0090-6689/88/167—1 73/$02.00/0
`Mepicat INSTRUMENTATION
`Copyright © 1988 by the Association for the Advancement of Medical Instrumentation
`
`Val. 22, No. 4
`
`Design and Evaluation of a New
`Pulse Oximeter Sensor
`
`
`
`eflectance
`
`Y. MENDELSON, PuD,J. C. KENT, MS, B. L. YOCUM, BS, AND M. J. BIRLE, BS
`
`The design and constructionof a newopticalreflectance sensorsuitable
`of an optical reflectance sensor suitable for noninvasive
`for noninvasive monitoring of arterial hemoglobin oxygen saturation
`monitoring of SpO, with a pulse oximeter. The experi-
`with a pulse oximeter is described. The reflectance sensor was inter-
`mental evaluation of the new sensorand verification that
`faced to a Datascope ACCUSAT pulse oximeter that was specially
`SpO, obtained with the reflectance sensor compare fa-
`adapted for this study to perform as a reflectance oximeter. We eval-
`vorably with:
`(a) SpO, measured simultaneously by a
`uated the reflectance sensor in a group of10 healthy adult volunteers.
`SpO,obtained from the forehead with the reflectance pulse oximeter
`finger sensor connected to a standard transmittance pulse
`and SpO,obtained fromafinger sensor that was connectedto a standard
`oximeter, and (b) HbO, measured by the IL 282 CO-
`ACCUSATtransmittance pulse oximeter were compared simultane-
`Oximeterfrom samplesof arterial blood in a group of 10
`ously to arterial blood samples analyzed by an IL 282 CO-Oximeter.
`healthy adult volunteers is presented.
`The equation for the best fitted linear regression line between the
`reflectance SpO, and HbO,values obtained from the reference IL 282
`CO-Oximeter in the range between 62 and 100% was: SpO, (%) =
`4.78 + 0.96 (IL); n = 110. The regression analysis revealed a high
`degree ofcorrelation (r = 0.98) and a relatively small standard error
`of the estimate (SEE = 1.82%), The meanandstandard deviationsfor
`the difference betweenthe reflectance SpO, and IL 282 measurements
`was 1,38 and 1.85%, respectively. This study demonstrates the ability
`to acquire accurate SpO,from the forehead using a reflectance sensor
`and a pulse oximeter.
`
`PULSE OXIMETRY
`
`The recent development of transmittance pulse ox-
`imeters by combining optical plethysmography with the
`spectrophotometric determination ofhemoglobin oxygen
`saturation in arterial blood (SpO,) has provided a widely
`used technique for monitoring hypoxemia.
`With transmittance pulse oximeters, sensor applica-
`tion is limited to several peripheral locations wherelight
`can be readily transmitted and detected, such as the
`fingertips, ear lobes, and toes on adults, and the foot or
`palmsoninfants. Alternatively, skin reflectance oximetry
`could enable SpO, measurement from more centrally
`located parts of the body such as the forearms, chest,
`and forehead, which cannot be monitored using conven-
`tional transillumination techniques.
`It appears that reflectance pulse oximetry may be par-
`ticularly suitable for direct assessment of fetal distress
`resulting from hypoxia during delivery, if used in addi-
`' tion to monitoring fetal heart rate by a scalp ECG elec-
`trode. Another suggested application of noninvasive re-
`flectance pulse oximetry is for monitoring SpO, in the
`external carotid artery through a sensor applied to the
`skin near the superficial temporal artery.!
`In this article we describe the design and construction
`
`From the Worcester Polytechnic Institute, Biomedical Engineering
`Program, Worcester, MA 01609.
`Address correspondenceand requests for reprints to Yitzhak Men-
`delson, PhD, Biomedical Engineering Program, Worcester Polytech-
`nic Institute, 100 Institute Rd., Worcester, MA 01609.
