BioMed CentralPage 1 of 11(page number not for citation purposes)BioMedical Engineering OnLineOpen AccessResearchA portable near infrared spectroscopy system for bedside monitoring of newborn brainAlper Bozkurt*1, Arye Rosen1, Harel Rosen2 and Banu Onaral1Address: 1School of Biomedical Engineering, Science and Health Systems, Drexel University, 3141 Chestnut Street Philadelphia, Pennsylvania 19104, USA and 2St.Peter's University Hospital, 254 Easton Ave, New Brunswick, New Jersey 08801, USAEmail: Alper Bozkurt* - alper.bozkurt@drexel.edu; Arye Rosen - arye.rosen@drexel.edu; Harel Rosen - rosenha1701@yahoo.com; Banu Onaral - banu.onaral@drexel.edu* Corresponding authorAbstractBackground: Newborns with critical health conditions are monitored in neonatal intensive careunits(NICU).InNICU,oneofthemostimportantproblemsthattheyfaceistheriskofbraininjury.Thereisaneedforcontinuousmonitoringofnewborn'sbrainfunctiontopreventanypotentialbraininjury.Thistypeofmonitoringshouldnotinterferewithintensivecareofthenewborn. Therefore, it should be non-invasive and portable.Methods:Inthispaper,alow-cost,batteryoperated,dualwavelength,continuouswavenearinfrared spectroscopy system for continuous bedside hemodynamic monitoring of neonatal brainispresented.ThesystemhasbeendesignedtooptimizeSNRbyoptimizingthewavelength-multiplexingparameterswithspecialemphasisonsafetyissuesconcerningburninjuries.SNRimprovement by utilizing the entire dynamic range has been satisfied with modifications in analogcircuitry.Results and Conclusion: As a result, a shot-limited SNR of 67 dB has been achieved for 10 Hztemporal resolution. The system can operate more than 30 hours without recharging when an off-the-shelf 1850 mAh-7.2 V battery is used. Laboratory tests with optical phantoms and preliminarydata recorded in NICU demonstrate the potential of the system as a reliable clinical tool to beemployed in the bedside regional monitoring of newborn brain metabolism under intensive care.I. BackgroundWhen light in the near infrared (NIR) range of the spec-trum is shone through the scalp, injected photons followvarious paths inside the head. Some of these photons areabsorbedbydifferentlayersofthetissuesuchasskin,skullandbrain.Othersexittheheadafterfollowingtheso-called "banana" pattern due to scattering effect of thetissue[1].Backscatteredphotonscanbedetectedbymeans of appropriate optical apparatus. When the absorp-tion spectrum of light is analyzed, it is seen that the mainabsorbersintheNIRrangearebloodchromophoresofoxygenatedanddeoxygenatedhemoglobin(HbO2 andHb, respectively). Water and lipid are relatively transpar-enttoNIRlight.Therefore,changesintheamplitudeofbackscattered light can be interpreted as changes in bloodchromophoreconcentrations.Theprocedureofestimat-ing blood chromophore concentrations by means of nearinfraredlightiscalledNearInfraredSpectroscopy(NIRS)[1]. Blood chromophore information can be usedPublished: 29 April 2005BioMedical Engineering OnLine 2005, 4:29 doi:10.1186/1475-925X-4-29Received: 01 March 2005Accepted: 29 April 2005This article is available from: http://www.biomedical-engineering-online.com/content/4/1/29 2005 Bozkurt et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.BioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 2 of 11(page number not for citation purposes)to estimate blood volume and tissue oxygenationwhichare indications of hemodynamic activity.DifferentapproachescanbeusedtoimplementNIRSsuch as time resolved, frequency domain and continuouswavetechniques[2].Amongthesemethods,continuouswave (cw) NIRS is the most practical one, where light withconstantamplitudeisinjectedtotissueandamplitudedecay of the light intensity due to absorption is analyzed.Changes in light amplitude are used to calculate changesinconcentrationsofbloodchromophores.DuetoitspracticalitycwNIRSsystemsallowforbedsideorhomemonitoring of blood chromophores for extended periods.Pulseoximetry,aNIRlightbasedtechniquesimilartoNIRS, is being widely used in current clinical practice. Theaimofpulseoximetryistodetectarterialbloodsatura-tion. CwNIRS further expands the application window ofNIR light by providing information about blood dynam-icsincapillaries.Although,thearterialsaturationobtained via pulse oximeter can only provide global infor-mationabouttheclinicalstateofthepatient,capillaryblooddynamicsstudiedwithcwNIRSiscapableofsup-plying local tissue oxygenation information which can beused for various clinical purposes.One of the important and promising application areas ofcwNIRStechniqueisclinicalneonatology.CwNIRSsys-tems are non-invasive and low-power systems. They allowrealtimemeasurementswithoutremovingnewbornsfrom nursery units, thereby not interfering with intensivecare.Thisisespeciallyimportantforthevulnerablepre-term neonate population. Some clinical situations, whereneonatalcwNIRScanbeemployed,areasphyxia,hypoglycemia,apnea,endotrachealsuctioning,amino-phyllineadministration,indomethacinandexogenoussurfactantadministrations[3].Byspanningsuchalargeapplication space, cwNIRS has a strong potential to be aninevitable