798
`
`IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 35, NO. 10, OCTOBER 1988
`
`Noninvasive Pulse Oximetry Utilizing Skin
`Reflectance Pho toplethy smog mphy
`
`Abstract-The major concern in developing a sensor for reflectance
`pulse oximetry is the ability to measure large and stable photople-
`thysmograms from light which is backscattered from the skin. Utiliz-
`ing a prototype optical reflectance sensor, we showed that by locally
`heating the skin it is possible to increase the pulsatile component of the
`reflected photoplethysmograms. Furthermore, we showed that addi-
`tional improvements in signal-to-noise ratio can be achieved by in-
`creasing the active area of the photodetector and optimizing the sepa-
`ration distance between the light source and photodetector. The results
`from a series of in vivo studies to evaluate a prototype skin reflectance
`pulse oximeter in humans are presented.
`
`I. INTRODUCTION
`ONINVASIVE monitoring of arterial hemoglobin
`
`N oxygen saturation (Sa02) based upon skin reflectance
`
`spectrophotometry was first described by Brinkman and
`Zijlstra in 1949 [l]. They showed that changes in Sa02
`can be recorded noninvasively from an optical sensor at-
`tached to the forehead. Their innovative idea to use light
`reflection instead of tissue transillumination, which is
`limited mainly to the finger tips and ear lobes, was sug-
`gested as an improvement to enable noninvasive monitor-
`ing of Sa02 from virtually any skin surface. More recent
`attempts to develop a skin reflectance oximeter utilizing
`a similar spectrophotometric approach were made by
`Cohen et al. [2] and Takatani [3]. All of those three non-
`invasive reflectance oximeters attempted to monitor Sa02
`by measuring the absolute light intensity diffusely re-
`flected (backscattered) from the skin.
`While those developments represent significant ad-
`vancements in noninvasive reflectance oximetry, limited
`accuracy as well as difficulties in absolute calibration were
`major problems with early reflectance oximeters. Al-
`though various methods have been proposed, to date, a
`versatile noninvasive reflectance oximeter, which can
`monitor Sa02 reliably from any location on the skin sur-
`face, is not yet available.
`Backscattered light from living skin depends not only
`on the optical absorption spectrum of the blood but also
`on the structure and pigmentation of the skin. In an at-
`tempt to overcome this problem, Mendelson et al. [4]
`
`Manuscript received June 17, 1987; revised May 9, 1988. This work
`was supported by the Whitaker Foundation and the National Science Foun-
`dation under Grant ECS-8404397.
`The authors are with the Biomedical Engineering Program, Worcester
`Polytechnic Institute, Worcester, MA 01609.
`IEEE Log Number 8822615.
`
`proposed to measure Sa02 based on the principle of skin
`reflection photoplethysmography . We showed that Sa02
`can be measured noninvasively by analyzing the pulsatile
`rather than the absolute, reflected light intensity Z, of the
`respective red and infrared photoplethysmograms accord-
`ing to the following empirical relationship [4]-[5]:
`Sa02 = A - B [Z,(red)/Z,(infrared)]
`( 1 )
`where A and B are empirically derived constants which
`are determined statistically during in vivo calibration in
`which the Zr ( red)/Zr( infrared) ratio calculated by the
`pulse oximeter is compared against direct blood Sa02
`measurements. Z, is obtained by a normalization process
`in which the pulsatile (ac) component of the red and in-
`frared photoplethysmograms is divided by the corre-
`sponding nonpulsatile (dc) component.
`In clinical applications where presently available trans-
`mission pulse oximeters cannot be used, there is a need
`for an optical sensor which is suitable for monitoring Sa02
`utilizing light reflection from the skin. Although the prin-
`ciples of reflection and transmission pulse oximetry are
`very similar, the major limitation of reflection pulse oxi-
`metry is the comparatively low level photoplethysmo-
`grams typically recorded from the skin. The feasibility of
`reflection pulse oximetry, therefore, is highly dependent
`on the ability to detect sufficiently strong reflection pho-
`toplethy smograms.
