The Effect of Optical Crosstalk on Accuracy of Reflectance

Research ArticleThe Effect of Optical Crosstalk on Accuracy of Reflectance-TypePulse Oximeter for Mobile Healthcare

Hyun Jae Baek JaeWook Shin and Jaegeol Cho

Department of Medical and Mechatronics Engineering Soonchunhyang University Asan Chungnam Republic of Korea

Correspondence should be addressed to Jaegeol Cho jaegeolchoschackr

Received 31 January 2018 Revised 5 September 2018 Accepted 16 September 2018 Published 21 October 2018

Guest Editor Ting Li

Copyright copy 2018 Hyun Jae Baek et al -is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

According to the theoretical equation of the pulse oximeter expressed by the ratio of amplitude (AC) and baseline (DC)obtained from the photoplethysmographic signal of two wavelengths the difference of the amount of light absorbed dependingon the melanin indicating the skin color is canceled by normalizing the AC value to the DC value of each wavelength-ereforetheoretically skin color does not affect the accuracy of oxygen saturationmeasurement However if there is a direct path for thelight emitting unit to the light receiving unit instead of passing through the human body the amount of light reflected by thesurface of the skin changes depending on the color of the skin As a result the amount of crosstalk that varies depending on theskin color affects the ratio of AC to DC resulting in errors in the calculation of the oxygen saturation value We made crosstalksensors and crosstalk-free sensors and performed desaturation experiments with respiratory gas control on subjects withvarious skin colors to perform oxygen saturation measurements ranging from 60 to 100 Experimental results showed thatthere was no difference in the measurement error of oxygen saturation according to skin color in the case of the sensor whichprevented crosstalk (minus08824 plusmn 22859 for Asian subjects 06741 plusmn 32822 for Caucasian subjects and 09669plusmn 22268 forAfrican American subjects) However a sensor that did not prevent crosstalk showed a large error in dark skin subjects(08258 plusmn 21603 for Asian subjects 08733plusmn 19716 for Caucasian subjects and minus30591 plusmn 39925 for African Americans) Basedon these results we reiterate the importance of sensor design in the development of pulse oximeters using reflectance-type sensors

1 Introduction

Since the Japanese biomedical engineer Takuo Aoyagi firstproposed the idea of using pulsatile light variation tomeasure arterial oxygen saturation in the 1970s a variety ofpulse oximeters have been studied and developed in manyuniversities and research institutes around the world [1 2]Pulse oximeters which are currently used in medical fa-cilities for patient monitoring in the intensive care unit andanesthesia in the operating room are most often used withfinger clip type sensors -is can be regarded as a trans-mittance-type sensor because the light emitting portionfor measuring the oxygen saturation is on one side of the

body and the light receiving portion is located on the op-posite side Because of these morphological features thetransmittance-type sensor is mainly used for finger or toeOn the other hand since the reflectance-type sensor has boththe light emitting portion and the light receiving portion onthe same plane it can be applied to various body parts anda representative example is forehead pulse oximeterReflectance-type sensors have not been short in their historyand have been developed in a variety of forms but they arestill not widely used in practice compared to transmittance-type sensors for many reasons [3 4] When a transmittance-type sensor is used for a recording of pulsation it can bemeasured ten times more strongly than when a reflection

HindawiJournal of Healthcare EngineeringVolume 2018 Article ID 3521738 8 pageshttpsdoiorg10115520183521738

type sensor is used [5] In addition since the sensor is wellfixed to the finger in the form of a clip the possibility ofmovement of the sensor and the possibility of entrance ofexternal light are small so that the quality of the signal to bemeasured is much more prevalent On the other hand inthe case of the reflectance-type sensor if the sensor is notfixed with the adhesive the sensor slips over the skinresulting in motion noise Nevertheless as the mobile healthdevices have recently been attracting attention there isa growing demand for pulse oximeters using reflectance-type sensors [6] As a mobile healthcare device modernreflectance-type pulse oximeter has led to changes inoptical sensor configurations Conventional reflectance-type pulse oximetry sensors consist of LEDs with two ormore wavelengths and a photodetector which are spaced4ndash11mm apart [7] As shown in Figure 1(a) an opticalbarrier is required to prevent direct coupling between thelight emitting part and the light receiving part and isapplied in the form of a partition between the LED andthe photodetector in the optical sensor Typically LEDsand photodetectors are packaged and physically sepa-rated using a black rubber material LEDs and photo-detector are exposed on the surface of the sensor orcoated with transparent epoxy In both cases LEDs andphotodetector which are completely separated are indirect contact with the skin As a result the photon ir-radiated from the light emitting part can reach the lightreceiving part only through the human body In recentyears optical sensor technology for pulsation measure-ment has become popular and the LED-photodetectorpair is integrated with the analog front-end makingit miniaturized and becoming a single component Forexample Maximrsquos MAX30100 includes both opticalsensors and measurement modules in a small size of56 times 28 times12 mm [8] Although the sensor module itself iscompletely physically separated from the light emittingunit and the light receiving unit a gap is formed betweenthe sensor module and the device while the module ismounted on the device Also in some cases there is alsocross talk caused by cover grass as described in Figure 1(b)As a result unlike conventional sensors photons directlycoupled to the inside of the sensor or reflected from theskin surface without passing through the human body aremeasured together at the light receiving part of pulseoximetry sensor