part of clinical neonatalogy by monitoring cer-ebralhemodynamicsofpre-andfull-termnewborns.Until now, a number of NIRS systems have been reportedbyvariousgroupstoassessoxygenationrelatedchangesoninfantbrain[4-7].Variousclinicalstudieshavebeenrun by using these systems [8-13].Inthispaper,wepresentanddiscussdesignissuesofaportable, battery-operated, low-cost, low-noise, fast cwN-IRS system which has been designed for bedside cerebralhemodynamic monitoring of newborns in various clinicalstudies. To our knowledge, none of the reported NIRS sys-tems provide these features taken together, which is a cru-cialfactorintransferringthetechnologytoNICUforcontinuousbedsideclinicalmonitoring,particularly,ofregionalbrainmetabolism.Moreover,thesystempro-vides access to raw photon data coming from detectors forfurtheroff-linesignalprocessing.ThereareanumberofopenbiomedicalsignalprocessingquestionssuchasremovingphysiologicalnoiseandmotionartifactsfromNIRSdataaswellashemodynamicdetectionofevokedevents which are still waiting to be explored by biomedi-cal signal processing experts.As a preliminary evaluation, we present pilot results dem-onstratingthatthesystemiscapableofmonitoringchangesinHbO2 andHbconcentrationsinIntralipidsolution and in infant brain.II. MethodsA. Principles of cwNIRS Measurement DeviceThe change in the amount of blood chromophores in thetissue can be predicted by means of Modified Beer Lam-bert Law as explained below [14,15].When the NIR light source and detector are located as infig 1, the detector receives backscattered photons. Some ofthe injected photons are lost as a result of scattering andabsorptionduetodifferentstructuresinthetissue.Theattenuation of light between the source and detector canbe formulated as follows:where Iin is incident light, Iout is the detected light and ODis the optical density for wavelength . Therefore, OD canbe defined as attenuation in intensity of light as a functionof wavelength . This attenuation is the superposition ofabsorption(A)andscattering(S)oflightwithwave-length . Thus,Propagation of photons between source and detector andparameters for Modified Beer Lambert Law Figure 1Propagation of photons between source and detector and parameters for Modified Beer Lambert LawI Iout inOD ( )10 1BioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 3 of 11(page number not for citation purposes)Blood chromophores HbO2 and Hb are the main absorb-ersoflightintheNIRregionofthespectra.Therefore,absorption of light can be formulated as:wherei, isthespecificextinctioncoefficientofbloodchromophore for wavelength , Ci is the concentration ofblood chromophores and L is the pathlength of light at .Pathlengthcanbeexpressedintermsofsourcedetectorseparation as:L = dDPF (4)Where d is the distance between the light source and thedetector and DPF is the differential pathlength factor. Thedifferential pathlength factor is the correction in the meanphoton pathlength for scattering and defined as [16]:wherea, istheabsorptioncoefficientands', isthereducedscatteringcoefficientatwavelength.Here,thereduced scattering coefficient is used instead of scatteringcoefficient due to the anisotropy of the scattering.When scattering, therefore the differential pathlength fac-tor,isassumedtobeconstant,twosuccessivemeasure-ments yield the differential OD value of:where the effect of scattering is cancelled.Sinceeachchromophorehasaspecificextinctioncoeffi-cientanddifferentialpathlengthfactor,measurementwith two wavelengths leads to:where and and are defined as and.Therefore,Equation9providesatransformationfromlightoutputchange to change in blood chromophore concentrations.Byusingbloodchromophoreconcentrationswedefinetwo parameters, namely,OXY = CHbO2 - CHb (10)andBV = CHbO2 + CHb (11)OXYandBVareestimatesproportionalto,respectively,oxygenation and blood volume changes in the tissue dueto hemodynamic activation.Calculation of parametersAbsorption and reduced scattering coefficients cannot bemeasured with cwNIRS method, therefore, are calculatedbyusingthevaluesgivenintheliterature[17,18].Anassumptionof85%saturationand100Mtotalhemo-globinconcentration[19]resultsintheabsorptionandreducedscatteringcoefficientsof.By using these coefficients in equation 5, DPF values arecalculated as 5.0055 and 4.6564 for 730 nm and 850 nm,respectively. These values are consistent with the measure-mentsintheliterature[20].Ontheotherhand,isobtained from the values given in the literature[15] as:By inserting into equation 8, the relationship betweenlight intensity change and blood chromophore concentra-tion is derived as:ODIISoutin + ( ) log attenuation A102A C Li ii Hb HbO ( ) ,,23DPFdsas a

_,

+

1]11123111 31 21 2( ),,/, ,/55 ( ) ( )OD OD OD C dDPFfinal initial i ii Hb HbO , , ,,26 ( ) OD M C 7

1]11

1]11ODODODCCCHbOHb122,M d DPFT ( )( ) . 8 DPF

1]11HbO HbOHb Hb2 1 2 21 2,,, ,DPFDPFDPF

1]111200 ( )C M OD19 anmanmsnmsnm 730 850 730 8500 015 0 023 0 80 0 95 . , . , . , . mm-11

1]10 4383 1 15961 3029 0 7861. .. .mM cm-1 -1

1]11

1]1 CCODHbOHb20 0673 0 12000 0993 0 0404730. .. .nnmnmOD

