

Position

120585119901+1

119892= 120585119901

119892+ Δ120585119901+1

119892 (18)

where 120585best is the global best in the group 120585best119892 is theindividual best 120596 is the inertia weight rand1 and rand2 are

Start

Evaluation of searching point

using objective function

as equation (16)

Modification of searching point using equations (17) and (18)

Reach maximum iteration or

Reach convergent condition

Stop

Yes

No

Step 1

Step 2

Step 3

Step 4

Generate an initial random parameter 120585pand population size G

120585p = [1205851p 120585

2p 120585

gp 120585

Gp ]

Figure 5 The flow chart of PSO algorithm

the uniformly random numbers between 0 and 1 and 1198881

and 1198882 are the acceleration parameters where the first termis the ldquocognitive componentrdquo and the second term is theldquosocial componentrdquo Parameters 1198861 1198871 1198862 and 1198872 are constantvalues of which the experienced values are 1198881 from 25 to05 and 1198882 from 05 to 25 respectively [21] and 119901max isthe maximum number of allowable iterations Therefore aPSO algorithm with time-varying acceleration coefficients(TVAC) [20] could efficiently converge to the global optimalsolution The flow chart of the PSO algorithm is shown inFigure 5

4 Experimental Results and Discussion

Following the direction of blood flow from the arterial anas-tomosis site (A) to the venous anastomosis site (V) multisitemeasurements were used to detect the blood velocities usingthe Doppler ultrasound device A 750MHz transducer wasexcited by sinusoidal tone bursts with 5-cycle long pulses atpulse repetition frequencies (PRFs) of 1948 kHzsim2727 kHzFor the pulsatile flow experiments the stroke rate was setat 75sim30 beatsmin at a peak flow velocity of 100msecsim

140msec using a PRF trigger with FPGA control [2223] The relevant hemodynamic parameters and transducerparameters for the experimental setup are shown in Table 2A 12 bit high-resolution digitizer with 200MHz sampling


Table 2 The related parameters of experimental setup

Hemodynamic parametersBlood density 1055 kgm3

Hematocrit (Ht) 40 for a normal adult120583plasma = 00035 pas

Dynamic viscosity 001063Nsm2

Room temperature = 26∘CHeart rate 100sim125HzHydraulic diameter Ultrasound image examination for Follow-up examination

Specification of transducerCenter operation frequency 1198910 750MHzOutput voltage plusmn50V (119881pp = 100V)

Pulse repetition frequency PRF 1948 kHzsim2727 kHzVelocity range 10mssim14ms

Transmitted pulse duration 066 120583sPulse cycle 5 cyclesDoppler angle 45sim60 degrees

frequency was used to record backscattering and originalsignals by PRF trigger synchronize controlThen one can cal-culate the Doppler frequency shift by those two signals in thePC system The velocities ratios of superacritical Reynolds(Resupra) numbers and resistive (Res) indexes were calculatedusing (2) and (3)The proposed adaptive CRA based classifierwas developed on a PC AMD Athlon II times2 245 291 GHzwith 175GBRAM and Matlab software (Mathworks NatickMA) To demonstrate the effectiveness of the proposedmethod for AVS stenosis evaluation 40 subjects were chosen(IRB VGHKS13-CT12-11) to be divided into 30 training dataand 10 testing data Preliminary diagnosis results confirmedthe specific degree by ultrasonic image examination andobservation by clinical physicians as shown in Figure 6Feasibility tests and case studies were used to validate theproposed estimation model as detailed below

41 Feasibility Tests with the Adaptive CRA The 30 subjectswere divided into three groups as training data The overalltraining results are shown in Figure 7 The performance ofthe CRA classifier is affected by the nonlinear gray gradetransformation as seen in (9) As the recognition coefficient120585 gradually increased gray grades were distinctly separatedinto three groups The decision might become increasinglynonlinear for a high-dimensional classification space There-fore recognition coefficient 120585 ≫ 0 was selected and thehue angle 119867119862 had high confidence to transform betweenthe RGB primary color space and the HSV color space Thesaturation value 119878 also indicated the confidence indices andconfirmed the values approached to 1

According to the training data the CRA classifier has6 inputs and 3 outputs as shown in Figure 4 The firststage is calculation of the Euclidean distances (EDs) betweenreference patterns and comparative patterns The most sim-ilar choice is the smallest ED Then the overall EDs areconverted to gray grades and are grouped into three classesThen the hue angle 119867119862 is carried out to identify the

Figure 6 Ultrasonic image examination date May 15 2013 Kaoh-siung Veterans General Hospital Tainan Branch

possible class In the second stage updating the recognitioncoefficient is performed by the proposed PSO algorithmFor the adaptive application the PSO with time-varyingacceleration coefficients (TVAC) is given by population size119866 = 20 for each iteration acceleration coefficients 1198861 =25 and 1198871 = 05 for the cognitive component accelerationcoefficients 1198862 = 05 and 1198872 = 25 for the social component[20 21] and the maximum allowable number 119901max = 100With the convergent condition MESF lt 120576 = 005 it can beseen that the adaptive CRA can find the optimal parameter120585best = 79883 and minimize the mean squared error asshown in Figures 7(a) and 7(b) The training results with theadaptive color relation classifier are shown in Figure 7(c)

In comparison Table 3 shows comparative performancesusing the proposed method and multilayer networks suchas artificial neural networks (ANNs) and support vectormachines (SVMs) [4 5 8] Stethoscope auscultation wasused to measure the phonoangiography (PCG) signalsThen feature selection takes the specific frequencies andthe magnitude of the characteristic frequencies using thetime-frequency methods and Burg methods [6 7 9 10] Itprovides time-frequency resolution to extract features from


Table 3 Comparison of performances between the proposed method and artificial neural network (ANN)

Task MethodCRA ANN and SVM

Network structure Hybrid decision-making and network structure Multilayer network

Training data Yes YesMinor Major

Feature selection Hemodynamic analysis with dimensionless numbers Time-frequency analysis and Burg method [6 7 9 10]

Inferencelearning algorithm PSO algorithm [20 21] (i) Least mean square algorithm(ii) Gradient descent method

Parameter assignment Yes YesMinor Major

Adjustable parameter Yes YesMinor Major

Iteration training Yes YesConvergent condition Yes Yes

0 5 10 15 20 25 30 35 40 45 500

01

02

03

04

Number of iterations

Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)