`This paper describes the considerations in designing a
`skin reflectance sensor for noninvasive monitoring of
`Sa02. The ability to detect improved photoplethysmo-
`graphic waveforms through the use of skin heating and
`multiple photodetectors are discussed. Results from a se-
`ries of in vivo studies to evaluate a prototype skin reflec-
`tance pulse oximeter in humans are presented.
`
`11. BACKGROUND
`A. Principle of Pulse Oximetry
`Pulse oximetry has been invented by Aoyagi et al. [6]
`and further refined by Nakajima et al. [7] and Yoshiya et
`al. [8]. This unique approach is based on the assumption
`that the change in light absorbed by tissue during systole
`is caused primarily by the arterial blood. Consequently,
`they showed that changes in light transmission through a
`pulsating vascular bed can be used to obtain an accurate
`noninvasive measurement of Sa02.
`The main advantage of employing a photoplethysmo-
`
`0018-9294/88/1000-0798$01 .OO O 1988 IEEE
`
`Page 1
`
`VALENCELL EXHIBIT 2011
`IPR2017-00315
`
`

`

`MENDELSON AND OCHS: NONINVASIVE PULSE OXIMETRY
`
`graphic technique is that only two wavelengths are re-
`quired, thereby greatly simplifying the optical sensor.
`Furthermore, the requirement for blood ‘ ‘arterialization”
`which was essential in previous nonpulsatile oximeters,
`such as the eight wavelength Hewlett-Packard (HP) ear
`oximeter [9], has been eliminated. Hence, there is no need
`for continuous skin heating. Moreover, skin pigmenta-
`tion, which can cause variable light attenuation, does not
`seem to affect the accuracy of pulse oximeters. This is
`because the ratio of the transmitted redhnfrared light in-
`tensity, from which Sa02 is calculated, is obtained by a
`normalization process in which the ac component of the
`red and infrared photoplethysmograms is divided by the
`corresponding dc components.
`The basic optical sensor of a noninvasive pulse oxi-
`meter consists of a red and infrared light emitting diodes
`(LED’s) and a silicone photodiode. The wavelength of
`the red LED is typically chosen from regions of the spec-
`tra where the absorption coefficient of Hb and Hb02 are
`markedly different (e.g., 660 nm). The infrared wave-
`length, on the other hand, is typically chosen from the
`spectral region between 940 and 960 nm where the differ-
`ence in the absorption coefficients of Hb and Hb02 is rel-
`atively small. The photodiode used has a broad spectral
`response that overlaps the emission spectra of the red and
`infrared LED’s.
`The light intensity detected by the photodetector de-
`pends, apart from the intensity of the incident light,
`mainly on the opacity of the skin, reflection by bones,
`tissue scattering, and the amount of blood present in the
`vascular bed. The amount of light attenuated by the blood
`varies according to the pumping action of the heart. Con-
`sequently, as tissue blood volume increases during sys-
`tole, a greater portion of the incident light is absorbed by
`the arterial blood causing a rapidly alternating signal. De-
`pending on the physiological state of the microvascular
`bed, typically, these alternating light intensity amounts to
`approximately 0.05-1 percent of the total light intensity
`either transmitted through or backscattered from the skin.
`Since pulse oximeters rely on the detection of arterial
`pulsation, significant reduction in peripheral blood flow,
`such as in hypotension or hypothermia, can limit the re-
`liability of the measurement. Nevertheless, the fact that
`no user calibration or site preparation is required, and the
`availability of small, light weight, and easy to apply sen-
`sors has made transmission pulse oximeters very popular
`in various clinical applications.
`
`B. Rejection Versus Transmission Pulse Oximetry
`In transmission pulse oximetry, sensor application is
`obviously limited to areas of the body, such as the finger
`tips, ear lobes, toes, and in infants the foot or palms where
`transmitted light can be readily detected. Other locations,
`which are not accessible to conventional transillumination
`techniques, i.e., the limbs, forehead, and chest may be
`monitored in principle using a reflection Sa02 sensor as
`shown schematically in Fig. 1.