1]11850BioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 4 of 11(page number not for citation purposes)B. cwNIRS Instrumentationa. Description of system designThe cwNIRS system consists of three parts: probe, controlcircuit and processing unit (fig 2). The probe constitutesthe interface between the control system and the subject.Itholdsthelightsourceanddetectorinanappropriategeometry. Operation of the light source and detectors aremanipulatedbythecontrolcircuitwhichcanbesubdi-vided as transmitter and receiver. Transmitter and receiverarecontrolledbythecomputersoftwareforcoherentdetectionoftwowavelengths.Thecomputeralsostoresanddisplaysreceivedlightinformationafterapplyingnecessary signal processing schemes.The design of the system can be explained through the useof four subtitles: transmitter, probe, receiver and compu-terprocessing.Theroleofeachpartisdisplayedinthedetailed block diagram of the system (fig 3)1. TransmitterThe transmitter part of the control circuit is composed ofanadjustableLEDdriverandwavelengthselector.ThepurposeoftheadjustableLEDdriveristoregulatethelight output of the LED in order to compensate for absorp-tion differences in various tissue types. For subjects withlighterskincolor,sufficientamountofoutputcanbeachievedbyapplyinglowerpowerlightwhereaslargerintensity of light is necessary for darker skin color. This isbecause of the relative amount of melanin in skin. Sincemelaninconcentrationisconstantduringthemeasure-ment, it does not affect the results related to concentrationchange.However,darkerskincolorcausesadecreaseinsignaltonoiseratio(SNR)byabsorbingmorelight.Therefore,thesamelevelofSNRcanbeachievedbyadjusting the intensity of NIR light.Light intensity can be changed by varying the current pass-ingthroughthelightsource.Theuserdeterminestheamountofcurrentandentersitintotheuserinterface.Then,thesoftwareadjuststhecurrentbychangingtheresistance of a digital potentiometer in driver circuit. Therange of current that can pass through the light source isfrom 0 to 100 mA. Values larger than 100 mA have poten-tialtodamagethelightsource;therefore,thesoftwarewarns the user about these values through the user-inter-face and does not initiate the operation. The device alsohas an indicator light on the electronics box which turnsontowarntheuserwhenacurrentlargerthan100mApasses through the light source due to an undesired shortcircuit.As described in the previous section, two different wave-lengthsarerequiredtoresolvetwotypesofbloodchromophores.Weemployedthesetwowavelengthstogetherbymeansoftimemultiplexing.Thisisimple-mentedbyamultiplexerICwhichiscontrolledbythesoftware.Atthesametime,datafromcorrespondingdetector channel were registered to satisfy coherent detec-tion.Themultiplexerwasnotdirectlyconnectedtothelightsource,sincecurrentintheorderof10to100mApasses through the source. Relatively higher RON resistanceof the multiplexer causes a large voltage drop and neces-sary voltage may not be supplied to light source to turn iton. Instead, we connected multiplexer to operate an ana-log switch with a very low resistance value to turn on/offthelightsource.ThispermitstheuseofconventionalSystem containing computer, control box and probe Figure 2System containing computer, control box and probeBlock diagram of the cwNIRS system Figure 3Block diagram of the cwNIRS systemBioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 5 of 11(page number not for citation purposes)camcorder batteries with low voltage values to power thesystem.Duringeachcycle,aftertimemultiplexingtwowave-lengths, we allow an idle period where no wavelength isturned on. The reason for this idle period is that NIR lightsourceisasemiconductorjunctionandheatsupduringthe operation. Experimentally, it has been shown that anidletimehelpslightsourcetocooldowntosaferanges(see system performance part). Moreover, when NIR lightisshonethroughthetissue,detectorreadingsarecom-posed of ambient light penetrating through the tissue andtheoffsetoftheelectroniccomponentsinadditiontoopticalinformationcomingfromthebrain.Theseoffsetvaluescanbedeterminedbyrecordingdetectoroutputwhen NIR wavelengths are turned off. During each cycle,detector reading for idle period is used to correct the men-tioned offset values in order to increase the accuracy of thereadings.Furthermore,detectoroutputduringdarkperiodcanbeusedtomonitortheopticalisolationofdetectors from direct incidence of the ambient light. If off-set values are larger than expected, the probe is reattachedto the skin for better coupling and isolation.2. ReceiverThereceiverparttransferslightinformationobtainedfrom detectors to the computer after amplifying and ana-log filtering. The aim of the amplification part is to bringthe signal level close to the top of the dynamic range of theanalog to digital converter. This operation minimizes theerror that occurs during quantization process. The ampli-fication process is initiated by the user. The user hits a but-tonontheuserinterfaceafterobtainingasatisfactoryprobe-skin coupling during calibration process. The