`
`167
`
`The principle of pulse oximetry was proposed by Aoy-
`agi et al? and further developed by Yoshiya et al.° This
`unique approach is based on the changein light absorp-
`tion by tissue. The change is caused primarily by arterial
`blood pulsation. The pulsating arterioles in a vascular
`bed, by expanding and relaxing, modulate the amount
`of light absorbed by the tissue and thus produce a char-
`acteristic photoplethysmographic waveform. The changes
`in light absorption are used to measure SpO, noninva-
`sively,
`Initial attempts to develop a noninvasive oximeterthat
`can measure oxygen saturation by analyzing the absolute
`light intensity that is diffusely reflected from the skin
`were only partially successful, mainly becauseoflimited
`accuracy associated with variations in tissue attenuation
`and differences in skin pigmentation. Recently, we
`showed that accurate SpO, measurements can be made
`utilizing a reflectance sensor and the concept of pulse
`oximetry.* > We found that SpO, can be calculated from
`the ratio of the reflected red and infrared photople-
`thysmogramsbased on a normalization in which the pul-
`satile (ac) componentof the red and infrared photople-
`thysmogramsis divided by the respective nonpulsatile
`(dc) component. The conversionofthe red/infrared ratios
`to SpO, is performed by an empirical calibration of the
`oximeter, This process is performed by comparing the
`red/infrared ratios measured by the pulse oximeter with
`blood HbO, values obtained from an invitro oximeter.
`The excursions of photoplethysmographic signals de-
`tected by reflectance and transmittance sensors when
`placed on the forehead and finger, respectively, are in-
`versely related to changes in arterial blood pulsations.
`Although the amplitude of the pulsatile component of
`the two waveformsis different, the shapes of the pho-
`
`APPLE 1015
`
`APPLE 1015
`
`1
`
`

`

`168
`
`MEDICAL INSTRUMENTATION
`
`Volume 22, No. 4, Augusi i988
`
`toplethysmograms are virtually identical, as illustrated
`in Figure 1.
`
`SENSOR DESIGN
`
`The basic optical sensor of a pulse oximeter consists
`of a red and an infrared light emitting diode (LED) and
`a suitable photodetector. In a transmittance pulse ox-
`imeter sensor,
`the LEDs and the photodetector are
`mounted in opposition, whereas in a reflectance sensor,
`the LEDS and the photodetector are mounted side by
`side. The wavelength ofthe red LED is typically chosen
`from regions of the spectra where the absorption coef-
`ficient of Hb and HbO,are markedly different (e.g., 660
`nm). The infrared wavelength, on the other hand,
`is
`typically chosen from the spectral region where the dif-
`ference in absorption coefficients of Hb and HbO,is
`relatively small (¢.g., 930 nm). The spectral response of
`the photodetector must overlap the emission spectra of
`the red and infrared LEDs.
`Practically, the major limitation in reflection pulse ox-
`imetry is the comparatively low-level photoplethysmo-
`gramstypically recorded from low-density, vascularareas
`of the skin. The feasibility of reflection pulse oximetry,
`therefore, is essentially dependent on the ability to de-
`sign a sensorthat can detect sufficiently strong reflection
`photoplethysmographic signals from various locations on
`the body.
`The light from the LEDs in the reflectance sensoris
`diffused by the skin inall directions. This suggests that
`to detect most of the backscattered radiation from the
`skin, the photodetector must be able to detect light from
`an area concentric with the LEDs. The intensity of the
`backscattered light decreases in direct proportion to the
`square of the distance between the photodetector and
`the LEDs; thus the photodetector should be mounted
`close to the LEDs. We found experimentally that a sep-
`aration of 4-5 mm between the LEDs and photodetector
`provides the best sensitivity in terms of detecting ade-
`
`quately large pulsatile components. We also found that
`when multiple photodetectors are arranged at equal dis-
`tances around the LEDs,
`the total amount of backscat-
`tered light that can be detected by the reflectance sensor