`Although the specific clinical utility of reflectance pulse
`
`199
`1
`
`r
`
`m
`
`Y z
`w a
`M n
`w
`
`m
`
`Y z
`w a
`
`Fig. 1. Principle of reflectance pulse oximetry illustrating the optical sen-
`sor and the different layers of the skin.
`
`oximetry has yet to be determined, it appears that the
`technique may have potential application for neonatal
`monitoring. For example, a reflectance Sa02 sensor may
`be of considerable value in the assessment of fetal distress
`during delivery if used in addition to presently available
`screw-type scalp ECG electrodes. Furthermore, since the
`skin of the chest is supplied by branches of the internal
`thoracic artery, which in turn stem fwm blood vessels
`leaving the aorta above the ductus arteriosus, Sa02 mea-
`surements using a reflectance sensor attached to the chest
`may prove to be of clinical importance when monitoring
`newborn infants with a patent ductus arteriosus.
`
`111. METHODS
`
`A. Instrumentation
`I ) Reflectance Sa02 Sensor: We have constructed and
`tested a prototype reflectance sensor which consists of
`three parts: an optical sensor for monitoring Sa02, a feed-
`back-controlled heater for varying the local temperature
`of the skin under the sensor, and a laser Doppler probe
`for recording relative changes in skin blood flow under
`the sensor.
`A schematic diagram illustrating the front view of the
`combined sensor is shown in Fig. 2. The sensor assembly
`can be attached to the skin by means of a double-sided,
`ring-shaped, tape. This attachement technique is suffi-
`cient to maintain the sensor in place without exerting ex-
`cessive pressure that could significantly reduce local blood
`flow in the skin.
`The optical sensor for monitoring Sa02 consists of red
`and infrared LED’s with peak emission wavelength of 660
`and 950 nm, respectively, and a silicone p-i-n photo-
`diode. The half-power spectral bandwidth of each LED is
`approximately 20-30 nm. The LED’s (dimensions: 0.3
`X 0.3 mm) and photodiode (dimension: 2.0 X 3.0 mm)
`
`Page 2
`
`

`

`800
`
`-----+7
`
`SLIOING PLATE
`
`TEMPERATURE SENSDR
`
`nm tm
`
`PHOTODIOOE
`
`LASER OORER BLOOD FLON
`FIBER OPTIC SENSOR
`
`IR tm
`
`Fig. 2. Frontal view of the combined SaO,/laser Doppler skin blood flow
`sensor.
`
`chips were mounted on separate ceramic substrates. A
`small drop of clear epoxy resin was applied over the
`LED’s and photodiode for protection. For investigational
`purposes, the ceramic substrates containing the LED’s and
`photodiode were mounted on separate sliding plates. This
`arrangement provides convenient adjustment of the sepa-
`ration distance between the LED’s and the photodiode
`from 4 to 11 mm. Undesired specular light reflections
`from the surface of the skin, as well as direct light path
`between the LED’s and the photodiode, were minimized
`by recessing and optically shielding the LED’s and pho-
`todiode inside the sensor assembly.
`The feedback-controlled heater consists of a round ther-
`mofoil heating element (1.25 cm diameter) and a solid-
`state temperature transducer (Analog Devices AD590)
`mounted in close proximity to the surface of the sensor
`contacting the skin. The heater is capable of delivering a
`maximum power of 2 W. The temperature of the sensor
`can be adjusted between 34 and 45°C in 1 +/-0.1”C
`steps.
`The distal ends of two parallel glass optical fibers (diam.
`0.15 mm; separation 0.5 mm) were used for recording
`relative skin blood flow under the reflectance sensor. The
`fiber tips were mounted in close proximity to the LED’s
`and photodiode. The proximal ends of these optical fibers
`were coupled to a MEDPACIFIC Model LD 5000 Laser
`Doppler perfusion monitor (MEDPACIFIC Corp., Seat-
`tle, WA). A 5 mW, continuous wave, HeNe laser located
`inside the perfusion monitor generates a monochromatic
`beam of red (632.8 nm) light. This light passes to the
`skin through one optical fiber which illuminates a region
`of tissue that approximates a hemisphere with a radius of
`about 1 mm. The light entering the tissue is scattered by
`the moving red blood cells causing a frequency shift pro-
`portional to the blood flow according to the Doppler prin-
`ciple [ 101. A portion of the backscattered light from both
`the nonmoving tissue structures and the moving red blood
`cells is then collected by an adjacent optical fiber and
`projected onto a photodiode inside the LD 5000 monitor.