In this paper we emphasize the importance of sensordesign to prevent crosstalk when using a small modularreflectance-type pulse oximetry sensor according to recenttrends Specifically we analyzed the effect of opticalcrosstalk on the reflectance-type pulse oximeter accordingto the subjectrsquos skin color through theoretical and clinicalstudies First the BeerndashLambert law the theoreticalbackground of pulse oximetry was analyzed -en desa-turation test was performed to reduce the oxygen satura-tion to 70 by controlling the respiratory gas and the

oxygen saturation measurement results were comparedusing the sensors with and without crosstalk preventionDesaturation testing was approved by the Shenzhen Uni-versity Committee on Human Research and was conductedwith the consent of all subjects

2 Theoretical Formulation

-e principle of the pulse oximeter can be explained bymodified BeerndashLambert law [2 9 10] where I(λ) is de-tected light intensity Io(λ) is the incident light intensityε(λ) is a molar absorption coefficient C is the molarconcentration l(λ) is them mean path length and G(λ) isappropriate factors that account for the measurementgeometry-e signal that records changes in I(λ) caused bypulsatile cardiac activity is called photoplethysmogram(PPG) If we assume that the light absorbing materials aremelanin and other blood or skin-related components theamplitude for the red wavelength PPG signal (AC) can beexpressed as Equation (2) -e subscripts m and b denotemelanin and blood and A0 is a spectrum of all chromo-phores in human skin except m and b Also the baseline(DC) can be written as

A(λ ) lnIo(λ)

I(λ) ε(λ)Cl(λ C) + G(λ) (1)

ACRed I0 λRed( 1113857exp1113858minusεm λRed( 1113857Cmlm λRed( 1113857

minus εb λRed( 1113857Cblbminusdia λRed( 1113857minusA0 λRed( 11138571113859

minus I0 λRed( 1113857exp1113858minusεm λRed( 1113857Cmlm λRed( 1113857

minus εb λRed( 1113857Cblbminussys λRed( 1113857minusA0 λRed( 11138571113859

where A0 λRed( 1113857 A0prime λRed( 1113857 + G λRed( 1113857

(2)

DCRed I0 λRed( 1113857exp1113858minusεm λRed( 1113857Cmlm λRed( 1113857

minus εb λRed( 1113857Cblbminussys λRed( 1113857minusA0 λRed( 11138571113859(3)

-e ratio of the AC to the DC reflects the change inthe maximum blood volume during the systolicperiod (MBVSYS) -e oxygen saturation can be obtainedby using the ratio of MBVSYS for the infrared to thosefor the red since the red light is less absorbed in HbO2 thanin Hb and on the contrary in the case of infraredFrom (2) and (3) MBVSYS for the red and infraredwavelength PPG can be obtained as Equations (4) and (5)respectively where α(λ) minusεm(λ)Cmlm(λ) β(λ)

minusεb(λ)Cblb(λ) and c(λ) minusA0(λ) Finally the ratio for theoxygen saturation calculation can be derived asR [ACRedDCRed][ACIRDCIR] and is linked to theoxygen saturation in a function of SpO2 f(R)

2 Journal of Healthcare Engineering

ACRed

DCRedI0 λRed( )exp[α λRed( ) + βdia λRed( ) + c λRed( )]minus I0 λRed( )exp α λRed( ) + βsys λRed( ) + c λRed( )[ ]

I0 λRed( )exp α λRed( ) + βsys λRed( ) + c λRed( )[ ]

exp minusεb λRed( )Cblbminusdia λRed( )[ ]minus exp minusεb λRed( )Cblbminussys λRed( )[ ]

exp minusεb λRed( )Cblbminussys λRed( )[ ]

εb λRed( )Cb lbminussys λRed( )minus lbminusdia λRed( )[ ]

(4)

ACIR

DCIR εb λIR( )Cb lbminussys λIR( )minus lbminusdia λIR( )[ ] (5)

If there is crosstalk in pulse oximetry sensor like Fig-ure 1(b) detected light intensity will be I(λ) + Ic(λ) andthe MBVSYS for ref PPG is rewritten as Equation (6)

where subscript c denotes crosstalk As described in (7)Ic(λ)makes a change in the ratio value which again causesan oxygen saturation error