`is directly proportional to the numberof photodetectors,
`The optical reflectance sensor used in this study con-
`sists of two red (peak emission wavelength: 660 nm) and
`two infrared (peak emission wavelength: 930 nm) LED
`chips (dimensions: 0.3 < 0.3 mm), andsix silicon pho-
`todiodes (active area: 2.74 x 2.74 mm) arranged sym-
`metrically in a hexagonal configuration as shownin Fig-
`ure 2, To maximize the fraction of backscattered light
`collected by the sensor, the currents from all six pho-
`todiodes were summed, The LEDs and photodiodechips
`were mounted with conductive epoxy (Epo-tek H31,
`Epoxy Technology, Inc. Billerica, Massachusetts) on a
`ceramic substrate (dimensions: 13.2 x 13.2 x 0.25 mm)
`that was housedin a standard 24-pin (dimensions: 19 x
`19 mm) microelectronic package (AIRPAX, Cambridge,
`Maryland), which is commonly used for packaging elec-
`tronic circuits. The optical components were intercon-
`nected and wired to the package pins with |-mil (0.0254-
`mm diameter) aluminumwires, by a conventional ultra-
`sonic bonding technique. To minimize the amount of
`light transmission and reflection between the LEDs and
`the photodiodes within the sensor, a ring-shaped, opti-
`cally opaque shield of black Delrin (Dupont, Wilming-
`ton, Delaware) was placed between the LEDs and the
`photodiode chips. The optical components were encap-
`sulated inside the package using optically clear adhesive
`(NOA-63, Norland Products, Inc., New Brunswick, New
`Jersey). The microelectronic package was mounted in-
`side a black Delrin housing (dimensions: 3.2-cm diam-
`eter X 1.5-cm high). The sensor can be attached to the
`skin by meansof double-sided adhesive tape. The weight
`of the entire sensor assembly is approximately 11 g.
`
`SIGNAL PROCESSING
`
`The optical reflectance sensor was interfaced to a com-
`
`INFRARED REFLECTANCE PLETHYSMOGRAMS
`
`
`
`2 sec
`
`bt
`
`INFRARED TRANSMITTANCE PLETHYSMOGRAMS
`
`
`
`Figure 1. Relative infrared reflectance and transmittance photoplethysmograms recorded from the forehead and finger, respectively.
`
`2
`
`2
`
`

`

`A New REFLECTANCE PULSe Oximerer Sensor (Mendelson et af)
`
`1Go
`
`RED & INFRARED LEDs
`
`
`
`PHOTODIODE
`
`CERAMIC
`SUBSTRATE
`
`~
`
`CONDUCTIVE
`EPOXY
`
`
`
`——. DELRIN HOLDER
`
`- AIRPAX PACKAGE
`
`OPTICAL SHIELD
`
`DELRIN HOLDER
`
`(A)
`
`5 mm
`
`ceramic=isNe OPTICAL SHIELD
`
`SUBSTRATE
`AIRPAX PACKAGE
`
`PHOTODIODE
`
`LEDs
`
`OPTICALLY CLEAR E
`
`POXY
`
`
`
`Figure 2, Diagram (A and B) and photograph (C) of the reflectance pulse oximeter sensor.
`
`mercially available ACCUSAT (Datascope, Paramus, New
`Jersey) pulse oximeter.® The oximetercircuitry generates
`separate digital pulses to energize alternately the red
`and infrared LEDsin the sensor. The time-multiplexed
`current pulses from the photodiodes, which correspond
`to the red and infrared light intensities reflected from
`the skin, are first converted by the oximeter to propor-
`tional voltage pulses. The pulses are subsequently de-
`multiplexed into two separate channels. The red and
`infrared photoplethysmographic signals are then ampli-
`fied and high-pass filtered to separate the ac and de
`components of each photoplethysmographic waveform.
`Before the study began, an ACCUSATpulse oximeter
`was modified by adjusting the intensities of the red and
`infrared LEDs so that the de component of each pho-
`toplethysmogram was approximately equal to the cor-
`responding dc level obtained from transmittance pho-
`toplethysmograms as measured by a standard ACCUSAT
`sensor designed for a finger. The adjustment was per-
`formed while the reflectance and transmittance sensors
`were applied to the forehead andright index fingerof a
`white subject breathing ambient air. No further adjust-
`ments were made throughout the study, Thereflectanicé
`oximeter was adapted to provide a continuous readout
`of the ac and de components of the red and infrared
`photoplethysmograms.
`
`In addition to the modified ACCUSATpulse oximeter,
`a second standard ACCUSAT transmittance pulse ox-
`imeter was used to measure SpO, with a finger sensor.