Figure 7 (a) The mean squared error of optimal solution versus the number of iteration training (b) the optimal recognition coefficientversus the number of iteration training and (c) the training results with the adaptive color relation classifier

nonstationary PCG signals However significant frequen-cies need trial and error procedures using the cascadedlow-passhigh-pass filters and down-samplingup-samplingoperations Its technique is limited by the preprocessingtime computational time and significant feature selectionIn addition multilayer network classifier is a mechanism

that has been widely applied in the continuous modelingsystem with iteration training such as tuning networkparameters and automatic target adjustment Briefly thelearning performance heavily relies on the choices of theinitial conditions learning rates and convergent conditionsThe training process and classification efficiency become a


Table 4 Results of physical examinations

Measurement siteNumberDOS A L V Hue angle Hsaturation S

Ratio (A) Res Ratio (L) Res Ratio (V) Res

1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072

major problem when handling huge training data and italso becomes time consuming without a guaranteed globalminimum

In contrast the PSO algorithm is an optimization tech-nique that avoids the drawbacks of the least mean squarealgorithm and gradient descent method Due to difficultyin obtaining the partial derivatives of the nonlinear meansquared error function as (16) swarm intelligence can guidesearches using multiple particles rather than individualsIndividuals in a swarm approach the optimum through thepresent solution previous experiences and other individualsrsquoexperiences and can avoid trapping at a local minimum Anoptimization solution is sensitive to certain control factorssuch as the population sizes time-varying acceleration coef-ficients and the number of iteration training [33] Howeverwe have suggested that the suitable control factors terminatethe PSO algorithm to find the optimum solutions Using 30subjects our comparison indicates CRAusing PSO algorithmhas higher accuracies 95sim100 than multilayer networkswith accuracies of 70sim90

42 Physical Examinations Using testing data from the other10 subjects who agreed to participate in this study the overalltesting results are shown in Table 4 The tests used at least10 records at the measurement sites to calculate the flowvelocities and dimensionless numbers We have three maindegrees to decide the occlusion levels and the hue angles119867 correspond to define the level to quantify the similarityin each class The hue angles can determine the similaritylevels and more likely approach to the around of angles 240∘

(blue) 120∘ (green) and 0∘360∘ (red) from Class I to ClassIII respectively If 119867 is very similar its value will be close tothe primary color grades From these physical examinationswe observed the dysfunction of AVS could be screened by theproposed classifier

The CRA based classifier used a straightforward mathe-matical operation to construct a pattern recognition modelwith expandable or reducible training data Euclidean dis-tances are used to express the similarity degree between areference pattern and comparative patterns which representsthe largest similarity It was also designed as an adaptivepatternmechanismwith an adjustable recognition coefficientto enhance the higher screening accuracies Thus it canreduce the training data requirements

5 Conclusion

A CRA based classifier was proposed to screen the degreesof stenosis for an arteriovenous access occlusion evalu-ation Doppler ultrasound is a noninvasive measurementsupport to measure blood flow at multiple sites in an invivo examination The Doppler acoustic device with thesuggested 75MHzsim100MHz central operation frequencyand the sinusoidal tone bursts provides a good resolutionthat allows the measurement of the blood velocities inthe arteriovenous access The dimensionless numbers theratios of supracritical Reynolds and resistive indexes canbe calculated with the peak-systolic velocities the peak-diastolic velocities and the heart rates The trends in theratios of the supracritical Reynolds numbers and the resistive


indices indicate a promising way to evaluate the occlusionlevels at multiple measurement sites The proposed classifierhas a flexible pattern mechanism with add-in and delete-off training data and the capacity to adjust the recognitioncoefficient to enhance accuracy For 40 subjects dividedinto training and testing groups the proposed classifier wasvalidated to survey the level of occlusion for the clinicianThe results can be presented in colors as red green and blueand can be implemented in computer graphics applicationsFurther study could also implement the proposed screeningmethod in an advanced embedded systemor a programmablemicroprocessor and could be used to allow integration withhandheld healthcare systems

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work is supported in part by the National ScienceCouncil of Taiwan under Contract nos NSC 101-2218-E-244-003 and NSC 102-2218-E-218-006 duration August 1 2012sim

July 31 2014 and by the Institutional Review Board (IRB)of the Kaohsiung Veterans General Hospital Tainan Branchunder Contract no VGHKS13-CT12-11

References

[1] J J Sands ldquoVascular access monitoring improves outcomesrdquoBlood Purification vol 23 no 1 pp 45ndash49 2005

[2] A Asif F N Gadalean D Merrill et al ldquoInflow stenosisin arteriovenous fistulas and grafts a multicenter prospectivestudyrdquo Kidney International vol 67 no 5 pp 1986ndash1992 2005

[3] Y M Akay M Akay W Welkowitz J L Semmlow and J BKostis ldquoNoninvasive acoustical detection of coronary arterydisease a comparative study of signal processing methodsrdquoIEEE Transactions on Biomedical Engineering vol 40 no 6 pp571ndash578 1993

[4] Y Suzuki M Fukasawa T Mori O Sakata A Hattori and TKato ldquoElemental study on auscultaiting diagnosis support sys-tem of hemodialysis shunt stenosis by ANNrdquo IEEJ Transactionson Electronics Information and Systems vol 130 no 3 pp 401ndash406 2010

[5] Y Suzuki M Fukasawa O Sakata H Kato A Hattori and TKato ldquoAn auscultaiting diagnosis support system for assessinghemodialysis shunt stenosis by using self-organizingmaprdquo IEEJTransactions on Electronics Information and Systems vol 131no 1 pp 160ndash166 2011

[6] W L Chen C H Lin T S Chen P J Chen and C DKan ldquoStenosis detection algorithm using the Burg method ofautoregressive model for hemodialysis patients evaluation ofarteriovenous shunt stenosisrdquo Journal of Medical and BiologicalEngineering vol 33 no 4 pp 356ndash362 2013

[7] W-L Chen T Chen C-H Lin P-J Chen and C-D KanldquoPhonoangiography with a fractional order chaotic system-anew and easy algorithm in analyzing residual arteriovenousaccess stenosisrdquoMedical amp Biological Engineering ampComputingvol 51 no 9 pp 1011ndash1019 2013

[8] POVesquezMMMarco andBMandersson ldquoArteriovenousfistula stenosis detection using wavelets and support vectormachinesrdquo in Proceedings of the 31st Annual InternationalConference of the IEEE Engineering in Medicine and BiologySociety (EMBS rsquo09) pp 1298ndash1301 Minneapolis Minn USASeptember 2009