ACRedprime

DCRedprimeI0 λRed( )exp α λRed( ) + βdia λRed( ) + c λRed( )[ ] + Ic λRed( ) minus I0 λRed( )exp α λRed( ) + βsys λRed( ) + c λRed( )[ ] + Ic λRed( )

I0 λRed( )exp α λRed( ) + βsys λRed( ) + c λRed( )[ ] + Ic λRed( )

(6)

ACRedprime

DCRedprimeleACRed

DCRed Rprime

ACRedprime DCRedprime

ACIRprime DCIRprimeneACRedDCRed

ACIRDCIR

(7)

3 ExperimentsWe refer to the sensor described in Figure 1(b) as theconventional sensor and the sensor described in Figure 1(c)as the designed sensor e optical sensor moduleMAX30100 (Maxim Integrated Products Inc San Jose CA)was employed for both sensors [8] It combines two LEDswith a typical wavelength of 880 nm and 660 nm photo-detector optimized optics and 16-bit sigma-delta analog-to-digital converter In addition it has function for ambientlight cancellation and adjustable digital lter to reject powerinterference and ambient noise In order to control theMAX30100 system including LED drive circuit we used theCortex M4 STM32F401 (STMicroelectronics Dallas TX

USA) which oers the balance of dynamic power con-sumption and processing performance e only dierencewas that the designed sensor shown in Figure 1(c) employeda barrier structure for physically separating the LED andphotodetector on the cover glass e experimental ow-chart is shown in Figure 2 Two desaturation experimentswere conducted for each sensor (1) calibration testing and(2) validation testing For each subject and each experimenta series of desaturation runs were performed as shown inFigure 3 Each run started with a stabilized period at roomair and is followed by stabilized plateaus at various lowersaturation levels between 60 and 100 e following werethe value targets Run1mdash92 87 82 77 and 72 andRun2mdash93 88 83 78 and 73e objective of these

LED PD

(a)

LED PD

(b)

LED PD

(c)

Figure 1 Examples of reectance-type pulse oximetry sensors

Journal of Healthcare Engineering 3

targets was to spread the data points evenly over the desiredrange Achieving them exactly was not important oughevery eort was made to be within 2 A plateau was denedas stable when the readings of the reference pulse oximetermanufactured by Nellcor oximeter (N-550 Nellcor PuritanBennett Inc Pleasanton CA) have not changed more than1 for 10 seconds

For each subject this allows 24 stabilized plateaus to betaken when initial room air saturation is considered (seriesof desaturation runs Run1-Run1 and Run2-Run2) ere-fore for example 10 subjects produce 240 stabilized pla-teaus Reducing oxygen saturation level was performed bya skilled operator at Shenzhen University China byadjusting the inspired air-nitrogen-CO2 mixture for theSaO2 response predicted from the oxyhemoglobin dissoci-ation curve obtained by end tidal gas analysis [11] Acommon sense is that dark skinned subjects are known tohave a palm area that is not darker than the skin color ofother parts of the body But even so recent studies haveshown that they have a darker color than the palm colors ofAsian and Caucasian [12] Based on this we measured thePPG signals from the nger to investigate the eect of opticalcrosstalk on the pulse oximeter according to skin colorBefore starting the experiment we attached a sensor to the

index nger of the subject and xed it with adhesive tape sothat it would not move because of over breathing During themeasurement of the data the subjects were placed in a re-laxed semisupine posture with a mouthpiece for breathing ina chair about 30 degrees inclined

31 Calibration Testing e ratio of absorbencies at twowavelengths value (R [ACRedDCRed][ACIRDCIR]) wascalibrated empirically against reference oxygen saturationmeasured by N-550 pulse oximeter manufactured by Nellcor(Nellcor Puritan Bennett Inc Pleasanton CA) in volun-teers To relate the measured values of the ratio R to theoxygen saturation reading of the pulse oximeter the em-pirical calibration curve was derived by a second orderpolynomial SpO2 α + β middot R + c middot R2 e coecopycients α βand c were determined by regression analysis to give thecurve a best t to the reference SpO2 for each sensor type 10subjects were recruited for each calibration test of which 3were African American with dark skin color and the restwere Caucasian or Asian erefore total 240 stabilizedplateaus for each sensor type were used for calibration

32 Validation Testing On the other day the second ex-periment was performed to evaluate the accuracy of the

Subjectgroup 1(N = 10)