`The electrical output from this photodiode is processed by
`the perfusion monitor resulting in a continuous reading
`
`IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 35, NO. IO, OCTOBER 1988
`
`that is proportional to the skin blood flow under the sen-
`sor. The instrument was nulled electronically before each
`study by adjusting the output reading to zero after the sen-
`sor was positioned over a stationary surface of white scat-
`tering material. To avoid optical interference between the
`LED’s in the Sa02 sensor and the HeNe laser source, the
`reflectance pulse oximeter was turned off when skin blood
`flow measurements were performed.
`2) Rejectance Pulse Oximeter: The reflectance oxim-
`eter generates digital switching pulses to drive the red and
`infrared LED’s in the sensor alternately at a repetition rate
`of 1 KHz. The time multiplexed output current from the
`photodiode, which correspond to the red and infrared light
`intensities reflected from the skin, is first converted to a
`proportional analog voltage using a low noise operational
`amplifier configured as a current-to-voltage converter. The
`resulting output voltage is subsequently decomposed into
`two separate channels using two sample-and-hold circuits
`synchronously triggered by the same pulses driving the
`respective LED’s. The red and infrared photoplethysmo-
`grams produced are amplified and high-pass filtered (cut-
`off frequency 15 Hz) to separate the ac pulses from the dc
`signal of each photoplethysmogram. To enable further
`signal processing, the respective ac and dc signals of each
`photoplethysmogram were digitized at a rate of 100 sam-
`ples /s by an IBM-AT personal computer equipped with
`a Tecmar 12 bit resolution A/D-D/A data acquisition
`board. From the recorded signals, a computer algorithm
`calculates the Zr ( red) /Ir ( infrared) ratio for each heart-
`beat. These values are further averaged using a five-point
`running average algorithm. Another algorithm uses the
`averaged ratios to compute and display Sa02 according to
`(1). The A and B coefficients necessary for calculating
`Sa02 in the oximeter were determined previously in our
`laboratory based on a calibration study using the HP
`Model 47201A ear oximeter as a reference.
`
`B. In Vivo Studies
`
`Seven Caucasian volunteers participated in the studies
`which were approved by our institutional review board.
`The subjects, five males and two females, were healthy
`nonsmokers ranging in age from 21 to 29 years.
`To establish a reference for measuring Sa02, we used
`the HP 47201A ear oximeter. The oximeter was standard-
`ized before each test by placing the ear probe in a special
`standardization chamber inside the ear oximeter. The ear
`probe was then attached to the anti-helix portion of the
`ear pinna with a head mount and elastic head band ac-
`cording to the manufacturer recommendations.
`The sensor of the reflectance pulse oximeter was at-
`tached either to the volar side of the forearm or the ante-
`rior thigh region. In each case, the monitored arm or leg
`was immobilized in the horizontal position to minimize
`spurious movement artifacts.
`The experimental setup used in our studies is illustrated
`in Fig. 3.
`
`Page 3
`
`

`

`MENDELSON AND OCHS: NONINVASIVE PULSE OXIMETRY
`
`Hewlett-Packard
`
`Flow Meters
`n, U,
`
`Q13 cp
`
`80 1
`
`i21
`
`ull
`
`0 W
`
`W
`
`0
`
`Fig. 3. Experimental setup illustrating the closed loop rebreathing circuit
`for obtaining different inspired Oz/NZ concentrations and the attachment
`of the oximeter sensors to the subject’s ear and thigh.