[9] M Gram J T Olesen H C Riis et al ldquoStenosis detectionalgorithm for screening of arteriovenous fistulaerdquo in 15thNordic-Baltic Conference on Biomedical Engineering and Medi-cal Physics vol 34 of IFMBE Proceedings pp 241ndash244 SpringerBerlin Germany 2011

[10] W-L Chen C-D Kan C-H Lin and T Chen ldquoA rule-baseddecision-making diagnosis system to evaluate arteriovenousshunt stenosis for hemodialysis treatment of patients usingfuzzy petri netsrdquo IEEE Journal of Biomedical and Health Infor-matics vol 18 no 2 pp 703ndash713 2014

[11] D H Lee J H Ahn S S Jeong K S Eo and M SPark ldquoRoutine transradial access for conventional cerebralangiography a single operatorrsquos experience of its feasibility andsafetyrdquoThe British Journal of Radiology vol 77 no 922 pp 831ndash838 2004

[12] PWiese and B Nonnast-Daniel ldquoColour doppler ultrasound indialysis accessrdquoNephrology Dialysis Transplantation vol 19 no8 pp 1956ndash1963 2004

[13] F Basseau N Grenier H Trillaud et al ldquoVolume flowmeasure-ment in hemodialysis shunts using time-domain correlationrdquoJournal of Ultrasound in Medicine vol 18 no 3 pp 177ndash1831999

[14] T Ogawa O Matsumura A Matsuda H Hasegawa and TMitarai ldquoBrachial artery blood flow measurement a simpleand noninvasive method to evaluate the need for arteriovenousfistula repairrdquoDialysis ampTransplantation vol 40 no 5 pp 206ndash210 2011

[15] X Xu J T Yen and K K Shung ldquoA low-cost bipolar pulsegenerator for high-frequency ultrasound applicationsrdquo IEEETransactions on Ultrasonics Ferroelectrics and Frequency Con-trol vol 54 no 2 pp 443ndash447 2007

[16] X Xu L Sun J M Cannata J T Yen and K K Shung ldquoHigh-frequency ultrasound Doppler system for biomedical applica-tions with a 30-MHz linear arrayrdquo Ultrasound in Medicine andBiology vol 34 no 4 pp 638ndash646 2008

[17] E L Madsen M E Deaner and J Mehi ldquoProperties ofphantom tissuelike polymethylpentene in the frequency range20ndash70MHZrdquoUltrasound in Medicine and Biology vol 37 no 8pp 1327ndash1339 2011

[18] C-H Lin ldquoAssessment of bilateral photoplethysmography forlower limb peripheral vascular occlusive disease using colorrelation analysis classifierrdquo Computer Methods and Programs inBiomedicine vol 103 no 3 pp 121ndash131 2011

[19] R C Gonzales and R E Woods Digital Image ProcessingPrentice Hall Press 2002

[20] A Ratnaweera S K Halgamuge and H C Watson ldquoSelf-organizing hierarchical particle swarm optimizer with time-varying acceleration coefficientsrdquo IEEE Transactions on Evolu-tionary Computation vol 8 no 3 pp 240ndash255 2004

[21] J-X Wu C-H Lin Y-C Du and T Chen ldquoSprott chaossynchronization classifier for diabetic foot peripheral vascularocclusive disease estimationrdquo IET Science Measurement ampTechnology Manuscript vol 6 no 6 pp 533ndash540 2012

[22] L Sun C-L Lien X Xu and K K Shung ldquoIn vivo cardiacimaging of adult zebrafish using high frequency ultrasound


(45ndash75MHz)rdquo Ultrasound in Medicine and Biology vol 34 no1 pp 31ndash39 2008

[23] V Cucevic A S Brown and F S Foster ldquoThermal assessment of40-MHz pulsed Doppler ultrasound in human eyerdquoUltrasoundin Medicine and Biology vol 31 no 4 pp 565ndash573 2005

[24] L C Nguyen F T H Yu and G Cloutier ldquoCyclic changes inblood echogenicity under pulsatile flow are frequency depen-dentrdquo Ultrasound in Medicine and Biology vol 34 no 4 pp664ndash673 2008

[25] J-X Wu Y-C Du C-H Lin P-J Chen and T Chen ldquoA novelbipolar pulse generator for high-frequency ultrasound systemrdquoin Proceedings of the IEEE International Ultrasonics Symposium(IUS rsquo13) pp 1571ndash1574 Prague Czech Republic July 2013

[26] P Fish Physics and Instrumentation of Diagnostic MedicalUltrasound John Wiley amp Sons

[27] W Qiu Y Yu F K Tsang and L Sun ldquoAn FPGA-based openplatform for ultrasound biomicroscopyrdquo IEEE Transactions onUltrasonics Ferroelectrics and Frequency Control vol 59 no 7pp 1432ndash1442 2012

[28] R O Bude and J M Rubin ldquoRelationship between the resistiveindex and vascular compliance and resistancerdquo Radiology vol211 no 2 pp 411ndash417 1999

[29] P V Ennezat S Marechaux M Six-Carpentier et al ldquoRenalresistance index and its prognostic significance in patientswith heart failure with preserved ejection fractionrdquo NephrologyDialysis Transplantation vol 26 no 12 pp 3908ndash3913 2011

[30] A F Stalder A FrydrychowiczM F Russe et al ldquoAssessment offlow instabilities in the healthy aorta using flow-sensitive MRIrdquoJournal of Magnetic Resonance Imaging vol 33 no 4 pp 839ndash846 2011

[31] R Maekawa M R Smith and S W van Sciver ldquoPressuredrop measurements of prototype NET and CEA cable-in-conduit conductors (CICCs)rdquo IEEE Transactions on AppliedSuperconductivity vol 5 no 2 pp 741ndash744 1995

[32] M K Banerjee R Ganguly and A Datta ldquoEffect of pulsatileflow waveform andWomersley number on the flow in stenosedarterial geometryrdquo ISRN Biomathematics vol 2012 Article ID853056 17 pages 2012

[33] C-M Li Y-CDu J-XWu et al ldquoSynchronizing chaotificationwith support vector machine and wolf pack search algorithmfor estimation of peripheral vascular occlusion in diabetesmellitusrdquo Biomedical Signal Processing and Control vol 9 pp45ndash55 2014