R (ratio) data for 240 stabilizedplateau

Desaturation experiments

Day 1

Conventionalsensor


2nd order polynomial fitting

Calibration coefficientα β and γ




Day 2 Day 3

Designedsensor

Conventionalsensor

Designedsensor

SpO2 data for 216 stabilizedplateau


Validation analysisbias correlation RMSE

Calibration testing Validation testing

Figure 2 Flowchart of the experimental procedure


oxygen saturation value obtained through the equation de-rived by calibration testing For the conventional sensorshown in Figure 1(b) total 12 subjects were recruited and 3 ofthem were subjects with dark skin -is produces 216 sta-bilized plateaus for Caucasian or Asian and 72 stabilizedplateaus for African American Also 16 were recruited for thedesigned sensor shown in Figure 1(c) and 4 of them weredark skinned subjects (288 stabilized plateaus for Caucasianor Asian and 96 stabilized plateaus for African American)Comparison of calibrated pulse oximetry with reference pulseoximeter measurements (N-550 Nellcor Puritan Bennett IncPleasanton CA) was reported in terms of the correlationcoefficients and root mean squared error Also BlandndashAltman plot was used to evaluate the discrepancy betweenmeasurements Finally the bias (reference SpO2mdashestimatedSpO2 using calibrated equation)plusmnprecision (standard de-viation) of the oximeters was investigated according to dif-ferent ranges of oxygen saturation

4 Result

Figures 4 and 5 show correlation and BlandndashAltman plotsdescribe the accuracy results for the sensors shown inFigures 1(b) and 1(c) respectively For the sensor shown inFigure 1(b) with crosstalk (conventional sensor) the oxygensaturation estimation error was minus02083plusmn 32405 for all

subjects -e error by race was 08258plusmn 21603 forAsian subjects 08733plusmn 19716 for Caucasian subjects andminus30591plusmn 39925 for African Americans On the other handin the absence of crosstalk (designed sensor) there was nodifference in estimation error according to skin color asshown in Figure 3-e estimation error of oxygen saturationwas minus01587plusmn 25089 for all subjects minus08824plusmn 22859 forAsian subjects 06741plusmn 32822 for Caucasian subjects and09669plusmn 22268 for African American subjects -e corre-lation coefficient between the reference oxygen saturationand the estimated value was 09298 when measured withconventional sensor and 09639 when measured withdesigned sensor for all subjects By subject race the cor-relation coefficient was 08864 for the results of using theconventional sensor (with crosstalk) for African Americansubjects and 094 or higher for all except the case

In Figure 6 the boxplot shows median centerline the 1stand 3rd quartile (box outline) and minimum andmaximumvalues (whiskers) of SpO2 estimation error in differentranges of oxygen saturation -e presence of crosstalk in thereflectance-type pulse oximetry sensor (conventional sen-sor) tended to overestimate overall oxygen saturation insubjects with dark skin and the estimation error was par-ticularly large at low oxygen saturation Table 1 shows thebias between reference and estimated SpO2 according to race(skin color) and sensor type At the SpO2 level of less than

Stabilization

Desaturation run

Recovery

Desaturation run

Recovery

Trial = 2

Trial = 2

Trial = 1

Start

Plateau Take

sample

Figure 3 Schematic flow for the desaturation testing


60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100Es

timat

ed S

pO2 (

)

Correlation coefficient 09298RMSE 32471

AsianCaucasianAfrican

(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash20

ndash15

ndash10

ndash5

0

5

10

15

SpO

2 di

ffere

nce

Mean ndash02083

+196SD 61432

ndash196SD ndash65597


(b)

Figure 4 Accuracy of oxygen saturation estimation in a conventional sensor with optical crosstalk (a) correlation plot (b) BlandndashAltman plot

60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100

Estim

ated

SpO

2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash10

ndash5

0

5

10

15Sp

O2

diffe

renc

e

Mean ndash01587

+196SD 47588



(b)

Figure 5 Accuracy of oxygen saturation estimation in a designed sensor without optical crosstalk (a) correlation plot (b) BlandndashAltman plot

20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20

~70 70~80 80~90 90~100Reference SpO2 ()


(a)

20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20

~70 70~80Reference SpO2 ()

80~90 90~100


(b)

Figure 6 Boxplot of bias for the three dierent subject groups in dierent ranges of oxyhemoglobin saturation (a) Conventional Sensor (b)Designed Sensor


70 the largest error of minus1505plusmn 558 was found in theAfrican American subject group when measured usingconventional sensors Even at 80ndash90 oxygen saturationlevels of African Americans measured with conventionalsensors they showed a greater level of error than in all othercases -ere was no difference in the measurement accuracyof oxygen saturation according to the race in the case of thedesigned sensor which blocked the optical crosstalk

