DESIGN AND EVALUATION OF A REFLECTANCE OXYGEN SENSOR IN CRITICALLY ILL PATIENTS Setsuo Takatani, George P. Noon, Yukihiko Nose, and Michael E. DeBakey Department of Surgery, Baylor College of Medicine Houston, TX 77030, USA INTRODUCTION The optical pulse oximeter has gained broad clinical applications in rapidly and noninvasively measuring arterial hemoglobin oxygen saturation (Sa02)(Yelderman and New, 1983; Severinghaus, 1987; pologe, 1987). However, major shortcoming of the current transmission oximeters is the requirement of a cuvette through which light signals must be transmitted, thus limiting its application to fingertip or earlobe. More broad applications can be achieved with the reflectance mode. The reflectance pulse oximeter, as reported by Mendelson et al (1983), was the first to measure Sa02 from the fingertip. Since then, several reflectance pulse oximeters have been reported, but no clinical application has of yet been reported (Mendelson et aI, 1988; Mendelson and Ochs, 1988). The major difficulty of reflectance pulse oximeters is small pulsatile signal level, particularly at red wavelength, in comparison to the transmission mode. Also, sensor characteristics, and physiological factors such as arterial-to-venous blood volume distribution, local venous saturation, blood flow, tissue inhomogeneity, optical properties of tissue, etc, all affect the accuracy of the measurement. For proper design of the reflectance sensor, the effects of these variables on the Sp 02 computation methodology must be well understood. In this study, theoretical modeling of reflectance pulse oximetry was undertaken using the three-dimensional photon diffusion theory whose applicability to oximetry in whole blood and tissue has been well established (Takatani and Graham, 1979; Reynolds et al, 1976; Cohen and Longini, 1971). The reflectance pulse oximeter sensor was then designed, tested in animals, and used in the critically ill patients who were undergoing open heart and lung surgery. REFLECTANCE PULSE OXIMETRY THEORETICAL MODELING In modeling the reflectance pulse oximetry, the threedimensional photon diffusion theory was used. For the sensor geometry of Fig. I, the diffuse reflectance equation was
Oxygen Transport to Tissue XlV, Edited by W. ErdmlUUl and D.F. Bruley, Plenum Press, New York, 1992
247
Differential Equation,
(V 2
-
f.lD) p. (r, 9, z) = - (f.lD)So
Solution, Rx
UGHT DETECTOR
R(rJ
=~[_S_I I
dS+K".,
A~[1 _ e-dlOO(_1)"]
Diffuse reflectance,
Rd = [a2/(~ - r,))[R(r~ - R(r,)].
Fig. 1. Sensor model and diffuse reflectance equation.
described earlier by Takatani (1989). Equations 1 through 6 from Takatani were used to derive diffuse reflectance from tissue in terms of Sa02, Sv02, arterial and venous hemoglobin content ([Hb]a, [Hb]v). In addition, venous hemoglobin oxygen saturation (Sv02) was expressed as a function of Sa02, total hemoglobin content ([Hb]), blood flow (Q) and tissue metabolism (V02) by the following equation; Sv02 = Sa02 - V02 / { 1.32cc/g x [Hb] x Q }
(1 )
Thus, the Equation (1) allows to include the effect of the local blood flow and metabolism on the Sa02 measurement. The effects of sensor design including wavelength selection and sensor geometry, physiological parameters such as arterial-tovenous blood volume distribution, venous saturation, blood flow, tissue metabolism, tissue optical characteristics were studied to optimize the sensor design and Sa02 computation algorithm. The theoretical results revealed that dual wavelengths, 665 and 820 nm, in combination with the following computation algorithm will yield the best results (Fig. 2); Sa02 = A x Ratio + B,
(2 )
where A and B are the constants that depend on the sensor geometry and physiological parameters, and Ratio is given by; Ratio = (AC/DC)665nm / (AC/DC)S20nm with AC and DC being defined as the pulsatile amplitude and average level of the signals at each wavelength _ Th is methodology can also minimize the effects of physiological variables such as arterial-to-venous distribution, venous saturation, tissue optical characteristics. REFLECTANCE PULSE OXIMETER SYSTEM Optical Sensor Fig. 3a and 3b show the prototype optical sensor and its
248
schematics. The optical sensor consists of four light emitting diode (LED) chips for each wavelength spaced equally around the photodiode chip to sample tissue spectra. Four chips for each wavelength were employed to enhance signal-to-noise ratio and to average out inhomogeneous effect of tissue. An optical barrier was placed between the LEO and photodiode to prevent direct coupling effect and control the detection depth in tissue. The current sensor can detect the light returning from 0.5 rom or deeper in the tissue. This design, thus, minimizes the multiple scattering effect of the shallow layer,
100.---__~--------~~--------__.