`SpO,fromeachof the two pulse oximeters was acquired
`every 2s (0.5 Hz) using an AT&T 6300 personal com-
`puter. The conversion of the transmitted and reflected
`red/infrared ratios measured by each ACCUSAT pulse
`oximeter to SpO, was performed using the same internal
`calibration algorithms. The exact algorithm for calculat-
`ing SpO, was unavailable.
`
`IN VIVO EVALUATION
`
`The purpose of this study was to evaluate the per-
`formanceofthe reflectance pulse oximeter sensor during
`progressive steady-state hypoxia in humans and to com-
`pare the values obtained with the reflectance sensor to
`those from an ACCUSATtransmittance pulse oximeter
`and from the IL CO-Oximeter detecting the HbO, of
`simultaneously drawn arterial blood samples.
`Tests were performed on 10 healthy, nonsmoking, adult
`volunteers of different ages and skin pigmentations in
`compliance with the University of Massachusetts Med-
`ical Center review guidelines for humans experiments.
`The subject distribution was: one deeply pigmentedblack,
`two subjects of lightly pigmented Oriental descent, and
`
`3
`
`

`

`170
`
`Menpicat, INSTRUMENTATION
`
`Volurne 22, No. 4, Augusi 1988
`
`two darkly tanned andfive lightly tanned whites. Subject
`age ranged from 22-39 years (mean ++ SD: 29.2 + 5.6
`years), Measured hematocrit was in the range of 35-44%
`(mean + SD: 40 + 2.65%), Each volunteerwas informed
`of the complete procedure and possible risks associated
`with arterial cannulation and hypoxic gas breathing. Each
`volunteer received monetary compensation for partici-
`pation in the study..
`A modified Allen’s test for assessing the radial and
`ulnar arterial blood circulation to the hand was per-
`formed on each subject prior to arterial cannulation. A
`- Teflon cannula (22-gauge, 3.2-cm long) was inserted into
`the radial or ulnar artery of each subject after the sub-
`cutaneoustissue around the puncture site was anesthe-
`tized locally with a 1-ml injection of 1% lidocaine hy-
`drochloride (Xylocaine).
`All instruments warmedupforat least 30 min before
`the study. The transmittance sensor of the ACCUSAT
`pulse oximeter was attached to the index finger on the
`hand opposite that of the arm with the arterial cannula.
`The sensorof the reflectance pulse oximeter was attached
`to the middle of the forehead. Samplesofarterial blood
`(approximately 1 ml/sample) were drawn into 3-ml hep-
`arinized syringes and analyzed immediately by the
`2
`
`Instrumentation Laboratories {L, 282 CO-Oximeter
`(Instrumentation Laboratories, Lexington, Massachu-
`setts), Simultaneous measurements oftotal hemoglobin
`(Hb), oxyhemoglobin (HbO,), carboxyhemoglobin
`(HbCO), and methemoglobin (Hi) were obtained from
`each blood sample. The arterial cannula wasflushed with
`0.9% normal heparinized saline solution (1000 units/250
`ml) between blood samplings. Care was taken to ensure
`that the arterial line and the blood-sampling syringes
`werefree of air bubbles.
`A standard lead I ECG and the end-tidal CO, were
`continuously monitored by a Hewlett-Packard 78345A
`patient monitor (Hewlett-Packard, Andover, Massachu-
`setts). Each subject was placed in the supine position.
`A face mask wastightly fitted over the subject’s mouth
`and nose, and the subject was asked to breathe spon-
`taneously different O, and N, gas mixtures. The inspired
`O,/N, gas was supplied by a modified Heidbrink anes-
`thesia machine (Ohio Medical Products, Madison, Wis-
`consin), The breathing circuit was equipped with a CO,
`scrubber(soda lime). Inspired O, concentration was ad-
`justed between 10 and 100% and was monitored contin-
`uously with an IL 408 oxygen monitor that was inserted
`in the inspiratory part of the breathing circuit.
`
`
`
`
`
`R/IRREFLECTANCERATIO
`
`
`
`
`
`30
`
`60
`
`70
`
`80
`
`90
`
`100
`
`IL 282 CO-OXIMETER ¢%)
`Figure 3. Comparison of the iL 282 CO-Oximeter (x-axis) and the red/infrared ratios measured by the reflectance pulse oximeter (y-axis)
`during progressive steady-state hypoxia in 10 subjects. y = 3.51 — 0.030x; r =0.98; SEE = 0,060; n = 110; p < 0.001. Thesolid line
`represents the bestfitted linear regressionline.