5 Discussion and Conclusion

Since the pulse oximetry technology is already well known andcan be implemented easily it has been recently adopted bygeneral consumer device makers rather than medical devicecompanies to develop mobile healthcare products -ere arehowever few studies on reflectance-type pulse oximetry sen-sors that are used primarily for this purpose Recent publishedwearable pulse oximeter studies have used reflectance-typesensors but most of them have not been tested for subjectswith various skin colors especially dark skin color Lu et alproposed a reflection pulse oximeter embedded in the backcover of a smart handheld device [13] SpO2 was tested in therange of 80ndash100 for 16 subjects but there was no skin colorinformation of the subjects In these prototypes opticalcrosstalk between the back cover and the main board on whichthe LED and PD were mounted was not blocked so it isexpected that errors may occur in dark skin subjects especiallyin SpO2 which is lower than 80 Poh et al proposed anearphone-type reflective PPG measurement system and sug-gested that oxygen saturation can be measured by adding LEDof different wavelength [14] LED and phototransistor in theirprototypes were integrated into a small resin package andmounted on the earbud of the earphone -erefore althoughthe experimental results on oxygen saturation are not pre-sented if the optical barrier design is not carefully illustrated asshown in Figure 1(b) SpO2 measurement error is likely tooccur in dark skin Venema et al also proposed a similar system[15] -ey named it in-ear pulse oximetry because the pro-totype optical sensor was placed at the inner tragus-ey sealedan optical sensor into an ear mold -ey have customized theear mold to the individual but careful design is needed toensure that the surface of the optical sensor is in perfect contactwith the skin surface of the tragus to prevent direct crosstalk-ey showed the result of sudden SpO2 drop due to sleep apneaduring sleep but there was no measurement of dark skinsubject Recently studies have been actively conducted tomeasure PPG signals in a noncontact manner using a web

camera or a smartphone camera [16ndash18] It enables noncontactheart rate monitoring by detecting cardiac pulse induced subtlecolor variation on skin surface Especially in recent years it hasbeen tried to measure not only heart rate but also oxygensaturation using camera [19ndash22] Most studies have beenperformed about low-sampling rate (frame rate) noise due toambient light and light source for metrology and crosstalk innoncontact situations was simply overlooked as one of thenoise sources However it is expected that an error due to skincolor will occur when camera is used for pulse oximetry so it isnecessary to test subjects having various skin colors

In this paper we reviewed the importance of sensordesign to prevent direct crosstalk in reflectance-type pulseoximetry sensor through theoretical analysis and experi-ments According to the theoretical equation of the pulseoximeter expressed by the ratio of AC and DC obtained fromthe PPG signal of two wavelengths the difference of theamount of light absorption depending on the melanin in-dicating the skin color is canceled by normalizing the ACvalue to the DC value of each wavelength In other words itcan be explained that the term including the subscript ldquomrdquo inthe ratio formula of AC and DC shown in Equation (4) iscanceled in the final formula However if the crosstalk occursas in the case of the sensor shown in Figure 1(b) there isa term represented by a subscript ldquocrdquo of different valuesdepending on the color of the skin -erefore the presence ofcrosstalks increases the probability of error in the measure-ment of oxygen saturation depending on the skin colorExperimental results also show that most of the errors inoxygen saturation in conventional sensors suffer from opticalcrosstalks occurred in dark skin subjects In conclusionoptical barrier design to prevent direct light crosstalk is criticalto ensure accuracy regardless of skin color in reflective-typepulse oximeters for mobile healthcare devices

Data Availability

-e data used to support the findings of this study areavailable from the corresponding author upon request

Conflicts of Interest

-e authors declare that they have no conflicts of interest

Acknowledgments

-is work was supported by the Soonchunhyang UniversityResearch Fund

Table 1 Bias reference SpO2 minus estimated SpO2 for conventional and designed sensor in specified range of oxygen saturation

Sensor Subject lt70 70sim80 80sim90 90sim100

Conventional sensorAsian minus008plusmn 224 minus049plusmn 238 139plusmn 193 110plusmn 187

Caucasian minus159plusmn 192 minus045plusmn 185 120plusmn 192 129plusmn 177African minus1505plusmn 558 minus588plusmn 508 minus369plusmn 398 minus157plusmn 225

Designed sensorAsian minus183plusmn 218 minus095plusmn 275 minus078plusmn 216 minus079plusmn 192

Caucasian minus250plusmn 071 074plusmn 381 137plusmn 336 004plusmn 221African minus050plusmn 286 098plusmn 273 134plusmn 192 068plusmn 200

Data are presented as meanplusmn standard deviation


References

[1] J W Severinghaus and Y Honda ldquoHistory of blood gasanalysis VII Pulse oximetryrdquo Journal of Clinical Monigoringvol 3 no 2 pp 135ndash138 1987

[2] M W Wukitsch M T Pettersono D R Tobler andJ A Pologe ldquoPulse oximetry analysis of theory technologyand practicerdquo Journal of Clinical Monitoring vol 4 no 4pp 290ndash301 1988

[3] YMendelson J C Kent B L Yocum andM J Birle ldquoDesignand evaluation of a new reflectance pulse oximeter sensorrdquoMedical Instrumentation vol 22 no 4 pp 167ndash173 1988

[4] N Johnson V A Johnson J Fisher B Jobbings J Bannisterand R J Lilford ldquoFetal monitoring with pulse oximetryrdquoBJOG An International Journal of Obstetrics and Gynaecologyvol 98 no 1 pp 36ndash41 1991