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



fuzzy petri nets are carried out to perform the classificationtasks However phonography techniques are affected by themeasurement sites the sizes of the sampling window andthe parametric mathematical models for phonoangiography(PCG) signal analysis In addition Doppler sonographyintravascular ultrasound image colour duplex ultrasoundand X-ray angiography [11ndash13] are also used to detect theocclusion in these vascular accesses in clinical examinationsStationary instruments provide reliable techniques and highaccuracy in clinical environments However they need toappreciate the operational principles and require extra learn-ing due to the limitations of the patients themselves

The ultrasound flowmeter is also a reliable and highlyaccurate technique and provides a good resolution in detect-ing blood properties [14ndash16] However the Doppler powerand velocity across an access under a pulsatile blood flow andthe ultrasonic backscattering signal from pulsatile flowingblood are affected by the shear rate as well as the accelerationof the flow velocity and the turbulences To improve ultra-sonic signal quality and the SNR at a specific depth Dopplerultrasound with a bipolar pulse drive can achieve the bestresolutions in high- and low-frequency ranges of 20MHz to50MHz and 8MHz to 20MHz respectively which is enoughto characterize the properties of the blood flow throughsurface and deep vessels It is well developed compact fastand free of irradiation [15ndash17] The purpose of this study isto use the Doppler ultrasound device for measuring bloodflow such as peak-systolic velocity peak-diastolic velocityend-diastolic velocity and heart rate Then the dimension-less numbers supracritical Reynolds (Resupra) number andresistive index (Res) are computed using flow velocitiesand the quantitative degrees are applied to screen the DOSColor relational analysis (CRA) [18] as a classifier performsthe recognition task for assessment of arteriovenous accessocclusion using a hue-saturation-value color model [19] Ithas a flexible pattern mechanism with add-in and delete-offtraining data and does not demand strict statistical methodsand inference rules in real time screening applications [18]An adaptive modeling system with an adjusting recognitioncoefficient to enhance the patency rate using a particle swarmoptimization (PSO) algorithm was utilized to increase theaccuracy [20 21] For 30 follow-up patients the results showthe proposed screening system is more efficient for occlusionevaluation

The rest of this paper is organized as follows Section 2addresses themeasurement technological support and exper-imental setup and Section 3 describes the adaptive color rela-tion classifier In Sections 4 and 5 experimental results andconclusion are provided and the efficiency of the proposedscreening method is demonstrated

2 Measurement Technological Support andExperimental Setup

21 Doppler Ultrasound Measurement In clinical examina-tions auscultation and ultrasound examinations are usedto detect the arteriovenous shunt (AVS) occlusion Ultra-sound examinations such as Doppler ultrasound imaging

intravascular ultrasound imaging and ultrasound flowmeter[22 23] are reliable and highly accurate clinical techniquesIt can operate in A-scan and B-scan instrument using low-frequency and high-frequency sound waves which can beprocessed to create quantitative images formarking the vesselaxis and diameter of the 2D ultrasound image [24 25] andalso provides a good resolution in detecting blood propertiessuch as flow velocities aggregation and coagulation [14ndash16] as shown in Figure 1 In a follow-up examination itcan also be used to determine whether an abnormality isstable or a treatment is working Among these the ultrasoundflowmeter is noninvasive well developed compact andfree of irradiation measurement technique The purposeof this study is to use the Doppler ultrasound device formeasurements of physiologic blood flow in AVS access suchas Doppler ultrasound velocity flow peak-systolic velocitypeak-diastolic velocity end-diastolic velocity and heart rate

Figure 1 shows the Doppler angles correct with flowvelocities that can be estimated from the Doppler shift Themaximum frequency 119891max and mode frequency of Dopplerspectra are less changed along with the change in the samplevolume lengthwhen setting the sample volumeposition at thecenter of the blood vesselThe flow velocity within the samplevolume is measured as [26]

119881 =

119891119889119888

21198910 cos 120579

(1)

where 119888 is the speed of sound in the blood the nominalvelocity of ultrasound in tissue is approximately 1540msec119891119889 is the Doppler shift frequency 1198910 is the center frequencyof transmitted ultrasound and 120579 is the Doppler angle 50sim80degrees between the acoustic beam and blood flow

In a narrowed access turbulence occurswhen blood flowsthrough a stenosis cross-sectional area increasing blood vis-cosity and flow resistance In previous studies [27] Dopplershift usually encounters a velocity range of 010msecsim

140msec in clinical measurement and represents a fre-quency shift depending on the transducer frequency Thesingle transducer was excited by a pulse repetition frequency(PRF) using bipolar pulses [25 27] as shown in Figure 2Then a 12 bit high-resolution digitizer with 200MHz sam-pling frequency was used to record backscattering and orig-inal signals by PRF trigger synchronize control with FPGAcontrol (DE2-70 Altera San Jose CA USA) It has goodresolution within a focusing zone of 100mm times 100mmand attenuation from 250 dB to 200 dB Thus the Dopplerfrequency shift can be calculated by those two signals inan embedded system One advantage of the bipolar pulsesdesignwith FPGA control is that it drives a higher voltage andshorter pulse width than the traditional ultrasound systemThose features can provide better SNR of the signal and aresuitable to detect blood changes in a hemodialysis access withthe bipolar pulse control in the ranges of 75MHz to 10MHzThe bipolar pulse design is also a low-cost design that can beassembled easily into a handheld system

22 Hemodynamic Analysis A flow with periodic variationsis a so-called pulsatile flow in small blood vessels or large



Loop site (L site)

(a)

Venous transducer

DBlood flow

Turbulent flow

120579

Skin Fixed

120579

Fixed


High-frequency






P-P intervalVp

Vm

(b)








Re 119904 =

119881119901 minus 119881119898

119881119901

(2)



Dataacquisition


Blood fow

Expander

Limiter

θ

Timercountersystem

AD converter

Band passfilter

Preamplifier


Dopplerangle

(a)

Width (mm) 0 10 30

Dep

th (m

m)

(dB)

Bipolar pulse

Beam profilep

5-cycle

100V

25

20

15

10

5

20

15

10

5

0

(b)



Repeak = 169120572083Stminus027 (3)



119863119867 =

4119860

119875


119875 =

119873119897

sum

119897=0

119871 119897 119897 = 1 2 3 119873119897

(4)







DOS = 1 minus (

119889119867

119863119867

)

2

(5)






Ratio(A) =

100381610038161003816100381610038161003816100381610038161003816


100381610038161003816100381610038161003816100381610038161003816

=


Ratio(L) =


= 1

Ratio(V) =

100381610038161003816100381610038161003816100381610038161003816


100381610038161003816100381610038161003816100381610038161003816

=


(6)











[[[[[[[[[

[

Φ119888 (1)

