DST_Masimo signal extraction pulse oximetry

MASIMO SIGNAL EXTRACTION PULSE OXIMETRY JulianM.Goldman,MD,1M ichaelT.Petterson,RRT,1 Robert J.Kopotic,MSN,RN,RRT,1andSteven J.Barker,PhD,MD2From the1Masimo Corporation,2852Kelvin Avenue,Irvine,CA 92614;2Department of Anesthesiology,University of Arizona School of Medicine,1501N.Campbell/PO Box245114,Tucson,AZ85724, U.S.A.Received Jul15,1999.Accepted for publication Jan13,2000. Address Correspondence to Julian M.Goldman,MD,Masimo Corporation,2852Kelvin Avenue,Irvine,CA,92614,U.S.A.E-mail:JGoldman@ Goldman JM,Petterson MT,Kopotic RJ,Barker SJ.Masimo signal extraction pulse oximetry.J Clin Monit2000;16:475^483 ABSTRACT.Objective.To describe a new p lse oximetry technology and measurement paradigm developed by Masimo Corporation.Introduction.Patient motion,poor tissue perfusion,excessive ambient light,and electrosurgical unit interference reduce conventional pulse oximeter(CPO) measurement integrity.Patient motion frequently generates erroneous pulse oximetry values for saturation and pulse rate. Motion-induced measurement error is due in part to wide-spread implementation of a theoretical pulse oximetry model which assumes that arterial blood is the only light-absorbing pulsatile component in the optical path.Methods.Masimo Signal Extraction Technology(SETÕ)pulse oximetry begins with conventional red and infrared photoplethysmographic signals,and then employs a constellation of advanced tech-niques including radiofrequency and light-shielded optical sensors,digital signal processing,and adaptive¢ltration,to measure SpO2accurately during challenging clinical condi-tions.In contrast to CPO which calculates O2saturation from the ratio of transmitted pulsatile red and infrared light, Masimo SET pulse oximetry uses a new conceptual model of light absorption for pulse oximetry and employs the discrete saturation transform(DST)to isolate individual``saturation components''in the optical pathway.Typically,when the tissue under analysis is stationary,only the single saturation component produced by pulsatile arterial blood is present. In contrast,during patient motion,movement of non-arterial components(for example,venous blood)can be identi¢ed as additional saturation components(with a lower O2satura-tion).When conditions of the Masimo model are met,the saturation component corresponding to the highest O2satu-ration is reported by the instrument as SpO2.Conclusion. The technological strategies implemented in Masimo SET pulse oximetry e¡ectively permit continuous monitoring of SpO2during challenging clinical conditions of motion and poor tissue perfusion.KEY WORDS.Pulse oximetry;motion artifact;oximetry/instru-mentation/methods;signal processing,computer-assisted; adaptive¢lters;Masimo signal extraction pulse oximetry; signal extraction technology(SETÕ).INTRODUCTIONReview of conventional pulse oximetryConventional pulse oximetry(CPO)determines arterial oxygen saturation(SpO2)and pulse rate(PR)by using a theoretical model which was developed over twenty-¢ve years ago[1].This model assumes that arterial blood is the only light-absorbing pulsatile component in the optical path.The ratio of pulsatile transmitted red(RD) to infrared(IR)light-the``optical density ratio''-is u sed to calcu late SpO2through the use of a calibrationAdditional information about Masimo is available from: Journal of Clinical Monitoring and Computing16:475^483,2000.ß2001KluwerAcademic Publishers.Printed in the Netherlands.equation generated by human in vivo calibration studies[2].A simpli¢ed approach to CPO analysis can be de-scribed as follows:We are interested in the photoplethysmographic variations of the ratio of RD to IR light,but the intensity of transmitted light is unknown.Therefore, the transmitted RD and IR light signals are normalized by dividing the pulsatile transmitted light(known as the AC component),by the non-pulsatile transmitted light(the DC component).This can be expressed as: Normalized light light AC=light DC:The Optical Density Ratio,r,is a proportionality ratio which is unique for each saturation and wavelength of RD and IR light,and is related to SpO2through an empirically derived calibration equation:r normalized RD/normalized IRRD red light($660nm)IR infrared light($905nm):Limitations of conventional pulse oximetry Continuous and accurate measurement of SpO2is di¤-cult in the clinical environment.Excessive ambient light and electromagnetic interference may disrupt the measurement completely[3].Poor tissue perfusion may produce signal strength below the limit of detection, and this condition may be exacerbated in patients with dark pigmentation or with thick digits.Freund et al. showed that once a pulse