[5] V Konig R Huch and A Huch ldquoReflectance pulseoximetryndashprinciples and obstetric application in the zurichsystemrdquo Journal of Clinical Monitoring and Computingvol 14 no 6 pp 403ndash412 1998

[6] H Lee H Ko and J Lee ldquoReflectance pulse oximetrypractical issues and limitationsrdquo ICT Express vol 2 no 4pp 195ndash198 2016

[7] Y Mendelson and B D Ochs ldquoNoninvasive pulse oximetryutilizing skin reflectace photoplethysmographyrdquo IEEETransactions on Biomedical Engineering vol 35 no 10pp 798ndash805 1988

[8] Maxim Integrated ldquoMAX30100 pulse oximeter and heart-ratesensor IC for wearable healthrdquo 2018 httpsdatasheetsmaximintegratedcomendsMAX30100pdf

[9] W E L Brown and A V Hill ldquo-e oxygen-dissociation curveof blood and its thermodynamical basisrdquo Proceedings of theRoyal Society B Biological Sciences vol 94 no 661pp 297ndash334 1923

[10] H F Zhang K Maslov M Sivaramakrishnan G Stoica andL V Wang ldquoImaging of hemoglobin oxygen saturationvariations in single vessels in vivo using photoacustic mi-croscopyrdquo Applied Physics Letters vol 90 no 5 p 0539012007

[11] P E Bickler J R Feiner and J W Severinghaus ldquoEffects ofskin pigmentation on pulse oximeter accuracy at low satu-rationrdquo Anesthesiology vol 102 no 4 pp 715ndash719 2005

[12] Y Wang M R Luo M Wang K Xiao and M PointerldquoSpectrophotometric measurement of human skin colourrdquoColor Research and Application vol 42 no 6 pp 764ndash7742017

[13] Z Lu X Chen Z Dong Z Zhao and X Zhang ldquoA prototypeof reflection pulse oximeter designed for mobile healthcarerdquoIEEE Journal of Biomedical and Health Informatics vol 20no 5 pp 1309ndash1320 2016

[14] M Poh K Kim A Goessling N Swenson and R PicardldquoCardiovascular monitoring using earphones and a mobiledevicerdquo IEEE Pervasive Computing vol 11 no 4 pp 18ndash262012

[15] B Venema J Schiefer V Blazek N Blanik and S LeonhardtldquoEvaluating innovative in-ear pulse oximetry for unobtsivecardiovascular and pulmonary monitoring during sleeprdquoIEEE Journal of Translational Engineering in Health andMedicine vol 1 p 2700208 2013

[16] W Wang A C den Brinker S Stuikj and G de HaanldquoAlgorithmic principle of remote PPGrdquo IEEE Transactions onBiomedical Engineering vol 64 no 7 pp 1479ndash1491 2017

[17] M Kumar A Veeraraghavan and A Sabharwal ldquoDis-tancePPG robust non-contact vital signs monitoring using

a camerardquo Biomedical Optics Express vol 6 no 5pp 1565ndash1588 2015

[18] T Coppetti A Brauchlin S Muggle et al ldquoAccuracy ofsmartphone apps for heart rate measurementrdquo EuropeanJournal of Preventive Cardiology vol 24 no 12 pp 1287ndash1293 2017

[19] D Shao C Liu F Tsow et al ldquoNoncontact monitoring ofblood oxygen saturation using camera and dual-wavelengthimaging systemrdquo IEEE Transactions on Biomedical Engi-neering vol 65 no 6 pp 1091ndash1098 2016

[20] U Bal ldquoNon-contact estimation of heart rate and oxygensaturation using ambient lightrdquo Biomedical Optics Expressvol 6 no 1 pp 86ndash97 2015

[21] H-Y Tsai K-C Huang H-C Chang and C-H Chang ldquoAstudy on oxygen saturation images constructed from the skintissue of human handrdquo in Proceedings of IEEE InternationalInstrumentation and Measurement Technology Conference(I2MTC) pp 58ndash62 Minneapolis MN USA May 2013

[22] H-Y Tsai K-C Huang H-C Chang J-L A Yeh andC-H Chang ldquoA noncontact skin oxygen-saturation imagingsystem for measuring human tissue oxygen saturationrdquo IEEETransactions on Instrumentation and Measurement vol 63no 11 pp 2620ndash2631 2014


International Journal of

AerospaceEngineeringHindawiwwwhindawicom Volume 2018

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018


Active and Passive Electronic Components

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawiwwwhindawicom

Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Control Scienceand Engineering