soft-warereadsthelightoutputcomingfromthebrainandadjuststhevalueofadigitalpotentiometerinthegainamplifier such that the signal value is amplified to the fulldynamicrange.AquantizationSNRof75dBisguaran-teedastheresultofclosed-loopgainadjustmentproce-durewherequantizationresolutionis12bit.Thegainvalue is also displayed on the user-interface. Typical pho-ton sampling frequency is 60 kHz and wavelength multi-plexing frequency is 10 Hz.Beforedigitalconversion,thesignalislow-passfilteredwith cutoff frequency of 1.5 kHz to avoid aliasing duringthe sampling process.3. ProbeTheprobeisthemostcriticalpartofthesystemdesignsinceitestablishestheinterfacebetweenelectronicsandthe subject. It holds the light source and two detectors inan appropriate geometry.Detectors are positioned on opposite sides of the sourcewithasource-detectordistanceof2cmasseeninfig4.Thisgeometryallowsmonitoringoftwodifferentloca-tions of the brain, simultaneously. Approximate penetra-tion depth is 1 cm for this configuration [2]. Alternatively,detectorscanbelocatedonthesamesidewithdifferentsource-detector distances for multi-depth measurements.By monitoring different depths, absolute saturation of thebrain tissue can be detected.Inourresearch,wepreferlightemittingdiode(LED)asthe source of the NIR light instead of using laser. Higherlight intensity levels can be utilized with LED, since it is anon-coherentandnon-collimatedlightsource.Powerconsumption of LED is minimal with respect to the lasersource,hencethesystemcanbebatteryoperated.Laserlight source also suffers from extreme heating of the sem-iconductorjunction.Opticalfibersarerequiredtocarrylaser light to tissue.Therefore,moreinstrumentspaceisneededtoprovidecooling.Furthermore,lasersourceneedsadditionalpre-cautions due to its injury potential on the eye. It may notbe appropriate for use in neonatal intensive care units forvulnerable population of newborns. As a result, practicaland portable systems are quite difficult and inappropriateto build with laser source.SinglepackageLEDswhichcontain23wavelengthstogether are available in the market (e.g. Epitex, Inc). Weselected a source where 730 and 850 nm are enclosed in aTO18 package. Crosstalk between these two wavelengthsislesssignificantwithrespecttootherwavelengthpairs[21]. This combination of wavelengths is appropriateforourpurposeofresolvingtwochromophoreThe first prototype of the "flexible" probe holding low powerLED and detectors Figure 4The first prototype of the "flexible" probe holding low power LED and detectorsBioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 6 of 11(page number not for citation purposes)concentrations.LEDcanemitoutputpowerbetween3and17mWwherethetypicalpower usedis9mW.Thespectrum purity of this LED in the NIR region is around 30nm. Although it is wider with respect to diode lasers (56nm), 30 nm gives enough sharpness to resolve two wave-lengthssinceabsorptionspectraisrelativelyflataroundsource wavelengths.Monolithicphotodiode/preamplifierintegratedcircuitshousedinaclear8-pinDIPpackage(OPT101,Burr-Brown,Co.)wereusedtodetectthelightcomingfromthe brain. Active sensing area of this detector is 2.29 2.29mm2. Feedback resistance of 1 Mohm provides a transim-pedance (V/A) gain of 106 and bandwidth more than 100Hz.Themostchallengingaspectofthedesignhasbeenthecoupling of optical elements (optodes) to the scalp of thesubject. We have used flexible circuit technology to fit theoptodestothecurvedsurfaceofthesubject'shead.Thesource and detectors attached to the flexible circuit boardareembeddedintomedicallygradedfoam(fig4).Thispreventsrelativelysharpcornersofoptodestocomeincontact with the scalp of the subject. Moreover, it blockstheambientlightwhichtendstosaturatethephotodiodes.We developed various methods to affix the flexible probeto the surface of the head. The first method is to use dou-blesidedmedicallygradedstickytapewithonesideattached to the foam and other side attached to the scalp(Adchem). This provides a satisfactory and stable optode-scalp coupling. Another method uses a medically gradedself-stickysiliconmaterial(Implantech,Inc)insteadofsticky tape and the foam; optodes can easily be embeddedinthissiliconmaterial.Inyetanothermethod,weusedmedically graded foam again. In this case, stable optode-scalp coupling is provided by inserting the foam under ababy hat as in fig 5. This baby hat is generally utilized inNICU to hold incubator tubes running to newborn's noseand mouth. The performances of these methods were sim-ilar in terms of ease-of-use and coupling efficiency.In order to avoid the effect of ambient light further and toincreasetheSNR,weattachedanNIRfilter(EdmundOptics) with cut-off wavelength of 700 nm to the surfaceofthephotodiode.Thisintroducedaninsulationlayerbetweenphotodiodesandscalp,inadditiontorejectingnon-NIR wavelengths.An 1850 mAh Li-Ion camcorder battery with 7.2 V voltagesupply was used to power the transmitter, receiver and theprobe.Achargingsystemwasalsoaddedtotheboxtocharge the battery without taking it out of the box.4. Processing UnitA data acquisition card with 10 V dynamic range and 