oximeter failed intraopera-tively,it would not provide data for32%of the mean anesthesia time[4].In a study using computerized anesthesia record keeping systems,pulse oximetry failure periods of10minutes or longer occurred in9%of cases [5].The challenge of obtaining an accurate measure-ment is greater in sicker patients who may exhibit poor perfusion and a resultant poor signal to noise ratio. Moller et al.demonstrated in a study of20,802patients that perioperative pulse oximetry failure rate is positively correlated with the American Society of Anesthesiolo-gists physical status score[6].In this group of patients, the pulse oximetry failure rate was2.5%overall,but increased to7.2%when only physical status4patients were considered.The postanesthesia care unit presents a challenging monitoring environment.In a study of postanesthesia care unit performance,the average fre-quency of pulse oximetry alarms caused by``sensor displacement,motion artifacts,poor perfusion,or a combination of these factors''was once every8minutes and77%of these alarms were false[7].Clearly,pulse oximetry failures are well documented and interfere with the accepted practice standard of continuous assess-ment of oxygenation during general anesthesia and conscience sedation.E¡ect of motion on pulse oximetryThe e¡ect of motion on pulse oximetry accuracy may be more insidious than causing complete failure to measure SpO2or PR dropout.Barker et al.evaluated pulse oximetry accuracy during controlled oxygen desaturation in volunteers while introducing speci¢c arm motions with a mechanical motion generator[8]. During motion,CPO exhibited dropout rates of up to46%,and more ominously,the CPO values were erroneous up to20%of the measurement time. Figure1is illustrative of the comparative performance three oximeters during adult volunteer testing in one of the author's laboratories(Barker).This test was per-formed with IRB approval to characterize pulse oxime-try performance.An adult female breathing room air had three oximeter sensors applied to the motion(test) hand,and a control oximeter(Nellcor N-200,Nellcor PB,Pleasanton,CA)placed on the stationary hand. Saturation data was collected continuously with acom-Fig.1.Room air pulse oximeter motiontest,(A)before initiating motion,(B)20seconds after start of motion.Control pulse oximeter (N-200)was placed on the stationary hand.Pulse oximeters on motion hand:Masimo SET;Hewlett-Packard M1175A;Spacelabs 90308.476Journal of Clinical Monitoring and Computing Vol16No72000puter-based data acquisition system and speci¢c periods were identi¢ed for analysis:shortly prior to initiating a controlled``tapping''motion at3Hz(``A'')and20 seconds after initiating motion(``B'').(See Table1for oxygen saturation data.)In the example above,during motion the Masimo (Masimo Corporation,Irvine,CA)pulse oximeter con-tinued to correlate closely with the stationary reference oximeter,while the other two oximeters displayed saturation values23%and13%lower than the stationary reference unit.It is clear from this data and published reports that CPO can read falsely low saturation values during motion.What causes this measurement error?In the past this phenomenon has been ascribed to excessive uncharacterized``noise''(motion)overwhelming the ``signal''(arterial pulsations)and driving the optical density ratio(r)toward1[2].A ratio of1is reported as a saturation of approximately82%.However,this pro-posed mechanism does not explain why the displayed saturation may be substantially below82%-especially in poorly perfused patients[9].Accurate saturation measurement during motionWhile investigating techniques to improve the per-formance of pulse oximetry,researchers at Masimo Corporation realized that CPO failed during motion because the theoretical perfusion model upon which CPO is based is invalid during motion.During motion, venous blood and other non-arterial absorptive sub-stances generate a pulsatile optical signal as they move and attenuate the transmitted light.A fundamental discovery was that ratio of RD and IR is a composite of arterial and``non-arterial components''.During motion,these non-arterial saturation components con-tribute their own pulsatile signals which cannot be separated from the arterial signal by CPO.Theoreti-cally,the decrease in displayed SpO2will depend upon the oxygen saturation and relative contribution of these components.Engineering data strongly suggests that the primary non-arterial component is venous blood in the optical path,so as a matter of convenience and brevity we refer to this component as``venous''^ although this has not been proven conclusively.The Masimo pulse oximetry modelThe pulse oximetry model which Masimo developed