Journal of



Journal ofEngineeringVolume 2018

SensorsJournal of



RotatingMachinery


Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation


Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

Submit your manuscripts atwwwhindawicom

type sensor is used [5] In addition since the sensor is wellfixed to the finger in the form of a clip the possibility ofmovement of the sensor and the possibility of entrance ofexternal light are small so that the quality of the signal to bemeasured is much more prevalent On the other hand inthe case of the reflectance-type sensor if the sensor is notfixed with the adhesive the sensor slips over the skinresulting in motion noise Nevertheless as the mobile healthdevices have recently been attracting attention there isa growing demand for pulse oximeters using reflectance-type sensors [6] As a mobile healthcare device modernreflectance-type pulse oximeter has led to changes inoptical sensor configurations Conventional reflectance-type pulse oximetry sensors consist of LEDs with two ormore wavelengths and a photodetector which are spaced4ndash11mm apart [7] As shown in Figure 1(a) an opticalbarrier is required to prevent direct coupling between thelight emitting part and the light receiving part and isapplied in the form of a partition between the LED andthe photodetector in the optical sensor Typically LEDsand photodetectors are packaged and physically sepa-rated using a black rubber material LEDs and photo-detector are exposed on the surface of the sensor orcoated with transparent epoxy In both cases LEDs andphotodetector which are completely separated are indirect contact with the skin As a result the photon ir-radiated from the light emitting part can reach the lightreceiving part only through the human body In recentyears optical sensor technology for pulsation measure-ment has become popular and the LED-photodetectorpair is integrated with the analog front-end makingit miniaturized and becoming a single component Forexample Maximrsquos MAX30100 includes both opticalsensors and measurement modules in a small size of56 times 28 times12 mm [8] Although the sensor module itself iscompletely physically separated from the light emittingunit and the light receiving unit a gap is formed betweenthe sensor module and the device while the module ismounted on the device Also in some cases there is alsocross talk caused by cover grass as described in Figure 1(b)As a result unlike conventional sensors photons directlycoupled to the inside of the sensor or reflected from theskin surface without passing through the human body aremeasured together at the light receiving part of pulseoximetry sensor

In this paper we emphasize the importance of sensordesign to prevent crosstalk when using a small modularreflectance-type pulse oximetry sensor according to recenttrends Specifically we analyzed the effect of opticalcrosstalk on the reflectance-type pulse oximeter accordingto the subjectrsquos skin color through theoretical and clinicalstudies First the BeerndashLambert law the theoreticalbackground of pulse oximetry was analyzed -en desa-turation test was performed to reduce the oxygen satura-tion to 70 by controlling the respiratory gas and the

oxygen saturation measurement results were comparedusing the sensors with and without crosstalk preventionDesaturation testing was approved by the Shenzhen Uni-versity Committee on Human Research and was conductedwith the consent of all subjects

2 Theoretical Formulation

-e principle of the pulse oximeter can be explained bymodified BeerndashLambert law [2 9 10] where I(λ) is de-tected light intensity Io(λ) is the incident light intensityε(λ) is a molar absorption coefficient C is the molarconcentration l(λ) is them mean path length and G(λ) isappropriate factors that account for the measurementgeometry-e signal that records changes in I(λ) caused bypulsatile cardiac activity is called photoplethysmogram(PPG) If we assume that the light absorbing materials aremelanin and other blood or skin-related components theamplitude for the red wavelength PPG signal (AC) can beexpressed as Equation (2) -e subscripts m and b denotemelanin and blood and A0 is a spectrum of all chromo-phores in human skin except m and b Also the baseline(DC) can be written as

A(λ ) lnIo(λ)

I(λ) ε(λ)Cl(λ C) + G(λ) (1)

ACRed I0 λRed( 1113857exp1113858minusεm λRed( 1113857Cmlm λRed( 1113857

minus εb λRed( 1113857Cblbminusdia λRed( 1113857minusA0 λRed( 11138571113859

minus I0 λRed( 1113857exp1113858minusεm λRed( 1113857Cmlm λRed( 1113857

minus εb λRed( 1113857Cblbminussys λRed( 1113857minusA0 λRed( 11138571113859

where A0 λRed( 1113857 A0prime λRed( 1113857 + G λRed( 1113857

(2)

DCRed I0 λRed( 1113857exp1113858minusεm λRed( 1113857Cmlm λRed( 1113857

minus εb λRed( 1113857Cblbminussys λRed( 1113857minusA0 λRed( 11138571113859(3)

-e ratio of the AC to the DC reflects the change inthe maximum blood volume during the systolicperiod (MBVSYS) -e oxygen saturation can be obtainedby using the ratio of MBVSYS for the infrared to thosefor the red since the red light is less absorbed in HbO2 thanin Hb and on the contrary in the case of infraredFrom (2) and (3) MBVSYS for the red and infraredwavelength PPG can be obtained as Equations (4) and (5)respectively where α(λ) minusεm(λ)Cmlm(λ) β(λ)

minusεb(λ)Cblb(λ) and c(λ) minusA0(λ) Finally the ratio for theoxygen saturation calculation can be derived asR [ACRedDCRed][ACIRDCIR] and is linked to theoxygen saturation in a function of SpO2 f(R)