12-bitresolution(DAQcard1200,NationalInstruments)hasbeenusedtoconvertanaloginformationtodigitalwheresamplingratesupto100kHzcanbeachieved.5digitaloutputand2analoginputchannelsofthedataacquisition card have been sufficient for the entire opera-tion of the system.Thesystemhardware issoftware (visual C++)operated:thecustomsoftwaremanagestimemultiplexingopera-tionbetweensourceanddetectorpairs,thecoherentdetection from multiple sensors and incorporation of realtimedigital signal processingalgorithms. The user inter-face software displays changes in blood volume, oxygena-tionand,concentrationsofHbandHbO2 byimplementing Beer-Lambert Law in real time. In order tomonitor optode-scalp coupling during the operation, theuser interface also displays raw voltage data coming fromcontrolbox.Especially,theoffsetvaluereadduringtheidle period of the light source is an important parameterdemonstrating the ambient light leakage. Data is recordedin a file for further off-line processing.Typicaltimemultiplexed voltage output ofthedetectorscan be represented as seen in fig 6. Each square pulse cor-respondstodetectedphotonswhenaparticularwavelength is turned on. This signal can be interpreted asalow-passsignalmultipliedbyatrainofsquarepulses,which corresponds to convolution with sinc functions infrequency domain. Thus, processing of time multiplexedsignal is not straightforward. However, we can get rid oftimemultiplexingschemebyapplyingfiniteimpulseBaby hat method used for optode-scalp coupling (see"Acknowledgment") Figure 5Baby hat method used for optode-scalp coupling (see "Acknowledgment")BioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 7 of 11(page number not for citation purposes)response(FIR)filteringfollowedbydecimation.AsweconvolveFIRfilterwithdetectoroutput,weonlyrecordtheresultoftheconvolutionwhenallFIRfiltercoeffi-cients coincides with the samples of the square pulse as infig 6. To have a sharper cut-off, filter order is selected to beequaltothenumberofsamplesthateachsquarepulsecontains. We designed an FIR filter with 10 Hz cut-off fre-quency by using Dolph-Chebyshev window function witharipplefactorof40.TheoperationofFIRfilteringanddecimation are applied real time by the control software.After decimation process, an additional FIR low-pass filteris applied to the signal to clean various artifacts. The NIRSsignalthatcorrespondstohemodynamicactivityofthebrainhasaveryslowtimecourse[15].Theexpectedfre-quencyrangeofthehemodynamicactivityis0.060.16Hzinterval.Ontheotherhand,hemodynamicactivitysignaliscontaminatedbyotherfactorssuchasrespira-tion, arterial pulsations and motion artifacts. Fortunately,alltheseartifactsareinhigherfrequencyrangeswithrespecttohemodynamicactivity.AnonlineFIRfilterisused to clean the frequencies above 0.16 Hz which is outof region of interest. Unfiltered data is recorded in parallelto filtered data for further offline-processing.III. ResultsA. System PerformanceThe performance of the device was evaluated by means ofinitial laboratory tests on phantom. First, shot limited sig-nal to noise ratio (SNR) was measured. Shot noise is dueto the random arrival of photons in the photodiode. Sincethe cause of shot noise is the quantum nature of light, it isimpossible to remove it. To assess the shot noise limitedSNR,weusedanopticalphantom(a=0.01ands'=1.00 mm-1), which is free of motion artifact and physio-logical noise.We measured SNR for various values of time multiplexingparameters:dutyratioandpulserepetitionrates.SNRdepends on these factors in two ways. First, duty cycle andpulserepetitionratedeterminethenumberofcollecteddata points. More number of data points for each pulse issupposed to provide an increased SNR since the perform-ance of the FIR filter is improved. On the other hand, dutycycleandpulserepetitionrateaffecttheheatingofthesemiconductor junction of the LED. Longer duty cycle andpulse repetition rate cause an elevation in the temperatureof the semiconductor junction. Temperature increase addsan extra noise to the generated light output, thereby, sig-nificantly limiting the SNR. Furthermore, this is an unde-sired condition since elevated temperature is a risk for thesafety of the device on subject population. Therefore, weface a trade-off between filter performance and tempera-ture rise of the semiconductor junction for duty ratio andpulse repetition rate to determine SNR.The experimental results of temperature increase and SNRvalues for different duty ratios and pulse repetition ratesare presented in fig 7 and 8. 