`
`IV. RESULTS
`Several in vivo studies were performed using the pro-
`totype optical reflectance sensor and oximeter as de-
`scribed above. The primary objectives of the first study
`were to investigate the effect of 1) source/detector sepa-
`ration and 2) local skin heating on the pulsatile compo-
`nent of the red and infrared photoplethysmograms de-
`tected by the sensor. In a separate in vivo study, we
`compared SaO, values measured by the pulse oximeter
`from the forearm and thigh of different subjects during
`progressive hypoxemia with simultaneous recordings ob-
`tained from the HP ear oximeter in the range between 70-
`100 percent.
`A. Source/Detector Separation Studies
`The purpose of these studies was to determine the re-
`lationship between different LED/photodiode separations
`and the magnitude of the pulsatile component of each re-
`flection photoplethysmogram. We noticed that for a con-
`stant LED intensity, the light intensity detected by the
`photodiode decreases roughly exponentially as the radial
`distance from the LED’s is increased. The same basic re-
`lationship applies to both the dc and ac components of
`the reflected photoplethysmograms as shown in Fig. 4.
`This is expected since the probability that the incident
`photons will be absorbed as they traverse a relatively
`longer path length before reaching the detector is in-
`creased.
`Fig. 5 shows the relative pulse amplitude of the red and
`infrared reflected photoplethysmograms recorded from the
`forearm of one subject. In this study, the incident light
`intensities of the red and infrared LED’s were adjusted by
`varying the LED driving currents such that for each sep-
`aration distance the dc component of each photople-
`thy smogram remained relatively constant. Each point
`represents the average values obtained for five repeated
`experiments performed on the same subject. In each ex-
`periment, and for each separation distance, the data ac-
`quired were averaged over a 30 s time interval.
`As shown in Fig. 5, by increasing the separation dis-
`tance between the LED’s and photodiode from 4 to 11
`
`4
`
`12
`
`10
`8
`6
`SEPARATION DISTANCE Imml
`Fig. 4. The effect of LED/photodiode separation on the dc (U) and ac (0)
`components of the reflected infrared photoplethysmograms. Measure-
`ments were performed at a skin temperature of 43°C.
`
`; /
`
`-
`g 0.8
`4
`
`4
`
`e
`8
`7
`5
`lo
`8
`LED/PHOTODIODE SPACING [ rnm 1
`Fig. 5. Effect of LED/photodiode separation on the relative pulse ampli-
`tude of the red (+) and infrared (U) photoplethysmograms. The driving
`currents of the Fed ( U ) and infrared (*) LED’s required to maintain a
`constant dc reflectance from the skin are shown for comparison.
`
`11
`
`mm, we were able to achieve almost a two-fold increase
`in the pulse amplitude af the infrared photoplethysmo-
`gram. Furthermore, as illustrated in Fig. 6, the mean beat-
`to-beat variations of the infrared photoplethysmograms,
`which were determined by calculating the respective coef-
`ficients of variation (i.e., the standard deviation divided
`by the mean €or a 30 s time interval), decreased from about
`7 to 3 percent. This trend indicates that the photopleth-
`ysmograms became progressively more stable as the LED/
`photodetector separation was increased. Similar trends
`were also observed for the reflected red photoplethysmo-
`grams.
`B. Skin Heating Studies
`Practically, it is difficult to detect large reflection pho-
`toplethysmograms from skin areas which are not very vas-
`cular, such as the chest and the limbs. In this study, we
`attempted to determine if local skin heating, which is
`known to produce vasodilatation of the microvascular bed,
`could be used as a practical mean to increase the pulsatile
`component of the reflected photoplethysmograms. Like-
`wise, we sought to determine if skin heating could help
`to reduce the beat-to-beat variability in the pulsatile com-
`ponents of the recorded photoplethysmograms.
`
`Page 4
`
`

`

`802
`
`IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 35, NO. IO, OCTOBER 1988
`
`0
`
`4
`
`5
`
`9
`8
`6
`7
`SEPARATION DISTANCE (mm)
`Fig. 6 . Effect of LEDlphotodiode separation on the mean pulse amplitude
`(*) and the corresponding decrease in the beat-to-beat amplitude fluctua-
`tion (H) of the infrared photoplethysmograms expressed in terms of the
`coefficient of variation. Each pulse amplitude was normalized with re-
`spect to a separation distance of 4 mm.