Φ119888 (2)

Φ119888 (119896)

Φ119888 (119870)

]]]]]]]]]

]

=

[[[[[[[[[

[



d

d


d

d


]]]]]]]]]

]

(7)








2 [26]

0

1

2

3

4

5


Rat

ioV

pV

m


DOS lt 03

(a)

0

02

04

06

08

1

12

14

Tren

d pa

ttern


DOS lt 03


(b)



ED (119896) = radic

6

sum

119894=1

(Δ120601119894 (119896))2 Δ120601119894 (119896) =

1003816100381610038161003816120601119894 (0) minus 120601119894 (119896)

1003816100381610038161003816 (8)



120588 (119896) = 120585 exp [minus120585ED (119896)] (9)


120588Iave =

sum119873119868

119905=1120588I(119905)

119873I

120588IIave =

sum119873II119905=1

120588II

(119905)

119873II

120588IIIave =

sum119873III119905=1

120588III

(119905)

119873III

(10)


Template 1

Compare

Template 2

Compare

Template K

Compare

ED(1)

ED(2)

ED(K)

120588(1)

120588(2)

Template 10

Compare ED(10) 120588(10)

Template k


120588(K)

Euclidean distance

Gray grade

DOS gt 05

03 lt DOD lt 05

DOS lt 03

++

minus

Parameterestimation

Error

r

r

b

g

g

Primary color grade

Ratio(A)

Ratio(L)

Ratio(V)

Res(L)

Res(A)

Res(V)

Hue anglered

Saturationvalue S

Hue angleblue

Hue anglegreen

Class III

Class I

Class II

Desiredangle

Refe

renc

e pa

ttern

Tran

sfer

RG

B co

lor s

pace

to H

SV c

olor

spac



120588mim = min [120588Iave 120588

IIave 120588

IIIave]

120588max = max [120588Iave 120588

IIave 120588

IIIave]

(11)


as

119867 =

[60∘

times (

119892 minus 119887

Δ120588

) + 360∘] mod 360

∘ if 120588max = 120588

IIIave

60∘

times (

119887 minus 119903

Δ120588

) + 120∘ if 120588max = 120588

Iave

60∘

times (

119903 minus 119892

Δ120588

) + 240∘ if 120588max = 120588

IIave

(12)



119903 =


Δ120588

119892 =


Δ120588

119887 =


Δ120588

(13)


119878 =

120574 minus 120588min120574

120588min = 120588max

120574 = 120588max 120588max = 0

(14)


119867119862 = [

119867

360∘] 119867119862 isin [0 1] (15)




min (MSEF) = min(

1

119870

119870

sum

119896=1

[119879 (119896) minus 119867119862 (119896)]2) le 120576 (16)



119901= [120585119901

1 120585119901

2 120585

119901

119892 120585

119901

119866] The



Δ120585119901

119892 Δ120585119901

119892= [Δ120585

119901

1 Δ120585119901

2 Δ120585

119901

119892 Δ120585

119901



Δ120585119901+1

119892= 120596Δ120585

119901

119892+ 1198881rand1 (120585best119892 minus 120585

119901

119892)

+ 1198882rand2 (120585best minus 120585119901

119892)

1198881 = (1198871 minus 1198861)

119901

119901max+ 1198861 1198882 = (1198872 minus 1198862)

119901

119901max+ 1198862

(17)

Position

120585119901+1

119892= 120585119901

119892+ Δ120585119901+1

119892 (18)


Start



as equation (16)




Stop

Yes

No

Step 1

Step 2

Step 3

Step 4


120585p = [1205851p 120585

2p 120585

gp 120585

Gp ]
































0 5 10 15 20 25 30 35 40 45 500

01

02

03

04


Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)








1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











Loop site (L site)

(a)

Venous transducer

DBlood flow

Turbulent flow

120579

Skin Fixed

120579

Fixed


High-frequency






P-P intervalVp

Vm

(b)








Re 119904 =

119881119901 minus 119881119898

119881119901

(2)



Dataacquisition


Blood fow

Expander

Limiter

θ

Timercountersystem

AD converter

Band passfilter

Preamplifier


Dopplerangle

(a)

Width (mm) 0 10 30

Dep

th (m

m)

(dB)

Bipolar pulse

Beam profilep

5-cycle

100V

25

20

15

10

5

20

15

10

5

0

(b)



Repeak = 169120572083Stminus027 (3)



119863119867 =

4119860

119875


119875 =

119873119897

sum

119897=0

119871 119897 119897 = 1 2 3 119873119897

(4)







DOS = 1 minus (

119889119867

119863119867

)

2

(5)






Ratio(A) =

100381610038161003816100381610038161003816100381610038161003816


100381610038161003816100381610038161003816100381610038161003816

=


Ratio(L) =


= 1

Ratio(V) =

100381610038161003816100381610038161003816100381610038161003816


100381610038161003816100381610038161003816100381610038161003816

=


(6)











[[[[[[[[[

[

Φ119888 (1)

Φ119888 (2)

Φ119888 (119896)

Φ119888 (119870)

]]]]]]]]]

]

=

[[[[[[[[[

[



d

d


d

d


]]]]]]]]]

]

(7)








2 [26]

0

1

2

3

4

5


Rat

ioV

pV

m


DOS lt 03

(a)

0

02

04

06

08

1

12

14

Tren

d pa

ttern


DOS lt 03


(b)



ED (119896) = radic

6

sum

119894=1

(Δ120601119894 (119896))2 Δ120601119894 (119896) =

1003816100381610038161003816120601119894 (0) minus 120601119894 (119896)

1003816100381610038161003816 (8)



120588 (119896) = 120585 exp [minus120585ED (119896)] (9)


120588Iave =

sum119873119868

119905=1120588I(119905)

119873I

120588IIave =

sum119873II119905=1

120588II

(119905)

119873II

120588IIIave =

sum119873III119905=1

120588III

(119905)

119873III

(10)


Template 1

Compare

Template 2

Compare

Template K

Compare

ED(1)

ED(2)

ED(K)

120588(1)

120588(2)

Template 10

Compare ED(10) 120588(10)

Template k


120588(K)

Euclidean distance

Gray grade

DOS gt 05

03 lt DOD lt 05

DOS lt 03

++

minus

Parameterestimation

Error

r

r

b

g

g

Primary color grade

Ratio(A)

Ratio(L)

Ratio(V)

Res(L)

Res(A)

Res(V)