accounts for saturation values contributed by the true arterial signal,and by one or more motion or noise signals.It is assumed that under conditions of motion, the detected IR and RD signals comprise both the true arterial saturation signal and a venous(or non-arterial) motion noise signal.Masimo pulse oximetry model:IR S M 1 RD r aÂS r vÂM ; 2 where,IR is(normalized)pulsatile infrared signal,RD is(normalized)pulsatile red signal,S is IR signal vector from pulsatile arterial blood,M is IR motion signal vector generated by venous(or non-arterial)pulsation, r a is optical density ratio corresponding to arterial saturation(this gives us SpO2),and r v is optical density ratio corresponding to venous(or non-arterial)satura-tion.Masimo's approach to determining SpO2does not depend upon calculating the ratio of RD to IR light. Instead,the model expressed in equations1and2are manipulated to compute a reference signal(RS)which is then used by an adaptive¢lter(AF)and Discrete Saturation Transform(DST)to identify the correct arterial saturation.Note that the RD signal is described by its relationship to the S and M components of the IR signal by the optical density ratio values r a and r v.Computation of the reference signalThe RS is a vector computed as follows:RS r HÂIRÀRD(this is a definition of RSdeveloped by Masimo).All r a values corresponding to a saturation range of 0^100%are substituted for r H during the iterative deter-Table1.Displayed oxygen saturation values during room-air motion experiment.Three pulse oximeters were placed on the test(motion) hand,and reference saturation was obtained from the stationary handOximeter Stationary SpO2(``A''in Figure1)Motion SpO2 (``B''in Figure1)Control(non-motionhand)9898Masimo SETÕ9797Space Labs9875Hewlett-Packard9785First column^data displayed prior to initiation of motion.Second column^data displayed20seconds after initiation of motion.Goldman et al:Masimo Pulse Oximetry477道客巴巴mination of SpO 2.(The values of r H which are equal to r a and r v are not known until the iterative process is completed.)The RS is calculated as described above and used as one input to the AF.The discrete saturation transform (DST)The RS and IR signals are fed into an AF which identi¢es and removes (cancels)frequency components which are in common between both signals.The output of the AF is u sed to bu ild a power cu rve which identi¢es the arterial saturation and other saturation components.The calculations which generate the power curve from the RS and IR signals are known collectively as the DST (Figure 2).Varying r H is a technique to determine the value of r a .(Once we know r a ,SpO 2can be derived from the calibration Equation.)We cannot determine which value of r H is equal to r a until after r H has been swept through the range of possible values of r a which corre-spond to a saturation range of 0^100%.(Table 2sum-marizes the steps involved in performing the DST.)In order to facilitate a conceptual understanding of the operation of the DST,the RS Equation will be presented in a di¡erent form:Once again,the RS is de¢ned as r H ÂIR ÀRD :Sub-stituting Equations (1)and (2)above:RS r H ÂS M À r a ÂS r v ÂM .Reorganize:RSr H ÂS Àr a ÂS r H ÂM Àr v ÂM :RS r H Àr a ÂS r H Àr v ÂM :3This form of the RS equation illustrates the contribu-tion of S (desired arterial signal)and M (motion com-ponent)to the generation of the RS.It will also reveal the changes to RS when r H is equal to r a or r v .The operation of the DST can be clari¢ed with a few examples.In these ¢ve Examples the output of the RS and DST power curve will be examined as we test correct and incorrect values for r H under motion and no motionconditions.Fig.2.Block diagram of discrete saturation transform.The adaptive ¢lter generates the DST power curve as trial saturations are stepped through a range of 0to 100%.Figure depicts output when the trial SpO 2=95%.Table 2.Summary of discrete saturation transformDiscrete saturation transform processCollect block of red and infrared signalsCalculate the reference signal for a trial saturation value (starting,for example at 0%)Send the infrared signal and the reference signal to the adaptive ¢lterThe adaptive ¢lter removes frequency content in common between the IR and reference signalsRepeat the process for the full range of saturation values (0^100%)Identify the DST power peak which corresponds to the SpO 2478Journal of Clinical Monitoring and Computing Vol 16No 72000Example 1:No motion.We are stepping through a range of r H values,and have arrived at r H which represents the correct arterial ratio (r a ).r H r a :M =0(0vector)because there is no motion.Therefore,from Equation (3),RS =(0)ÂS +(r H Àr v )Â0=0.Since