ACRed






(4)

ACIR




ACRedprime



(6)

ACRedprime

DCRedprimeleACRed

DCRed Rprime



ACIRDCIR

(7)



LED PD

(a)

LED PD

(b)

LED PD

(c)











Day 1

Conventionalsensor







Day 2 Day 3

Designedsensor

Conventionalsensor

Designedsensor








4 Result




Stabilization

Desaturation run

Recovery

Desaturation run

Recovery

Trial = 2

Trial = 2

Trial = 1

Start

Plateau Take

sample



60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100Es

timat

ed S

pO2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash20

ndash15

ndash10

ndash5

0

5

10

15

SpO

2 di

ffere

nce

Mean ndash02083

+196SD 61432



(b)


60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100

Estim

ated

SpO

2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash10

ndash5

0

5

10

15Sp

O2

diffe

renc

e

Mean ndash01587

+196SD 47588



(b)


20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20

~70 70~80 80~90 90~100Reference SpO2 ()


(a)

20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20


80~90 90~100


(b)








Data Availability




Acknowledgments










References



























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


ACRed






(4)

ACIR




ACRedprime



(6)

ACRedprime

DCRedprimeleACRed

DCRed Rprime



ACIRDCIR

(7)



LED PD

(a)

LED PD

(b)

LED PD

(c)











Day 1

Conventionalsensor







Day 2 Day 3

Designedsensor

Conventionalsensor

Designedsensor








4 Result




Stabilization

Desaturation run

Recovery

Desaturation run

Recovery

Trial = 2

Trial = 2

Trial = 1

Start

Plateau Take

sample



60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100Es

timat

ed S

pO2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash20

ndash15

ndash10

ndash5

0

5

10

15

SpO

2 di

ffere

nce

Mean ndash02083

+196SD 61432



(b)


60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100

Estim

ated

SpO

2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash10

ndash5

0

5

10

15Sp

O2

diffe

renc

e

Mean ndash01587

+196SD 47588



(b)


20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20

~70 70~80 80~90 90~100Reference SpO2 ()


(a)

20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20


80~90 90~100


(b)








Data Availability




Acknowledgments










References



























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia










Day 1

Conventionalsensor







Day 2 Day 3

Designedsensor

Conventionalsensor

Designedsensor








4 Result




Stabilization

Desaturation run

Recovery

Desaturation run

Recovery

Trial = 2

Trial = 2

Trial = 1

Start

Plateau Take

sample



60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100Es

timat

ed S

pO2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash20

ndash15

ndash10

ndash5

0

5

10

15

SpO

2 di

ffere

nce

Mean ndash02083

+196SD 61432



(b)


60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100

Estim

ated

SpO

2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash10

ndash5

0

5

10

15Sp

O2

diffe

renc

e

Mean ndash01587

+196SD 47588



(b)


20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20

~70 70~80 80~90 90~100Reference SpO2 ()


(a)

20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20


80~90 90~100


(b)








Data Availability




Acknowledgments










References



























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



4 Result




Stabilization

Desaturation run

Recovery

Desaturation run

Recovery

Trial = 2

Trial = 2

Trial = 1

Start

Plateau Take

sample



60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100Es

timat

ed S

pO2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash20

ndash15

ndash10

ndash5

0

5

10

15

SpO

2 di

ffere

nce

Mean ndash02083

+196SD 61432



(b)


60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100

Estim

ated

SpO

2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash10

ndash5

0

5

10

15Sp

O2

diffe

renc

e

Mean ndash01587

+196SD 47588



(b)


20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20

~70 70~80 80~90 90~100Reference SpO2 ()


(a)

20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20


80~90 90~100


(b)








Data Availability




Acknowledgments










References



























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100Es

timat

ed S

pO2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash20

ndash15

ndash10

ndash5

0

5

10

15

SpO

2 di

ffere

nce

Mean ndash02083

+196SD 61432



(b)


60 65 70 75 80 85 90 95 100Reference SpO2 ()

60

70

80

90

100

Estim

ated

SpO

2 (

)



(a)

60 65 70 75 80 85 90 95 100Mean SpO2

ndash10

ndash5

0

5

10

15Sp

O2

diffe

renc

e

Mean ndash01587

+196SD 47588



(b)


20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20

~70 70~80 80~90 90~100Reference SpO2 ()


(a)

20

Bias

(ref

SpO

2 - es

t Sp

O2)

10

0

ndash10

ndash20


80~90 90~100


(b)








Data Availability




Acknowledgments










References



























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia







Data Availability




Acknowledgments










References



























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


References



























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


Documents

The Effect of Optical Crosstalk on Accuracy of Reflectance