`
`4
`
`

`

`A New REFLECTANCE PULSE OxIMETER GENSOR (Mendelsonef a/.)
`
`74
`
`arnmmompnmmCCAGOHCA
`
`oo
`
`la
`
`
`
`90 |
`
`.
`80F
`
`Ow
`
`Qa
`
`.
`
`Ww
`ur
`
`Za
`
`g?
`f-—
`
`mM 70
`
`0
`
`ud
`
`aZ
`
`-
`
`“A
`“7
`;
`2
`AY
`7
`
`60
`
`930
`
`30
`
`60
`
`70
`
`80
`
`90
`
`100
`
`IL 282 CO-OXIMETER (%)
`Figure 4. Comparison of SpO, measurements obtained from the reflectance pulse oximeter (y-axis) and the IL 282 CO-Oximeter (x-axis)
`during progressive steady-state hypoxia in 10 subjects. y = 4.78 + 0.96x;r = 0.98; SEE = 1.82; = 110; p < 0,001. The solid line represents
`the bestfitted linear regressionline. The dashedline representsidentity.
`
`Progressive hypoxemia was gradually induced by
`changing the inspired fractions of O, and N,. To provide
`a relatively uniform distribution of SpO, data points,
`samples were recorded during both desaturation and
`reoxygenation. Initially, the inspired O, concentration
`was changed in step decrements, each producing ap-
`proximately a 5% decrease in SpO, as determined from
`the ACCUSATtransmittance pulse oximeter display. The
`inspired O, was maintainedat each level until the pulse
`oximeter readings were stable. When the inspired O,
`reached 10%, corresponding to a saturation of approxi-
`mately 65%, the process was reversed, and the inspired
`O, was increased in a similar stepwise manner to 100%.
`SpO,from the ACCUSATandthereflectance pulse ox-
`imeters during blood sampling was acquired every 2 s
`(0.5Hz), using an AT&T 6300 personal computer.
`Noneof the subjects showed ECG abnormalities be-
`fore or during the study. All subjects tolerated the pro-
`cedure well, without adverse reactions.
`
`DATA ANALYSIS
`
`For each step change in inspired O,, readings from
`
`the ACCUSATtransmittance and reflectance pulse ox-
`imeters were averaged for 10 s before and after blood
`sampling and compared with the corresponding HbO,
`values measured by the IL 282 CO-Oximeter. To avoid
`operator biases, the data from each pulse oximeter were
`acquired automatically by the computer and later sub-
`jected to the samestatistical tests. Averaged readingsfor
`the 10 subjects were pooled and a least-squares linear
`regression analysis was performed. Student's t test de-
`termined the significance of each correlation; p < 0.001
`was considered significant.
`The SpO, displayed by two-wavelength pulse oxime-
`ters account only for the presence of HbO, and Hb in
`the blood. The presence of HbCO, Hi, or any other
`interfering substance in the blood is not accounted for.
`Therefore, the term often used to represent SpO, meas-
`ured by pulse oximeters is functional saturation,
`i.e.,
`HbO,/(Hb + HbO,), The IL 282 CO-Oximeter, on the
`otherhand, displays the percentage of oxygenated hemo-
`globin expressed as a fraction of the total hemoglobin
`present in the blood, i.e., HbO,/(Hb + HbO, + HbCO
`+.Hi), To compare SpO, measured by the pulse ox-
`imeters with corresponding readings from the IL 282
`
`5
`
`

`

`172
`
`MEbICAL INSTRUMENTATION
`
`Volume 22, No. 4, August 1986
`
`CO-Oximeter, both HbCO and Hi values from each blood
`sample were used to convert
`the IL, 282 readings to
`functional SpO,, according to the following relationship’:
`
`%SpO, (functional) = (HbO, x 100)/
`Pp
`2
`(100 — HbCO — Hi).