RS =0,the AF will not ¢nd any frequency content in common between RS and IR,and the out-put of the AF will be equal to IR.That is,there will be a high power peak plotted on the DST power curve because no energy (M)is being removed from the IR signal at that value of r H (i.e.,at the correct saturation).Example 2:Motion is now present in the signal,and as we sweep through values of r H ,we reach r H =r v .r H r v :Therefore,from Equation (3),RS =(r H Àr a )ÂS +(0)ÂM =(r v Àr a )ÂS.RS is now equal to the arterial signal (S),scaled by (r v Àr a ).The AF will remove the frequencies correlated with the arterial signal from IR,leaving only the motion-related power to be plotted in the DST power curve.This will result in a peak at the saturation corre-lated with the motion (venous)signal (see Figure 3,point A).Example 3:Motion is still present in the signal,and we have arrived at the correct value of r H .r H r a :Therefore,from Equation (3),RS =(0)ÂS +(r H Àr v )ÂM =(r a Àr v )ÂM.Now,RS is equal to the motion signal (M)scaled by (r a Àr v ).The AF will ¢nd and cancel frequency content correlated between RS (that is now equal to M)and the IR signal.The non-canceled frequencies from the IR signal will produce a power peak in the DST (see Figure 3,point B).Example 4:The same conditions exist as in Example 3,but r H is now slightly di¡erent than r a .(This Example explains the fallo¡in power on either side of the r H =r a power peak.)r H %r aRS %0 ÂS r H Àr v ÂM :In this Example,RS primarily contains the motion signal frequency components,but since the value of (r H Àr a )is not equal to zero,some of the arterial signal frequency components are present.The AF will cancel the motion component from the IR signal,and also cancel a small portion of the arterial signal component,leading to an output power that is a little lower than Example 3Figure 3,point C).Example 5We are sweeping through values of r H ,and the DST is evaluating a value of r H which is neither related to the arterial or non-arterial signals.motion is present and r H r a and r H r v :Therefore,the expressions (r H Àr a )and (r H Àr v )are both substantially di¡erent than zero,and the RS will con-tain the frequency content of both the arterial and the motion signals.Consequently,the AF will identify and remove a substantial amount of the correlated frequen-cies from the IR signal,and the DST will reveal the low AF output power Figure 3,point D).In vivo example of the e¡ects of motion on CPO and M asimo SETpulse oximetryFigure 4shows the normalized RD and IR signals from the studied in Figure 1.During this pre-motion period at an SpO 2of 97%,the amplitude of theIRFig.3.Points marked along axis of DSTgraph to explain Examples 2^5,(A)r H =r v ;(B)r H =r a ;(C)r H moving away from peak;(D)r H r a and r H r v .Goldman et al:Masimo Pulse Oximetry 479signal is greater than the RD signal resulting in an optical density ratio less than 1.The DST power curve from the signal in Figure 4is depicted in Figure 5.Note the single narrow power curve spike at 97%.This indicates a narrow distribution of saturation species in this RD and IR signal.After the motion begins (Figure 6),note that the relative amplitude of the and IR plethysmograms have changed.On close visual inspection it appears that the RD/IR ratio has changed,thus re£ecting a decrease in the patient's SpO 2as measured by CPO.However,this patient's SpO 2has not changed ^she is still breath-ing room air.The motion of non-arterial saturation components induced by the motion of the hand is being misinterpreted by the conventional pulse oximeters -which simply analyze the RD/IR ratio ^and report a decrease in SpO 2.Table 1and Figure 1reveal the CPO and Masimo SpO values on the stationary and moving hands during the pre-motion and motion periods.Examination of Figure 7reveals that the DST has correctly identi¢ed the 2during the motion trial.A new peak in the power curve has appeared at a satura-tion of approximately 78%(i.e.,r H =r v ).Based on the analysis of our engineering data,we believe that the primary saturation species which contributes this sec-ond power peak is venous blood which has become ``pulsatile''as a result of motion.In warm healthy subjects at rest who probably have a low peripheral arterial-venous O 2content di¡erence,and therefore have a high peripheral venous O 2saturation (S v O 2),the DST reveals that the S v O 2power peak occurs at a highsaturation.In patients with reduced peripheral blood £ow (such as during hypothermia,hypotension,or shock),there may be an increased digital arterial-venous O 2content di¡erence,and the S v O 2power peak may appear at a much lower saturation on the DST power curve.These same patients