Hue anglered

Saturationvalue S

Hue angleblue

Hue anglegreen

Class III

Class I

Class II

Desiredangle

Refe

renc

e pa

ttern

Tran

sfer

RG

B co

lor s

pace

to H

SV c

olor

spac



120588mim = min [120588Iave 120588

IIave 120588

IIIave]

120588max = max [120588Iave 120588

IIave 120588

IIIave]

(11)


as

119867 =

[60∘

times (

119892 minus 119887

Δ120588

) + 360∘] mod 360

∘ if 120588max = 120588

IIIave

60∘

times (

119887 minus 119903

Δ120588

) + 120∘ if 120588max = 120588

Iave

60∘

times (

119903 minus 119892

Δ120588

) + 240∘ if 120588max = 120588

IIave

(12)



119903 =


Δ120588

119892 =


Δ120588

119887 =


Δ120588

(13)


119878 =

120574 minus 120588min120574

120588min = 120588max

120574 = 120588max 120588max = 0

(14)


119867119862 = [

119867

360∘] 119867119862 isin [0 1] (15)




min (MSEF) = min(

1

119870

119870

sum

119896=1

[119879 (119896) minus 119867119862 (119896)]2) le 120576 (16)



119901= [120585119901

1 120585119901

2 120585

119901

119892 120585

119901

119866] The



Δ120585119901

119892 Δ120585119901

119892= [Δ120585

119901

1 Δ120585119901

2 Δ120585

119901

119892 Δ120585

119901



Δ120585119901+1

119892= 120596Δ120585

119901

119892+ 1198881rand1 (120585best119892 minus 120585

119901

119892)

+ 1198882rand2 (120585best minus 120585119901

119892)

1198881 = (1198871 minus 1198861)

119901

119901max+ 1198861 1198882 = (1198872 minus 1198862)

119901

119901max+ 1198862

(17)

Position

120585119901+1

119892= 120585119901

119892+ Δ120585119901+1

119892 (18)


Start



as equation (16)




Stop

Yes

No

Step 1

Step 2

Step 3

Step 4


120585p = [1205851p 120585

2p 120585

gp 120585

Gp ]
































0 5 10 15 20 25 30 35 40 45 500

01

02

03

04


Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)








1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Dataacquisition


Blood fow

Expander

Limiter

θ

Timercountersystem

AD converter

Band passfilter

Preamplifier


Dopplerangle

(a)

Width (mm) 0 10 30

Dep

th (m

m)

(dB)

Bipolar pulse

Beam profilep

5-cycle

100V

25

20

15

10

5

20

15

10

5

0

(b)



Repeak = 169120572083Stminus027 (3)



119863119867 =

4119860

119875


119875 =

119873119897

sum

119897=0

119871 119897 119897 = 1 2 3 119873119897

(4)







DOS = 1 minus (

119889119867

119863119867

)

2

(5)






Ratio(A) =

100381610038161003816100381610038161003816100381610038161003816


100381610038161003816100381610038161003816100381610038161003816

=


Ratio(L) =


= 1

Ratio(V) =

100381610038161003816100381610038161003816100381610038161003816


100381610038161003816100381610038161003816100381610038161003816

=


(6)











[[[[[[[[[

[

Φ119888 (1)

Φ119888 (2)

Φ119888 (119896)

Φ119888 (119870)

]]]]]]]]]

]

=

[[[[[[[[[

[



d

d


d

d


]]]]]]]]]

]

(7)








2 [26]

0

1

2

3

4

5


Rat

ioV

pV

m


DOS lt 03

(a)

0

02

04

06

08

1

12

14

Tren

d pa

ttern


DOS lt 03


(b)



ED (119896) = radic

6

sum

119894=1

(Δ120601119894 (119896))2 Δ120601119894 (119896) =

1003816100381610038161003816120601119894 (0) minus 120601119894 (119896)

1003816100381610038161003816 (8)



120588 (119896) = 120585 exp [minus120585ED (119896)] (9)


120588Iave =

sum119873119868

119905=1120588I(119905)

119873I

120588IIave =

sum119873II119905=1

120588II

(119905)

119873II

120588IIIave =

sum119873III119905=1

120588III

(119905)

119873III

(10)


Template 1

Compare

Template 2

Compare

Template K

Compare

ED(1)

ED(2)

ED(K)

120588(1)

120588(2)

Template 10

Compare ED(10) 120588(10)

Template k


120588(K)

Euclidean distance

Gray grade

DOS gt 05

03 lt DOD lt 05

DOS lt 03

++

minus

Parameterestimation

Error

r

r

b

g

g

Primary color grade

Ratio(A)

Ratio(L)

Ratio(V)

Res(L)

Res(A)

Res(V)

Hue anglered

Saturationvalue S

Hue angleblue

Hue anglegreen

Class III

Class I

Class II

Desiredangle

Refe

renc

e pa

ttern

Tran

sfer

RG

B co

lor s

pace

to H

SV c

olor

spac



120588mim = min [120588Iave 120588

IIave 120588

IIIave]

120588max = max [120588Iave 120588

IIave 120588

IIIave]

(11)


as

119867 =

[60∘

times (

119892 minus 119887

Δ120588

) + 360∘] mod 360

∘ if 120588max = 120588

IIIave

60∘

times (

119887 minus 119903

Δ120588

) + 120∘ if 120588max = 120588

Iave

60∘

times (

119903 minus 119892

Δ120588

) + 240∘ if 120588max = 120588

IIave

(12)



119903 =


Δ120588

119892 =


Δ120588

119887 =


Δ120588

(13)


119878 =

120574 minus 120588min120574

120588min = 120588max

120574 = 120588max 120588max = 0

(14)


119867119862 = [

119867

360∘] 119867119862 isin [0 1] (15)




min (MSEF) = min(

1

119870

119870

sum

119896=1

[119879 (119896) minus 119867119862 (119896)]2) le 120576 (16)



119901= [120585119901

1 120585119901

2 120585

119901

119892 120585

119901

119866] The



Δ120585119901

119892 Δ120585119901

119892= [Δ120585

119901

1 Δ120585119901

2 Δ120585

119901

119892 Δ120585

119901



Δ120585119901+1

119892= 120596Δ120585

119901

119892+ 1198881rand1 (120585best119892 minus 120585

119901

119892)

+ 1198882rand2 (120585best minus 120585119901

119892)

1198881 = (1198871 minus 1198861)

119901

119901max+ 1198861 1198882 = (1198872 minus 1198862)

119901

119901max+ 1198862

(17)

Position

120585119901+1

119892= 120585119901

119892+ Δ120585119901+1

119892 (18)