60 kHz of sampling rate hasbeenusedduringallmeasurementsandtemperaturevalue at the 30th minute of the operation has been takenwhentemperatureasymptoticallyconvergestoastablevalue.Increasesinbothpulserepetitionrateanddutycycle caused a temperature increase in the semiconductorjunction.TheexpectedriseinSNRduetoincreasednumberofsamplesissuppressedbyextranoiseinthelight output of the LED as result of temperature increase.So, overall SNR of the device decreased.As a result of this analysis, operation duty cycle and pulserepetitionratewereselectedtobe12.5%and10Hz,respectively.ThisresultsinarelativelysmallamountoftemperatureincreasewhereasaSNRof67dBwasSummary of the filtering operation Figure 6Summary of the filtering operationTemperature increase & SNR vs. pulse repetition rate (dutyratio = 12.5%) Figure 7Temperature increase & SNR vs. pulse repetition rate (duty ratio = 12.5%)BioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 8 of 11(page number not for citation purposes)obtained. Such high values of SNR could be achieved as aresult of powering the entire system including the controlbox and processing unit by battery. An off-the-shelf 1850mAhLithium-Ionbatteryoperatesthesystemformorethan 30 hours.Temperatureincreaseisanimportantfactortoevaluatethe safety of the system. NIR light is a non-ionizing typeoflightanditsonlyknownpotentialhazardisthermalinjury of the skin. There are two factors that contribute tothe temperature increase of the skin: radiated heat due toNIR light absorption by the skin and conducted heat duetoheatingofsemiconductorjunctionsinsidetheLED,since LED is attached to the skin. Our previous study[22]demonstratesthattemperatureincreaseduetoradiatedheatisaround0.51C,whereas,temperatureincreaseduetoconductedheatcanbeashighas10C.Ontheotherhand,theheatingofthesemiconductorjunctioncanbecontrolledbyadjustingthepulserepetitionrateand duty cycle, therefore, the effect of conductive heat canbeminimized.Whenaboveproposedpulserepetitionrateanddutycycleareused,thecombinedtemperatureincrease of the skin is around 1.5C and has no harmfuleffect to the subject under normal conditions. Therefore,theriskofthesystemforanypotentialburninjuryisminimal.Straylightrejectionhasfoundtobelargerthan99%under normal illumination levels. This has been increasedeven more with the use of NIR filter. Inter-channel cross-talk between detectors was measured to be less than 0.1%.B. Preliminary Evaluationa. Liquid phantom experimentA dynamic liquid phantom simulating the optical proper-ties of the tissue was used in order to test the efficacy of thesystem. The basis of the phantom was formed by a scatter-ingsolutionofIntralipid(LiposynIII)andphosphatebufferedsalinewithpH=7.4.27.2mlofintralipidwasadded to 1000 ml of water in order to obtain an overallreduced scattering coefficient of 0.8 mm-1 (730 nm). Thesolutionwasplacedinacylindricalbeaker.Amagneticstirring rod was also placed in the beaker and the phan-tom was stirred during the course of the experiment. Afterthebasiswasformed,redbloodcellsobtainedfromhealthy human blood were added to scattering solution toachieve a volume fraction of 1.5% and total hemoglobinconcentration of 26 M. This is a typical value for normalphysiologicalconditionswithanassumptionof4.0%blood volume and 40% hematocrit. The hemoglobin sat-uration was measured to be 85%.In order to induce deoxygenation on the liquid phantom,4 g of dry Bakers yeast was added to the solution. The tem-perature of the phantom was maintained at 37C to keeptheyeastactive.Thisconditionwasmaintainedoveracourse of about 2030 min. Deoxygenation was observeduntilhemoglobinsaturationreachedasteadystateat25%.Afterdeoxygenationofyeast-intralipidsolutionreached steady state, oxygenation was simulated again bydeliveringextraoxygentothephantomfromanoxygentank. Oxygen supply was maintained until a steady statelevelofoxygenationwasobtained.Steadystatehemo-globin saturation was 89%.During this procedure, two sets of data were collected byattaching the cwNIRS probe to the side of the beaker. Thefirst measurement was performed to observe the effect ofHbO2 andHbadditiontotheintralipidsolution.Sincetherewerenohemoglobinmoleculesintheintralipidsolution,thismeasurementprovidedtheabsolutecon-centrations of added HbO2 and Hb as seen in figure 9. Inthis figure, the ratio of HbO2 and Hb is consistent with themeasured saturation value of 85%.In the second measurement, our aim was to observe deox-ygenation as a result of yeast addition. Here, the baselinewas taken as the steady state intralipid-hemoglobin solu-tionwithoutyeast.YeastactivationrelatedchangesinHbO2 and Hb concentrations can be seen in figure 10 inaddition to BV and OXY curves. When yeast was added tothe solution, deoxygenation was due to the consumptionof oxygen byyeast which reachedsteady state within 15minutes.Steadystatevaluesareconsistentwiththemeasured saturation of 25%. Re-oxygenation, when extraoxygenwassupplied,isdisplayedinthesamefigure.During the procedure of deoxygenation and oxygenation,as expected, blood volume did not change significantly.b. Monitoring of newborn subjectsThevalidityofthesystemwastestedonanewborninNICUofSt.Peter'sUniversityHospital(NJ)byTemperature increase & SNR vs. duty ratio (pulse repetitionrate = 10 Hz) Figure 8Temperature increase & SNR vs. duty ratio (pulse repetition rate = 10 Hz)BioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 9 of 11(page number not for citation purposes)monitoringoxygenatedanddeoxygenatedhemoglobinconcentrationsonthetemporalregionofthebrainasaresponse to a standard auditory brainstem response (ABR)basedhearingscreeningtest.Ahearingscreeningtest,which is mandatory by law in many states, is applied