also exhibitsubstantiallyFig.4.RD (---)and IR (ö)plethysmograms prior to onset ofmotion.Fig.5.DSTpower curve under no-motion condition.Note easily identi¢ed peak in power curve at SpO 2=97%(con¢rmed by referee instrument on controlhand).Fig.6.RD (---)and IR (öö)plethysmograms 20seconds after onset of motion.The RD to IR peak-to-trough ratios are larger than the pre-motion ratios (depicted in Figure 4)for both the small and large waveforms (equivalent to SpO 2values of 54%and 94%,respectively).CPO techniques may incorporate an average of these individual SpO 2values and report a falsely low SpO 2.480Journal of Clinical Monitoring and Computing Vol 16No 72000lower spurious desaturations during motion when moni-tored with CPO.Digital signal processingDigital signal processing(DSP)refers to the digitiza-tion of analog signals and the subsequent manipulation of these signals on a digital computer[10].Modern dedicated high-speed DSP computer chips are opti-mized to e¤ciently execute DSP algorithms.However, as we learned in the audio¢eld's transition from vinyl LPs(analog processing)to compact digital disks(digital processing),the conversion from analog processing to digital processing does not by itself confer a performance bene¢t.This analogy also applies to pulse oximetry. The DST could be implemented in analog circuitry, but this would be prohibitively expensive.Furthermore, the ready availability of the DSP permits the implemen-tation of sophisticated ambient noise¢ltering schemes. These¢ltering techniques are combined with advanced circuit board design,electromagnetic shielding of theentire signal path,and optical shielding of the sensors, to virtually eliminate interference from ambient light, electrosurgical unit radiofrequency interference,and other sources of environmental signal contamination.Low noise optical probes(LNOPÕ)Masimo SET LNOP oximetry sensors have several unique design features.The recessed photodetector cup, present on all single patient use LNOP sensors,is de-signed to enhance performance by minimizing direct interference from ambient light and stabilizing the detector area on the skin surface.Additionally,the single patient-use sensors are fabricated of a durable substrate which remains clinically viable for substan-tially longer than competitive sensors.Furthermore, swabbing with a standard alcohol wipe can rejuvenate the sensor's adhesive.A sensor longevity study showed average neonatal sensor viability of11.1days[11].Pulse rate detectionMasimo's proprietary pulse rate algorithms may perform independently of SpO2calculation,but use the same principles as the saturation ing Signal Extraction Technology,photoplethysmographic data is analyzed in both the time and frequency domain to extract the best possible information under challenging clinical conditions.Plethysmograms are evaluated to ensure that they meet speci¢c morphologic conditions before being accepted as valid for pulse rate analysis. The pulse rate algorithm is particularly robust,and is capable of reliable bradycardia monitoring in the neo-nate[12,13].CLINICAL IMPLICATIONSThe historical performance shortcomings of CPO have stimulated compensatory strategies to facilitate the clinical implementation of a£awed technology.For example,to address the high rate of false alarms,e¡orts have been directed towards intelligent processing of oximetry alarms[14].Perhaps more signi¢cantly,the sickest patients^those who may be at highest risk for hypoxemia-induced sequelae^may require``tricks'' such as vasodilator therapy or digital nerve blocks to improve pulse oximeter performance[15,16].Typically, clinicians do not have recourse to experimental alarm-reduction software or the time to optimize performance, so continuous patient monitoring is compromised. Continuous intraoperative assessment of oxygenation is an accepted anesthesia practice standard.The relatively high incidence of intraoperative pulse oximetry failures is due,^in part^to poor tissue perfusion caused by hypothermia or hypotension.Poor perfusion may cause low pulsatile signal amplitude which then requires marked signal ampli¢cation to identify SpO2data. Poor perfusion may also produce low tissue blood£ow, Fig.7.DSTpower curve under motion condition.Note presence of two peaks in power curve.The motion of non-arterial blood generates the left-most peak.The peak corresponding to the higher saturation value is the correct SpO2.Goldman et al:Masimo Pulse Oximetry481which results in low peripheral venous oxygen satura-tion.Therefore,motion during low perfusion produces an unfavorable signal-to-noise ratio,which contributes to pulse oximetry failure.The hardware and software features incorporated in Masimo SET