Start



as equation (16)




Stop

Yes

No

Step 1

Step 2

Step 3

Step 4


120585p = [1205851p 120585

2p 120585

gp 120585

Gp ]
































0 5 10 15 20 25 30 35 40 45 500

01

02

03

04


Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)








1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












DOS = 1 minus (

119889119867

119863119867

)

2

(5)






Ratio(A) =

100381610038161003816100381610038161003816100381610038161003816


100381610038161003816100381610038161003816100381610038161003816

=


Ratio(L) =


= 1

Ratio(V) =

100381610038161003816100381610038161003816100381610038161003816


100381610038161003816100381610038161003816100381610038161003816

=


(6)











[[[[[[[[[

[

Φ119888 (1)

Φ119888 (2)

Φ119888 (119896)

Φ119888 (119870)

]]]]]]]]]

]

=

[[[[[[[[[

[



d

d


d

d


]]]]]]]]]

]

(7)








2 [26]

0

1

2

3

4

5


Rat

ioV

pV

m


DOS lt 03

(a)

0

02

04

06

08

1

12

14

Tren

d pa

ttern


DOS lt 03


(b)



ED (119896) = radic

6

sum

119894=1

(Δ120601119894 (119896))2 Δ120601119894 (119896) =

1003816100381610038161003816120601119894 (0) minus 120601119894 (119896)

1003816100381610038161003816 (8)



120588 (119896) = 120585 exp [minus120585ED (119896)] (9)


120588Iave =

sum119873119868

119905=1120588I(119905)

119873I

120588IIave =

sum119873II119905=1

120588II

(119905)

119873II

120588IIIave =

sum119873III119905=1

120588III

(119905)

119873III

(10)


Template 1

Compare

Template 2

Compare

Template K

Compare

ED(1)

ED(2)

ED(K)

120588(1)

120588(2)

Template 10

Compare ED(10) 120588(10)

Template k


120588(K)

Euclidean distance

Gray grade

DOS gt 05

03 lt DOD lt 05

DOS lt 03

++

minus

Parameterestimation

Error

r

r

b

g

g

Primary color grade

Ratio(A)

Ratio(L)

Ratio(V)

Res(L)

Res(A)

Res(V)

Hue anglered

Saturationvalue S

Hue angleblue

Hue anglegreen

Class III

Class I

Class II

Desiredangle

Refe

renc

e pa

ttern

Tran

sfer

RG

B co

lor s

pace

to H

SV c

olor

spac



120588mim = min [120588Iave 120588

IIave 120588

IIIave]

120588max = max [120588Iave 120588

IIave 120588

IIIave]

(11)


as

119867 =

[60∘

times (

119892 minus 119887

Δ120588

) + 360∘] mod 360

∘ if 120588max = 120588

IIIave

60∘

times (

119887 minus 119903

Δ120588

) + 120∘ if 120588max = 120588

Iave

60∘

times (

119903 minus 119892

Δ120588

) + 240∘ if 120588max = 120588

IIave

(12)



119903 =


Δ120588

119892 =


Δ120588

119887 =


Δ120588

(13)


119878 =

120574 minus 120588min120574

120588min = 120588max

120574 = 120588max 120588max = 0

(14)


119867119862 = [

119867

360∘] 119867119862 isin [0 1] (15)




min (MSEF) = min(

1

119870

119870

sum

119896=1

[119879 (119896) minus 119867119862 (119896)]2) le 120576 (16)



119901= [120585119901

1 120585119901

2 120585

119901

119892 120585

119901

119866] The



Δ120585119901

119892 Δ120585119901

119892= [Δ120585

119901

1 Δ120585119901

2 Δ120585

119901

119892 Δ120585

119901



Δ120585119901+1

119892= 120596Δ120585

119901

119892+ 1198881rand1 (120585best119892 minus 120585

119901

119892)

+ 1198882rand2 (120585best minus 120585119901

119892)

1198881 = (1198871 minus 1198861)

119901

119901max+ 1198861 1198882 = (1198872 minus 1198862)

119901

119901max+ 1198862

(17)

Position

120585119901+1

119892= 120585119901

119892+ Δ120585119901+1

119892 (18)


Start



as equation (16)




Stop

Yes

No

Step 1

Step 2

Step 3

Step 4


120585p = [1205851p 120585

2p 120585

gp 120585

Gp ]
































0 5 10 15 20 25 30 35 40 45 500

01

02

03

04


Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)








1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation
















2 [26]

0

1

2

3

4

5


Rat

ioV

pV

m


DOS lt 03

(a)

0

02

04

06

08

1

12

14

Tren

d pa

ttern


DOS lt 03


(b)



ED (119896) = radic

6

sum

119894=1

(Δ120601119894 (119896))2 Δ120601119894 (119896) =

1003816100381610038161003816120601119894 (0) minus 120601119894 (119896)

1003816100381610038161003816 (8)



120588 (119896) = 120585 exp [minus120585ED (119896)] (9)


120588Iave =

sum119873119868

119905=1120588I(119905)

119873I

120588IIave =

sum119873II119905=1

120588II

(119905)

119873II

120588IIIave =

sum119873III119905=1

120588III

(119905)

119873III

(10)


Template 1

Compare

Template 2

Compare

Template K

Compare

ED(1)

ED(2)

ED(K)

120588(1)

120588(2)

Template 10

Compare ED(10) 120588(10)

Template k


120588(K)

Euclidean distance

Gray grade

DOS gt 05

03 lt DOD lt 05

DOS lt 03

++

minus

Parameterestimation

Error

r

r

b

g

g

Primary color grade

Ratio(A)

Ratio(L)

Ratio(V)

Res(L)

Res(A)

Res(V)

Hue anglered

Saturationvalue S

Hue angleblue

Hue anglegreen

Class III

Class I

Class II

Desiredangle

Refe

renc

e pa

ttern

Tran

sfer

RG

B co

lor s

pace

to H

SV c

olor

spac



120588mim = min [120588Iave 120588

IIave 120588

IIIave]

120588max = max [120588Iave 120588

IIave 120588

IIIave]

(11)


as

119867 =

[60∘

times (

119892 minus 119887

Δ120588

) + 360∘] mod 360

∘ if 120588max = 120588

IIIave

60∘

times (

119887 minus 119903

Δ120588

) + 120∘ if 120588max = 120588

Iave

60∘

times (

119903 minus 119892

Δ120588

) + 240∘ if 120588max = 120588

IIave

(12)