`
`RESULTS
`
`A total of 110 pairs of data points were used in the
`regression analysis, which gave the estimated slopes and
`intercepts of the regression lines. An average of 11 blood
`samples was collected from each subject. Each pair of
`data points represents a different hypoxic level. Regres-
`sion analysis of the HbO, values obtained from the IL
`282 CO-Oximeter(x-axis) vs the normalized red/infrared
`ratios (y-axis) as measured by the reflectance pulse ox-
`imeter is shown in Figure 3. The equation for the best
`fitted linear regression line was: y = 3.51 — 0,030x;
`r = —0.98; SEE = 0.060; p < 0.001. A comparison of
`SpO, readings from the reflectance pulse oximeter
`(y-axis) and the IL 282 CO-Oximeter(x-axis) is shown
`in Figure 4. The equationfor the bestfitted linear regres-
`sion line was: y = 4.78 + 0.96x; r = 0.98; SEE = 1.82;
`
`p < 0.001. Figure 5 shows the comparison of Sp, values
`measured by the ACCUSATreflectance pulse oximeter
`(y-axis) and the ACCUSATtransraittance pulse oximeter
`(x-axis), The linear regression equation for this compar-
`ison is: y = 5.85 + 0.95x; r = 0.98; SEE = 2.23; p <
`0.001. The standard deviations of the mean differences
`betweenthe reflectance oximeter SpO, and IL 282 HbO,
`values for four different saturation ranges are summa-
`rized in Figure 6.
`
`DISCUSSION
`
`Pulse oximetry has become a widely utilized medical
`technology, particularly in anesthesia andintensive care.
`Pulse oximeters offer significant monitoring advantages
`because of their reliability, simple operation, and the
`benefit of providing continuous SpO, monitoring.
`Noninvasive monitoring of oxygen saturation based
`upon skin-reflectance spectrophotometry was first de-
`scribed by Brinkman and Zijlstra.2 They showed that
`changes in oxygen saturation can be recorded noninva-
`sively from an optical sensor attached to the forehead.
`
`The use of light reflection instead of tissue transillumi-
`
`REFLECTANCESp02
`
`90
`
`80 f
`
`70 -
`
`60 +
`
`90
`
`60
`
`70
`
`80
`
`90
`
`100
`
`TRANSMITTANCE SpO2
`Figure 5. Comparison of SpO, measured by the reflectance pulse oximeter (y-axis) and the finger transmittance pulse oximeter (x-axis)
`I
`during progressive steady-state hypoxia in 10 subjects. y = 5.85 + 0.95x; r = 0,98; SEE = 2.23;n = 110; p < 0.001. The solid line represents
`the bestfitted linear regression line. The dashed line represents identity.
`
`6
`
`

`

`
`
`A New Rercectance Puise OxiMeTer Sensor (Mendelson ef al.)
`
`
`
` STANDARDDEVIATIONOFMEANDIFFERENCES
`
`173
`
`perlormed successfully using a reflectance sensor con-
`nected to a standard transmittance pulse oximeter with-
`out significant modifications of hardware and software.
`
`SUMMARY
`
`We compared simultaneous SpO, from a reflectance
`pulse oximetersensorattached to the forehead and from
`a transmittance pulse oximeter with a sensorattached to
`a finger with HbO,fromarterial blood samples in a group
`of 10 healthy adult volunteers. A high degree of corre-
`lation was found for SpO, between 62 and 100%. Relative
`to arterial blood samples, the SEE for the reflectance
`pulse oximeter was 1.82%. We concludethatin situations
`in which a transmittance pulse oximeter cannot be used
`reliably, the forehead may be considered as a suitable
`alternative site for monitoring SpO, with a reflectance
`pulse oximeter sensor.
`
`We gratefully acknowledge the clinical assistance of Albert
`Shahnarian, PhD, Gary W. Welch, MD, PhD, and Robert M.
`Giasi, MD, Department of Anesthesiology, University of Mas-
`sachusetts Medical Center, Worcester, Massachusetts. We are
`indebted to Paul A. Nigroni, Datascope Corporation, Paramus,
`NewJersey, and Kevin Hines, Semiconductor Division, Analog
`Devices, Wilmington, Massachusetts, for their technical as-
`sistance. Financial support for this study was provided by the
`Datascope Corporation.