`
`10
`
`1 1
`
`7
`
`x h
`
`T1°
`
`t 9
`
`O J
`
`34 3 5
`
`36 37 38 39 40 41
`TEMPERATURE ('C)
`Fig. 7. Effect of skin temperature on the mean pulse amplitude ( 0 ) and the
`corresponding decrease in the coefficient of variation (M) of the infrared
`photoplethysmograms. Each pulse amplitude was normalized with re-
`spect to a separation distance of 4 mm.
`
`42 43 44
`
`45
`
`Measurements were performed at a constant LED/pho-
`todiode separation of 6 mm while the subject was breath-
`ing ambient air. After attaching the reflectance sensor to
`the forearm, the surface of the skin was gradually heated
`to 45°C in 1°C step increments. The time needed to
`achieve a desired skin temperature depends on factors such
`as skin type> local blood flow, heat conductivity of the
`skin, and the temperature of the surrounding environ-
`ment. Typically, we found that at each temperature set-
`ting, 5 min were sufficient for the skin temperature to
`reach steady state.
`As shown in Fig. 7, by increasing the local skin tem-
`perature from 34" to 45"C, we were able to obtain a five-
`fold increase in the pulse amplitude of the infrared pho-
`toplethysmograms. Moreover, by heating the skin, the
`vascular bed under study becomes vasodilated and, there-
`
`fore, the reflected photoplethy smograms become more
`stable resulting in smaller beat-to-beat amplitude fluctua-
`tions. Consequently, as our data show, the mean coeffi-
`cient of variation decreased from approximately 7 to 2
`percent. Similar trends were also observed for the re-
`flected red photoplethysmograms.
`The effect of local skin heating on the pulsatile com-
`ponent of the reflected photoplethysmograms is shown in
`Fig. 8. The relative skin blood flow for each temperature
`setting is also shown for comparison. It is clearly seen
`that as the temperature of the skin was increased from its
`initial value of 29" to 43"C, the pulse amplitude of the
`red and infrared photoplethysmograms increased accord-
`ingly. Furthermore, the mean pulse amplitude of the re-
`corded waveforms remained relatively constant over a pe-
`riod of approximately 20 min after the heater was turned
`
`Page 5
`
`

`

`MENDELSON AND OCHS: NONINVASIVE PULSE OXIMETRY
`
`803
`
`HN-
`2 9 O C ( 1 . 5 )
`
`- - -4 HO--
`43OC(4.0)
`43OC(4.5)
`43'C(4.8)
`
`35OC(4.5)
`
`34OC(3.0)
`
`33OC(2.1)
`
`4
`3ZoC(1.8) 3OoC(1.6)
`
`'
`
`1
`
`
`
`IR
`(VI
`
`0
`
`Fig. 8. Simultaneous recording of the infrared and red photoplethysmo-
`grams from the forearm at different skin temperatures. The numbers in
`parenthesis indicate the relative skin blood flows (scale: 0-10). Each
`record lasted approximately 15 s. The time elapsed between consecutive
`recordings is 10 min. HN = heater turned on, HO = heater turned off.
`
`off. Thereafter, the pulse amplitude started to diminish.
`After about 50 min, the pulse amplitude returned to its
`initial level.
`
`C. Hypoxemia Studies
`Preliminary studies using our prototype reflectance sen-
`sor during progressive steady-state hypoxemia were con-
`ducted on a group of seven healthy adult volunteers.
`Each subject was placed in a reclining position and
`asked to breathe different fractions of 02/N2 gas while
`maintaining spontaneous respiration. The inspired O2 /N2
`gas mixture was supplied through a fitted face mask by a
`closed-loop rebreathing circuit equipped with a COz
`scrubber and a one-way breathing valve. The fractional
`inspired O2 concentration ( F I 0 2 ) was adjusted between
`10 and 100 percent using separate gas flowmeters. The
`exact inspired FI02 was monitored continuously with an
`Instrumentation Laboratory Model 408 oxygen monitor
`(Instrumentation Laboratories Inc. , Lexington, MA)
`which was inserted in the inspiratory limb.