90
70
~~-r~~-rOT~-r~~-r~~~~
0.2
0.4
0.6
0.8
1.2
1.4
RATIO
Fig. 2. Comparison of three Sa02 computation algorithms. AC/OC model gives the best linear relation with respect to Sa0 2·
particularly at the vicinity of the light source and detector and measures diffusely scattering light in the deeper region. In addition to measuring wavelengths of 665 and 820 nm, nearinfrared wavelength (940 nm) LEOs were also included in the sensor: these are used to warm the tissue in case the tissuesensor interface temperature decreases. A thermocouple was mounted at the sensor surface to continuously monitor the sensor-tissue interface temperature. The dimension of the sensor is approximately 1 cm square and it is placed at the center of a plexi-glass mount (3 cm diameter) whose main purpose is to prevent heat loss to the ambient. Oximeter System Fig. 4 shows the block diagram of the oximeter system. The LEOs are excited sequentially with a narrow width pulse (10 micro-second) at 1 kHz. The measured reflectance from tissue is
249
Optical
Optical
8llnier
Bemer
a a D
Subltnlte
LED TOP VIEW
SIDE VIEW
b
a
Fig. 3. Prototype reflectance pulse oximeter sensor (a) and schematic diagram of the sensor (b).
TISSUE
MAIN OXIMETER UNIT proq. Gain Aq).,
SIB, Multiplexer, l/V
De~ultiplu.er,
Pre-Amp
LPF Circuiu
Hitsubishi "MAXY" PC
'I'aIp. ~(r)
Fig. 4. Block diagram of the reflectance pulse oximeter control and data acquisition system.
amplified, smapled, and AC and DC components of each wavelength are computed. A personal computer, Maxy (Mitsubishi), is used to compute the Sa02 on-line using the Equation (2) with predetermined A and B constants. RESULTS AND DISCUSSION Prior to clinical applications, the prototype sensor was evaluated in animals. The mongrel dogs with weight ranging from 10 to 20 Kg were anesthetized and their respiration was controlled. The femoral artery was cannulated to obtain blood samples and to measure the Sa02 and [Hb] using an IL-282 250
bench-top CO-Oximeter. A commercially available transmission pulse oximeter was placed on the earlobe to obtain a reference Sa02. The oxygen content of the gas ventillating the dogs was varied to alter Sa02. Fig. 5 shows the correlation plot between the Sa02 measured by the reflectance pulse oximeter vs. those by the IL-282 CO-Oximeter. For clinical evaluation of the sensor, the critically ill patients who might have low Sa02 such as lung and heart diseases were selected. After obtaining the consent from the patients, the sensor was placed on the forehead or cheek with a double-sided tape where the best pulsatile signals were obtainable.
100 N =83 Y =0.909 X + 7.11 r = 0.9823
tIP
90 N
0
III
80
~
70
til
U
Z
0(
Eo! U
60
r.. ~
50
~ ~
40
40
50
60
70
80
90
100
IL-282 Sa0 2 ( %)
Fig. 5. Correlation plot between the Sa02 by the reflectance pulse oximeter and those by the IL-282 CO-Oximeter in dogs.