tonewborns to screenfor hearing problems. In ourexperi-mental protocol, we used hearing screening test to createa controlled auditory stimulation in the temporal regionoftheinfantbrain.Localtemporalactivationhasbeenobserved as a response to auditory stimuli in fMRI studiesin the literature[23].All studies were carried out under an IRB approved proto-col and informed consent was obtained prior to studies.In this study, a five day old 37 week estimated gestationalage male infant underwent an ABR based hearing screen-ing test while he was being monitored by the cwNIRS sys-tem.InanABRbasedhearingscreeningtest,EEGelectrodes are attached with adhesive to the baby's scalp.Whilethebabysleeps,clickingsoundsareappliedthrough tiny earphones in the baby's ears. The test meas-ures the brain's electrical activity in response to the soundsand displays the result as "passed" or "failed". The nonin-vasive procedure takes only a few minutes.In our study, the hearing screening test was repeated twice.During the first test, the probe was placed on the temporalregion of the brain which corresponds to the auditory cor-tex. In the second test, the probe was located on the fore-head of the newborn for the control measurement (fig. 5).Baselinedatawerecollectedfor20secondsbeforeeachtest.Afterbaseline,ahearingscreeningdevicefromNatus, Inc. (ALGO 3) was used for auditory stimulationbyapplying34clicksperseconduntiltheinfantpassedthetest.Bothtestslastedapproximately2minutes.TheALGO 3 hearing screening system displayed the results as"passed"forbothtestswhichconfirmedthatwecreatedauditorystimulisuccessfully.Theentireprocedurewasrepeated twice to validate the self-consistency of the data.Averagechangesinthehemoglobinconcentrationsinaddition to OXY and BV signals during stimulation withrespect to baseline period are presented in figure 11. Errorbars indicate the results of two successive experiments. Anincreaseinbothoxygenatedanddeoxygenatedhemo-globin concentrations was observed in the temporal (test)region during auditory stimulation. Observed increase inblood volume was a result of blood rush to the local tis-sue.Ontheotherhand,averagedeoxyhemoglobinChanges in hemoglobin concentration during red blood celladdition Figure 9Changes in hemoglobin concentration during red blood cell additionChanges in hemoglobin concentration (1), OXY and BV val- ues (2) during the yeast test Figure 10Changes in hemoglobin concentration (1), OXY and BV val-ues (2) during the yeast testBioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 10 of 11(page number not for citation purposes)concentrationchangewasbiggerthanoxyhemoglobinconcentrationwhichwasanindicationofoxygencon-sumption as a result of the local activity. In the forehead(control) probe, only small fluctuations around the base-line were recorded which were uncorrelated with the stim-ulation. Observed changes are consistent with other NIRSbased neonatal measurements in the literature [10,11].IV. ConclusionWe present the design of a low-cost, battery operated con-tinuous wave NIRS system that is intended for continuousbedside brain hemodynamic monitoring of newborns inneonatalintensivecareunits.Suchasystemcanpoten-tiallyassistcliniciansinassessingfunctionalchangesincerebraloxygenationandbloodvolumewithoutremov-ing the baby from the NICU.Design parameters were defined and optimized for a safeand effective system performance. The most critical designissuewasthetrade-offbetweenthetemperatureoftheLED and the performance of the digital filter. SNR of 67dB was obtained for a temporal resolution of 10 Hz. Thesystem can be operated for 30 hours with an off-the-shelf1850 mAh Li-Ion battery. Stray light rejection was satisfac-tory and inter-channel crosstalk of the channels was lessthan 0.1%.Preliminaryexperimentsperformedbothinthelabora-tory and in a clinical setting suggest that the system can beusedtotrackfunctionalchangesofbloodvolumeandoxygenation. This system can be used for bedside moni-toringofneonatesundergoingvariousclinicalstudiessuch as apnea, asphyxia, hypoglycemia, endotracheal suc-tioning,surfactant,aminophyllineandindomethacinadministrations.Authors' contributionsABandHRcarriedoutthein-vivoandin-vitroexperi-mentsandABdraftedthemanuscript.BOandARcon-ceivedofthestudy,andparticipatedinitsdesignandcoordination.AllauthorsreadandapprovedthefinalmanuscriptAcknowledgementsWe acknowledge, with thanks, Dr. Kambiz Pourrezaei, Frank Kepics, and Gunay Yurtsever for their helpful discussions on technical issues. We also extend our appreciations to Dr. Britton Chance and his team for allowing us to use their laboratory equipment and for helpful comments on the man-uscript. Written and signed consent was obtained from parents of the new-born to publish figure 5 as a part of our article. This work has been sponsored in part by funds from the Defense Advanced Research Projects Agency (DARPA) Augmented Cognition Program and the Office of Naval Research (ONR), under agreement numbers N00014-02-1-0524 and N00014-01-1-0986.References1. Villringer A, Chance B:Non-invasive optical spectroscopy