`
`REFERENCES
`
`1. Cheng EY, Hopwood M, Kay J: Pulse oximetry: Evaluation of a
`new sensor. Proceedings of the Twenty-second AAM! Annual
`Meeting, p 20, 1987
`2. Aoyagi T, Kishi M, Yamaguchi K, Watanable S: Improvementof
`the earpiece oximeter. Thirteenth Annual Meeting Jpn Soc
`Med Electronics Biomed Eng 90-91, 1974
`3. Yoshiya |, Shimada Y, Tanaka K: Spectrophotometric monitoring
`of arterial oxygen saturation in the finger tip. Med Biol Eng
`Comput 18:27-32, 1980
`4, Mendelson Y, Cheung PW, Neuman MR, Fleming DG, Cahn
`SD: Spectrophotometric investigation of pulsatile blood flow
`for transcutaneous reflectance oximetry. Adv Exp Med Biol
`159: 93-102, 1983
`5. Mendelson Y, Ochs BD: Sensor development for reflectance |
`pulse oximetry. Proc Thirty-ninth ACEMB, p 222, 1986
`6. Mendelson Y, Kent JC, Shahnarian A, Welch GW, Giasi RM:
`Evaluation of the Datascope ACCUSAT pulse oximeter in
`healthy adults. J Clin Monit 4: 59-63, 1988
`\L-282 CO-Oximeter Operator's Manual: Lexington, Massachu-
`setts, Instrumentation Laboratory, Inc., 1977
`8. Brinkman R, Zijlstra WG: Determination and continuous regis-
`’
`tration of the percentage oxygen saturationin clinical condi-
`tions. Arch Chir Neerl 1:177-183, 1949
`9. Gohen A, Wadsworth N: A light emitting diode skin reflectance
`oximeter. Med Biol Eng Comput 10:385-391, 1972
`10, Takatani S: On the theory and development of a noninvasive
`tissue reflectance oximeter. PhD dissertation, Case Western
`Reserve Univ, Cleveland, Ohio, 1978
`11. Discussion IX: In Payne JP, Severinghaus JW (eds): Pulse Ox-
`imetry, Springer-Verlag, 1986 .
`
`7,
`
`
`
`S:
`0 - ae
`BELOW 70
`
`4
`
`ee Oe
`70-80
`
`eal g
`80-90
`
`ey _ os
`ABOVE 90
`
`IL 282 CO-OXIMETER (%)
`
`Figure 6. Standard deviations of the mean differences between the
`reflectance pulse oximeter and the IL 282 CO-Oximeter. N is the
`number of paired data points included in the statistical analysis.
`
`nation was suggested to enable noninvasive monitoring
`from virtually any skin surface. More recently, Cohen
`and Wadsworth? and Takatani!® attempted to develop a
`skin reflectance oximeterutilizing a similar spectropho-
`tometric approach. In those three reflectance oximeters,
`oxygen saturation was calculated fromthe absolute light
`intensity diffusely reflected (backscattered) from the skin,
`Although these developments represent significant ad-
`vancements in noninvasive oximetry, the major problems
`were limited accuracy, poor reproducibility, and diffi-
`culties in absolute calibration.
`Available transmittance pulse oximeters can be used
`only on a few specific peripheral locations. The approach
`presented in this article demonstrates that SpO, can be
`measured from an alternate site, specifically the fore-
`head. This technique providesa clinically acceptable al-
`ternative to presently available transmittance pulse
`oximeters. Although we found that reflectance photo-
`plethysmograms can be detected from several locations
`on the body (e.g., forearm, chest, and back), the rela-
`tively small, photoplethysmographicsignals lead to prac-
`tical problems when processed by the pulse oximeter.
`Therefore, the choice of the forehead as a site for our
`study was based onthefact thatat this location, relatively
`large reflectance photoplethysmographic signals can be
`detected.
`Therelationship between the red/infrared ratios meas-
`ured by the reflectance pulse oximeter and HbO, meas-
`ured by the IL 282 CO-Oximeter produced a regression
`relationship similar in slope and intercept to that ob-
`served from transmittance pulse oximeters.! This sug-
`gests that SpO, monitoring from the forehead can be
`
`7
`
`