A total of 18 patients were studied in the operating room over the duration of 4-5 hours; of the patients studied, two were black and one was a 4 day old infant. Fig. 6 shows the correlation plot between the Sa02 measured by the reflectance pulse oximeter vs. those analyzed by the IL-282 CO-Oximeter. The errors may have developed during blood sampling, because the patient's Sa02 were unstable. Although the better accuracy was obtained in dog experiments, dog experiment was well controlled and Sa02 was stable during blood sampling. Concerning the application site, cheek showed better signal level and stability over 4-5 hour monitoring. As for the sensor-skin interface temperature and signal level, when the interface temperature was maintained at or above 35 C, usually there was no problem in detection of signal from the cheek. Since the head is exposed to the anesthesiologist during surgery, application of the sensor is easier in comparison to the fingertip. Also, since the circulation to the cheek or forehead is related to the circulation to the head, the 251
100,---------------------------~
x
95 N
oCIS
III
x 90
x
x x
x xx x
xxx
x
x
x 85 N=38 Y =0.9424 X+ 4.22 r =0.926 OO~~~,-~~,-~~,-~_r,-~~
00
85
90
IL-282 Sa02
95
100
(%)
Fig. 6. Correlation plot between the Sa02 by the reflectance pulse oximeter and those by the IL-282 CO-Oximeter in critically ill patients. monitoring from the head area may be indicative of the circulation to the vital organ. The reflectance pulse oximeter can be a powerful alternative to the transmission oximeter in the operating room and intensive care unit. ACKNOWLEDGEMENT This research was partially supported by a grant in aid from the Colin Electronics Inc., Komaki, Japan. REFERENCES Cohen, A. and Longini, R.L., 1971, Theoretical determination of the blood's relative oxygen saturation in vivo. Med BioI Eng 9:61-69. Mendelson, Y., Cheung, P.W., Neuman, M.R., Fleming, D.G. and Cahn, S.D., 1983, Spectrophotometric investigation of pulsatile blood flow for transcutaneous reflectance oximetry. Oxygen Transport to Tissue VI:93-102. Mendelson, Y., Kent, J.C., Yocum, B.L. and Birle, J., 1988, Design and evaluation of a new reflectance pulse oximeter sensor. Medical Instrumentation 22(4):167-173. Mendelson, Y. and Ochs, B.D., 1988, Noninvasive pulse oximetry utilizing skin reflectance photoplethysmography. IEEE Trans Biomed Engr 35(10):798-805. Pologe, J.A., 1987, Pulse oximetry; Technical aspects of machine design. Int Anesthesiol Clin 25:137-153. severinghaus, J.W. and Naifeh, K.H., 1987, Accuracy of response of six pulse oximeters to profound hypoxia. Anesthesiology 67:551-558. 252
Reynolds, 1.0., Johnson, C.C., and Ishimaru, A., 1976, Diffuse reflectance from a finite blood medium: Application to the modeling of fiber optic catheters. Appl Opt 15:2059-2067. Takatani, S. and Graham, M.D., 1979, Theoretical analysis of diffuse reflectance from a two-layer tissue model. IEEE Trans Biomed Engr 26(12):656-664. Takatani, S., 1989, Toward absolute reflectance oximetry: I. Theoretical consideration for noninvasive tissue reflectance oximetry. Oxygen Transport to Tissue XI:91-102. Yelderman, M. and New W., 1983, Evaluation of pulse oximetry. Anesthesiology 59:349-352.
253
Oxygen Transport to Tissue XlV, Edited by W. ErdmlUUl and D.F. Bruley, Plenum Press, New York, 1992
247
Differential Equation,
(V 2
-
f.lD) p. (r, 9, z) = - (f.lD)So
Solution, Rx
UGHT DETECTOR
R(rJ
=~[_S_I I
dS+K".,
A~[1 _ e-dlOO(_1)"]
Diffuse reflectance,
Rd = [a2/(~ - r,))[R(r~ - R(r,)].
Fig. 1. Sensor model and diffuse reflectance equation.
described earlier by Takatani (1989). Equations 1 through 6 from Takatani were used to derive diffuse reflectance from tissue in terms of Sa02, Sv02, arterial and venous hemoglobin content ([Hb]a, [Hb]v). In addition, venous hemoglobin oxygen saturation (Sv02) was expressed as a function of Sa02, total hemoglobin content ([Hb]), blood flow (Q) and tissue metabolism (V02) by the following equation; Sv02 = Sa02 - V02 / { 1.32cc/g x [Hb] x Q }
(1 )
Thus, the Equation (1) allows to include the effect of the local blood flow and metabolism on the Sa02 measurement. The effects of sensor design including wavelength selection and sensor geometry, physiological parameters such as arterial-tovenous blood volume distribution, venous saturation, blood flow, tissue metabolism, tissue optical characteristics were studied to optimize the sensor design and Sa02 computation algorithm. The theoretical results revealed that dual wavelengths, 665 and 820 nm, in combination with the following computation algorithm will yield the best results (Fig. 2); Sa02 = A x Ratio + B,
(2 )
where A and B are the constants that depend on the sensor geometry and physiological parameters, and Ratio is given by; Ratio = (AC/DC)665nm / (AC/DC)S20nm with AC and DC being defined as the pulsatile amplitude and average level of the signals at each wavelength _ Th is methodology can also minimize the effects of physiological variables such as arterial-to-venous distribution, venous saturation, tissue optical characteristics. REFLECTANCE PULSE OXIMETER SYSTEM Optical Sensor Fig. 3a and 3b show the prototype optical sensor and its
248
schematics. The optical sensor consists of four light emitting diode (LED) chips for each wavelength spaced equally around the photodiode chip to sample tissue spectra. Four chips for each wavelength were employed to enhance signal-to-noise ratio and to average out inhomogeneous effect of tissue. An optical barrier was placed between the LEO and photodiode to prevent direct coupling effect and control the detection depth in tissue. The current sensor can detect the light returning from 0.5 rom or deeper in the tissue. This design, thus, minimizes the multiple scattering effect of the shallow layer,
100.---__~--------~~--------__.