andimaging of human brain function. Trends in Neurosciences 1997,20(10):435-42.2. StrangmanG,BoasDA,SuttonJP:Non-invasiveneuroimagingusing near-infrared light. Biological Psychiatry 2002, 52(7):679-93.3. Skov L, Brun NC, Greisen G: Neonatal Intensive Care: ObviousYet Difficult Area for Cerebral Near Infrared Spectroscopy.Journal of Biomedical Optics 1997, 2(1):7-14.4. Wyatt JS, Cope M, Delpy DT, Wray S, Reynolds EOR: Quantifica-tionofcerebraloxygenationandhaemodynamicsinsicknewborn infants by near infrared spectrophotometry. Lancet1986, 8515:1063-1066.5. Chance B, Anday E, Nioka S, Zhou S, Hong L, Worden K, Li C, MurrayT, Ovetsky Y, Pidikiti D, Thomas R: A novel method for fast imag-ing of brain function, non-invasively, with light. Optics Express1998, 2:411-423.6. Siegel AM, Marota JJA, Boas DA: Design and evaluation of a con-tinuous-wavediffuseopticaltomographysystem.OpticsExpress 1999, 4:287-298.7. Schmidt FEW, Fry ME, Hillman EMC, Hebden JC, Delpy DT: A 32-channeltime-resolvedinstrumentformedicalopticaltomography. Review of Scientific Instruments 2000, 71:256-265.8. du Plessis AJ: Near-infrared spectroscopy for the in vivo studyof cerebral hemodynamics and oxygenation. Current Opinion inPediatrics 1995, 7:632-639.9. MeekJH,FirbankM,ElwellCE,AtkinsonJ,BraddickO,WyattJS:Regionalhemodynamicresponsestovisualstimulationinawake infants. Pediatric Research 1998, 43:840-843.10. Sakatani K, Chen S, Lichty W, Zuo H, Wang Y: Cerebral blood oxy-genationchangesinducedbyauditorystimulationinnew-borninfantsmeasuredbynearinfraredspectroscopy.EarlyHuman Development 1999, 55:229-236.11. Zaramella P, Freato F, Amigoni A, Salvadori S, Marangoni P, Suppjei A,Schiavo B, Chiandetti L: Brain auditory activation measured bynear-infraredspectroscopy(NIRS)inneonates.PediatricResearch 2001, 49:213-219.12. TagaG,AsakawaK,HirasawaK,KonishiY:Hemodynamicresponses to visual stimulation in occipital and frontal cortexof newborn infants: a near-infrared optical topography study.Early Hum Dev 2003, 75(Suppl):S203-10.13. Hintz SR, Benaron DA, Siegel AM, Zourabian A, Stevenson DK, BoasDA: Bedside functional imaging of the premature infant brainduringpassivemotoractivation.JournalofPerinatalMedicine2001, 29:335-343.14. Cope M, Delpy DT: System for long-term measurement of cer-ebral blood flow and tissue oxygenation on newborn infantsbyinfraredtransillumination.MedBiolEngComput1998,26:289-294.15. Cope M: The Development of a Near-Infrared SpectroscopySystemandItsApplicationforNoninvasiveMonitoringofAverage changes in hemodynamic parameters in response tohearing screening test Figure 11Average changes in hemodynamic parameters in response to hearing screening testPublish with BioMed Centraland every scientist can read your workfree of charge"BioMed Central will bethe mostsignificantdevelopment for disseminating the results ofbiomedical research in our lifetime."Sir Paul Nurse, Cancer Research UKYour research papers will be:available free of charge to the entire biomedical communitypeer reviewed and published immediately upon acceptancecited in PubMed and archived on PubMed Central yours you keep the copyrightSubmit your manuscript here:http://www.biomedcentral.com/info/publishing_adv.aspBioMedcentralBioMedical Engineering OnLine 2005, 4:29 http://www.biomedical-engineering-online.com/content/4/1/29Page 11 of 11(page number not for citation purposes)CerebralBloodandTissueOxygenationintheNewbornInfant. University College London, London; 1991. 16. Boas DA, Franceschini MA, Dunn AK, Strangman G: "NoninvasiveImaging of Cerebral Activation with Diffuse Optical Tomog-raphy.". In In Vivo Optical Imaging of Brain Function Edited by: FrostigRD. Boca Raton: CRC Press; 2002:193-221. 17. Wray S, Cope M, Delpy DT, Wyatt JS, Reynolds EO: Characteriza-tion of the near infrared absorption spectra of cytochromeaa3 and haemoglobin for the non-invasive monitoring of cer-ebral oxygenation. Biochim Biophys Acta 1988, 933(1):184-92.18. Van der Zee P, Cope M, Arridge SR, et al.: Experimentally meas-ured optical pathlengths for the adult's head, calf and fore-armandtheheadofthenewborninfantasafunctionofinteroptode spacing. Adv Exp Med Biol 1992, 316:143-153.19. Hazinski MF: Pediatric Evaluation and Monitoring Considera-tions.InHemodynamicMonitoring:InvasiveandNoninvasiveClinicalApplicationEditedby:DarovicGO.Philadelphia,PA:W.B.Saunders;2002:471-514. 20. BenaronDA,KurthCD,StevenJM,Delivoria-PapadopoulosM,Chance B: Transcranial optical path length in infants by near-infraredphase-shiftspectroscopy.JClinMonit1995,11(2):109-17.21. Yamashita Y, Maki A, Koizumi H: Wavelength dependence of theprecisionofnoninvasiveopticalmeasurementofoxy-,deoxy-,andtotal-hemoglobinconcentration.MedPhys2001,28(6):1108-14.22. Bozkurt A, Onaral B: Safety assessment of near infrared lightemitting diodes for diffuse optical measurements. BioMedicalEngineering OnLine 2004, 3:9.23. AndersonAW,MaroisR,ColsonER,PetersonBS,DuncanCC,Ehrenkranz RA, Schneider KC, Gore JC, Ment LR: Neonatal audi-toryactivationdetectedbyfunctionalmagneticresonanceimaging. Magn Reson Imaging 2001, 19(1):1-5.