119903 =


Δ120588

119892 =


Δ120588

119887 =


Δ120588

(13)


119878 =

120574 minus 120588min120574

120588min = 120588max

120574 = 120588max 120588max = 0

(14)


119867119862 = [

119867

360∘] 119867119862 isin [0 1] (15)




min (MSEF) = min(

1

119870

119870

sum

119896=1

[119879 (119896) minus 119867119862 (119896)]2) le 120576 (16)



119901= [120585119901

1 120585119901

2 120585

119901

119892 120585

119901

119866] The



Δ120585119901

119892 Δ120585119901

119892= [Δ120585

119901

1 Δ120585119901

2 Δ120585

119901

119892 Δ120585

119901



Δ120585119901+1

119892= 120596Δ120585

119901

119892+ 1198881rand1 (120585best119892 minus 120585

119901

119892)

+ 1198882rand2 (120585best minus 120585119901

119892)

1198881 = (1198871 minus 1198861)

119901

119901max+ 1198861 1198882 = (1198872 minus 1198862)

119901

119901max+ 1198862

(17)

Position

120585119901+1

119892= 120585119901

119892+ Δ120585119901+1

119892 (18)


Start



as equation (16)




Stop

Yes

No

Step 1

Step 2

Step 3

Step 4


120585p = [1205851p 120585

2p 120585

gp 120585

Gp ]
































0 5 10 15 20 25 30 35 40 45 500

01

02

03

04


Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)








1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Template 1

Compare

Template 2

Compare

Template K

Compare

ED(1)

ED(2)

ED(K)

120588(1)

120588(2)

Template 10

Compare ED(10) 120588(10)

Template k


120588(K)

Euclidean distance

Gray grade

DOS gt 05

03 lt DOD lt 05

DOS lt 03

++

minus

Parameterestimation

Error

r

r

b

g

g

Primary color grade

Ratio(A)

Ratio(L)

Ratio(V)

Res(L)

Res(A)

Res(V)

Hue anglered

Saturationvalue S

Hue angleblue

Hue anglegreen

Class III

Class I

Class II

Desiredangle

Refe

renc

e pa

ttern

Tran

sfer

RG

B co

lor s

pace

to H

SV c

olor

spac



120588mim = min [120588Iave 120588

IIave 120588

IIIave]

120588max = max [120588Iave 120588

IIave 120588

IIIave]

(11)


as

119867 =

[60∘

times (

119892 minus 119887

Δ120588

) + 360∘] mod 360

∘ if 120588max = 120588

IIIave

60∘

times (

119887 minus 119903

Δ120588

) + 120∘ if 120588max = 120588

Iave

60∘

times (

119903 minus 119892

Δ120588

) + 240∘ if 120588max = 120588

IIave

(12)



119903 =


Δ120588

119892 =


Δ120588

119887 =


Δ120588

(13)


119878 =

120574 minus 120588min120574

120588min = 120588max

120574 = 120588max 120588max = 0

(14)


119867119862 = [

119867

360∘] 119867119862 isin [0 1] (15)




min (MSEF) = min(

1

119870

119870

sum

119896=1

[119879 (119896) minus 119867119862 (119896)]2) le 120576 (16)



119901= [120585119901

1 120585119901

2 120585

119901

119892 120585

119901

119866] The



Δ120585119901

119892 Δ120585119901

119892= [Δ120585

119901

1 Δ120585119901

2 Δ120585

119901

119892 Δ120585

119901



Δ120585119901+1

119892= 120596Δ120585

119901

119892+ 1198881rand1 (120585best119892 minus 120585

119901

119892)

+ 1198882rand2 (120585best minus 120585119901

119892)

1198881 = (1198871 minus 1198861)

119901

119901max+ 1198861 1198882 = (1198872 minus 1198862)

119901

119901max+ 1198862

(17)

Position

120585119901+1

119892= 120585119901

119892+ Δ120585119901+1

119892 (18)


Start



as equation (16)




Stop

Yes

No

Step 1

Step 2

Step 3

Step 4


120585p = [1205851p 120585

2p 120585

gp 120585

Gp ]
































0 5 10 15 20 25 30 35 40 45 500

01

02

03

04


Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)








1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












min (MSEF) = min(

1

119870

119870

sum

119896=1

[119879 (119896) minus 119867119862 (119896)]2) le 120576 (16)



119901= [120585119901

1 120585119901

2 120585

119901

119892 120585

119901

119866] The



Δ120585119901

119892 Δ120585119901

119892= [Δ120585

119901

1 Δ120585119901

2 Δ120585

119901

119892 Δ120585

119901



Δ120585119901+1

119892= 120596Δ120585

119901

119892+ 1198881rand1 (120585best119892 minus 120585

119901

119892)

+ 1198882rand2 (120585best minus 120585119901

119892)

1198881 = (1198871 minus 1198861)

119901

119901max+ 1198861 1198882 = (1198872 minus 1198862)

119901

119901max+ 1198862

(17)

Position

120585119901+1

119892= 120585119901

119892+ Δ120585119901+1

119892 (18)


Start



as equation (16)




Stop

Yes

No

Step 1

Step 2

Step 3

Step 4


120585p = [1205851p 120585

2p 120585

gp 120585

Gp ]
































0 5 10 15 20 25 30 35 40 45 500

01

02

03

04


Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)








1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation


































0 5 10 15 20 25 30 35 40 45 500

01

02

03

04


Mea

n sq

uare

d er

ror

of o

ptim

al so

lutio

n

Convergence

(a)

0 10 20 30 40 500

2

4

6

8


Opt

imal

reco

gniti

onco

effici

ent

(b)

5

10

15

20

25

30

50

100

150

200

250

300

350

Hue angle

Hue

ang

le

Class II

Class III

Class III

Class I

Subj

ect n

umbe

r

(c)








1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













1 09129 111 0733 100 071 102 063 11647906480

2 09056 141 087 100 050 129 074 12269009707

3 05346 121 066 100 060 120 065 21992409023

4 04866 070 072 100 059 098 067 58031906140

5 04066 089 078 100 074 089 066 134868705659

6 03960 088 088 100 076 075 061 120801207880

7 01845 085 066 100 060 098 063 198206306034

8 01830 093 065 100 072 082 053 209989505357

9 01467 097 064 100 051 102 066 191179708221

10 00985 104 063 100 044 085 037 218095309072






5 Conclusion






Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













Acknowledgments



References






































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation
























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









research article multiple-site hemodynamic analysis of ...research article multiple-site hemodynamic...

Documents