`The skin reflectance sensor was attached to the volar
`side of the foream and maintained at a constant temper-
`ature of 43°C. The spacing between the LED's and the
`photodiode in these experiments was set to 6 mm.
`Initially, the F I 0 2 was changed in step decrements, each
`producing a 5 percent decrease in Sa02 as measured by
`the reference HP ear oximeter. At each Sa02 level, the
`inspired F,02 was maintained at a constant level until both
`oximeters displayed stable readings.
`For each step change in F,02, SaOz readings from our
`prototype reflectance pulse oximeter were averaged over
`60 s time intervals and compared to the corresponding
`SaOz values measured simultaneously by the HP ear ox-
`imeter. The averaged readings from all seven subjects
`were then pooled and a linear regression analysis was per-
`formed.
`
`A comparison between the reflectance pulse oximeter
`and the HP ear oximeter readings obtained from all seven
`subjects is shown in Fig. 9, A total of 66 pairs of data
`points were used in this regression analysis. Linear
`regression analysis of this experimental data resulted in a
`slope of 0.93 and a positive y intercept of 6.22 percent ( r
`= 0.96; S.E.E = 2.20). The mean and standard deviation
`for the differences between the skin reflectance pulse ox-
`imeter and the HP SaOZ readings were found to be -0.001
`+ / - 2.27, respectively.
`In order to determine if repeatable Sa02 measurements
`can be made also from body sites other than the forearm,
`we performed a similar series of experiments in which the
`sensor was applied to the thigh region of three different
`subjects. In these experiments, a total of 24 data points
`were obtained simultaneously from the reflectance pulse
`oximeter and the HP ear oximeter. Linear regression anal-
`ysis of this data set revealed a slope of 0.93 and a positive
`y intercept of 6.5 percent ( r = 0.99; S.E.E. = 1.56).
`The mean and standard deviation for the differences be-
`tween the skin reflectance pulse oximeter and the HP SaOz
`readings were found to be -0.001 + / - 1.61 , respec-
`tively.
`The response of our prototype skin reflectance oximeter
`was further compared against simultaneous recordings of
`Sa02 from the fingertip and earlobe made by a transmis-
`sion pulse oximeter (Nellcor Model N-100, Nellcor, Inc.,
`Hayward, CA) and the HP ear oximeter, respectively. The
`recordings, which are shown in Fig. 10, were obtained by
`asking the subject to hyperventilate and then hold his
`breath consecutively.
`D. Multiple Photodetector Arrangement
`The incident light emitted from the LED's diffuses in
`the skin in all directions. This is evident from the circular
`pattern of backscattered light surrounding the LED's.
`Therefore, by collecting the backscattered radiation using
`
`Page 6
`
`

`

`IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 35, NO. 10, OCTOBER 1988
`
`804
`
`100
`
`85
`
`EO
`
`75
`
`~~
`
`t
`
`T m - m
`NO. Of Subjects = 7
`Corr. Coef. = 0.964
`S. E. E. = 2.20
`
`,
`
`++
`
`,’i
`
`65
`
`70
`
`75
`
`EO
`
`E5
`
`90
`
`95
`
`100
`
`HP Eor Oximeter Sa02 ( X )
`Fig. 9. Comparison of SaOp recorded simultaneously from the forearm and
`the ear by the skin reflectance pulse oximeter and the HP ear oximeter,
`respectively.
`
`601
`
`20 sec
`
`todetectors instead of only one, we modified our sensor
`and mounted two additional photodiodes similar in size
`and spectral response to that used originally. This enabled
`us to triple the total active area of the photodetector and
`thus collect a greater fraction of the backscattered light
`from the skin. Fig. 11 shows the spatial arrangement of
`all three photodiodes which were mounted symetrically
`with respect to the red and infrared LED’s. Also shown
`in this figure is the relative pulse amplitude of the red and
`infrared photoplethy smograms recorded from the forearm
`when the output currents of several photodiodes were
`summed simultaneously. As expected, we can see that by
`using multiple photodetectors a larger fraction of the
`backscattered radiation from the skin can be collected and,
`therefore, larger photoplethysmograms can be recorded.