facilitate accurate performance in poorly perfused patients[17^19].The importance of continuous oxygenation moni-toring during conscience sedation of pediatric dental patients is readily acknowledged as well.Here too,the limitations of CPO become apparent.In a study of63 pediatric patient dental visits,patient movement was responsible for87^90%of the235desaturation episodes [20].Recently,renewed attention has been directed toward the phenomenon of late postoperative hypo-xemia.Postoperatively,some patients may be at risk for hypoxemia contributing to morbidity and mortality because they are located in reduced surveillance envi-ronments[21].Telemetric pulse oximetry monitoring^ presumably the ideal approach to monitoring these patients^has not gained broad acceptance due to nuisance alarms and failures to monitor e¡ectively.In fact,the poor performance of conventional pulse oxi-metry in critically ill patients has caused some investi-gators to question the utility of CPO monitoring[22]. These same concerns apply to the potential bene¢ts and frustratingly poor performance of CPO for monitoring newborns[23].Several clinical studies have demonstrated Masimo SET pulse oximetry's accuracy and performance bene-¢ts,in particular,a signi¢cant reduction of false alarms [9,24^26].False alarms create a greater problem than mere nuisance interruptions:They may compromise patient safety by desensitizing clinicians to the detection of true patient emergencies[27,28].The frequency of false alarms may be easily reduced by increasing pulse oximetry signal averaging times or increasing the alarm latency period,but these approaches delay hypoxemia detection and may miss rapid,transient,desaturations [29,30].Therefore,it is noteworthy that Masimo SET technology permits a reduction of false alarms without sacri¢cing the detection of true desaturations and pulse rate changes[8,13,24].Patients whose clinical disorders and physical activity confound accurate measurement by CPO are at risk for misdiagnosis and subsequent mismanagement.For ex-ample,artifactually low SpO2values caused by venous blood motion during exercise testing may result in over-prescription of supplemental oxygen or delayed discharge from the health-care facility.Similarly,signal dropout or freezing of an oximeter display during motion may result in missed oxygen desaturations during cardio-respiratory monitoring in neonates at risk for apnea of prematurity.In contrast,reliable pulse oximetry may facilitate oxygen titration and reduce the incidence of serious neonatal complications such as retinopathy of prematurity and bronchopulmonary dysplasia.Reliable pulse oximetry monitoring may directly reduce perioperative costs.Routine use of supplemental oxygen to prevent postoperative hypoxemia has been challenged as expensive and potentially unnecessary.In one study,withholding of supplemental oxygen from adult postoperative patients who had room air SpO2 values of94%or greater,would have resulted in an estimated annual savings of over$600,000in that hos-pital alone[31].The availability of a monitor which can accurately measure SpO2and pulse rate in moving or poorly perfused patients may stimulate a re-evaluation of the clinical applications of pulse oximetry.Clinical areas which demand increased vigilance in detecting hypoxe-mia,but due to environmental or patient characteristics have precluded reliable pulse oximetry monitoring, should be reexamined[32].One such example is heli-copter transport of critically ill neonates,where Masimo SET pulse oximetry was shown to continuously mon-itor these patients while CPO failed during takeo¡and landing[33].CONCLUSIONMasimo SET pulse oximetry marks the¢rst new pulse oximetry paradigm since the invention of pulse oximetry over twenty-¢ve years ago.In view of the documented limitations of CPO,and proven and theoretical advan-tages of Masimo SET technology,studies are underway to re-evaluate pulse oximetry's role in improving pa-tient safety and reducing the cost of care in clinical areas where reliable continuous oxygenation monitoring is indicated.REFERENCES1.Severinghaus JW,Honda Y.History of blood gas analysis.VII.Pulse oximetry.J Clin Monit1987;3:135^1382.Wukitsch MW,Petterson MT,Tobler DR,Pologe JA.Pulse oximetry:Analysis of theory,technology,and practice.J Clin Monit1988;4:290^3013.Trivedi NS,Ghouri AF,Shah NK,Lai E,Barker SJ.E¡ects of motion,ambient light,and hypoperfusion on pulse oximeter function.J Clin Anesth1997;9:179^183 4.Freund PR,Overand PT,Cooper J,Jacobson L,Bosse S,Walker B,Posner KL,Cheney FW.A prospective study of intraoperative pulse oximetry failure.J Clin Monit 1991;7:253^258482Journal of Clinical Monitoring and Computing Vol16No72000。