`V. DISCUSSION
`The sensor designed for this study enabled us to ex-
`amine the effect of LED/photodiode separation distance
`as well as skin heating on the pulse amplitude of the pho-
`toplethysmograms detected by a reflectance pulse ox-
`imeter sensor.
`One of the requirements in designing a reflectance pulse
`oximeter sensor is to determine the optimum separation
`distance between the LED’s and the photodiode. Ob-
`viously, this distance should be selected such that pho-
`toplethysmograms with maximum pulsatile components
`could be detected. Generally, the pulsatile component of
`the reflected photoplethsmograms depends not only on the
`systolic blood pulse in the peripheral vascular bed but also
`on the amount of arterial blood within the illuminated tis-
`sue volume.
`The selection of each LED driving current determines
`the effective penetration depth of the incident light. For a
`given LED/photodiode separation, it it clear that with
`higher levels of incident light, a larger pulsatile vascular
`bed will be illuminated. Consequently, the reflected pho-
`toplethysmograms will contain a larger ac component.
`Practical considerations, however, limit the driving cur-
`rent of each LED to the manufacturer specified maximum
`power dissipation. Alternatively, by placing the photo-
`detector too close to the LED’s, the large dc component,
`which is mainly due to multiple scattering of the incident
`photons by the blood-free stratum corneum and epidermis
`layers in the skin, will cause the photodetector to become
`saturated.
`It is important to point out that although the HP ear
`oximeter which was used as a reference in our studies is
`not an acceptable primary standard for measuring SaO,?,
`its accuracy and reliability as a noninvasive oximeter have
`been widely established [ 1 11-[ 131.
`Our experience using the prototype reflectance sensor
`has shown that the pulse amplitude of the reflection pho-
`toplethysmograms depends among other factors on the
`position of the photodiode relative to the LED’s. The se-
`lection of a particular separation distance, however, in-
`volves a tradeoff. On one hand, larger photoplethysmo-
`grams can be detected by mounting the photodiode further
`
`; p . J - - - - ~
`
`SKIN REFLECTANCE OXIMETER
`
`60’
`
`Fig. 10. Simultaneous recordings of SaO, from the HP ear oximeter, Nell-
`cor pulse oximeter and the prototype reflectance pulse oximeter.
`
`R
`
`IR
`
`Fig. 11. Reflection photoplethysmograms recorded from the forearm using
`a combination of three photodiodes. The circles indicate the relative lo-
`cation of the photodiodes with respect to the LED’s (*). The closed
`circles indicate the photodiodes which were used to collect the reflected
`light as shown by the corresponding traces.
`
`several photodetectors, considerably larger photople-
`thysmograms could be detected.
`To demonstrate the advantage of using multiple pho-
`
`Page 7
`
`

`

`MENDELSON AND OCHS: NONINVASIVE PULSE OXIMETRY
`
`apart from the LED’s. On the other hand, higher LED
`driving currents are necessary to overcome the absorption
`of the incident light due to a longer optical path length.
`The results of our studies also validated our hypothesis
`that skin heating is a feasible method for increasing the
`size of the reflected photoplethysmograms. We noticed
`that by heating the skin surface to 45”C, a five-fold in-
`crease in the pulsatile component could be achieved. We
`noticed also that the improvement due to skin heating can
`last up to 20 min from the time the temperature of the skin
`has reached 45°C and the heater was turned off. It is im-
`portant to mention that the ability to measure accurate
`SaOz values by the prototype pulse oximeter sensor was
`independent of the exact skin temperature. We found that
`a minimum skin temperature of approximately 40°C is
`generally sufficient in order to detect adequate stable pho-
`toplethysmograms. Our experience in healthy adults has
`shown that at this temperature the heated sensor can re-
`main in the same location for at least three hours without
`any apparent skin damage. It should be noted also that the
`principal objective of sk