90
70
~~-r~~-rOT~-r~~-r~~~~
0.2
0.4
0.6
0.8
1.2
1.4
RATIO
Fig. 2. Comparison of three Sa02 computation algorithms. AC/OC model gives the best linear relation with respect to Sa0 2·
particularly at the vicinity of the light source and detector and measures diffusely scattering light in the deeper region. In addition to measuring wavelengths of 665 and 820 nm, nearinfrared wavelength (940 nm) LEOs were also included in the sensor: these are used to warm the tissue in case the tissuesensor interface temperature decreases. A thermocouple was mounted at the sensor surface to continuously monitor the sensor-tissue interface temperature. The dimension of the sensor is approximately 1 cm square and it is placed at the center of a plexi-glass mount (3 cm diameter) whose main purpose is to prevent heat loss to the ambient. Oximeter System Fig. 4 shows the block diagram of the oximeter system. The LEOs are excited sequentially with a narrow width pulse (10 micro-second) at 1 kHz. The measured reflectance from tissue is
249
Optical
Optical
8llnier
Bemer
a a D
Subltnlte
LED TOP VIEW
SIDE VIEW
b
a
Fig. 3. Prototype reflectance pulse oximeter sensor (a) and schematic diagram of the sensor (b).
TISSUE
MAIN OXIMETER UNIT proq. Gain Aq).,
SIB, Multiplexer, l/V
De~ultiplu.er,
Pre-Amp
LPF Circuiu
Hitsubishi "MAXY" PC
'I'aIp. ~(r)
Fig. 4. Block diagram of the reflectance pulse oximeter control and data acquisition system.
amplified, smapled, and AC and DC components of each wavelength are computed. A personal computer, Maxy (Mitsubishi), is used to compute the Sa02 on-line using the Equation (2) with predetermined A and B constants. RESULTS AND DISCUSSION Prior to clinical applications, the prototype sensor was evaluated in animals. The mongrel dogs with weight ranging from 10 to 20 Kg were anesthetized and their respiration was controlled. The femoral artery was cannulated to obtain blood samples and to measure the Sa02 and [Hb] using an IL-282 250
bench-top CO-Oximeter. A commercially available transmission pulse oximeter was placed on the earlobe to obtain a reference Sa02. The oxygen content of the gas ventillating the dogs was varied to alter Sa02. Fig. 5 shows the correlation plot between the Sa02 measured by the reflectance pulse oximeter vs. those by the IL-282 CO-Oximeter. For clinical evaluation of the sensor, the critically ill patients who might have low Sa02 such as lung and heart diseases were selected. After obtaining the consent from the patients, the sensor was placed on the forehead or cheek with a double-sided tape where the best pulsatile signals were obtainable.
100 N =83 Y =0.909 X + 7.11 r = 0.9823
tIP
90 N
0
III
80
~
70
til
U
Z
0(
Eo! U
60
r.. ~
50
~ ~
40
40
50
60
70
80
90
100
IL-282 Sa0 2 ( %)
Fig. 5. Correlation plot between the Sa02 by the reflectance pulse oximeter and those by the IL-282 CO-Oximeter in dogs.
A total of 18 patients were studied in the operating room over the duration of 4-5 hours; of the patients studied, two were black and one was a 4 day old infant. Fig. 6 shows the correlation plot between the Sa02 measured by the reflectance pulse oximeter vs. those analyzed by the IL-282 CO-Oximeter. The errors may have developed during blood sampling, because the patient's Sa02 were unstable. Although the better accuracy was obtained in dog experiments, dog experiment was well controlled and Sa02 was stable during blood sampling. Concerning the application site, cheek showed better signal level and stability over 4-5 hour monitoring. As for the sensor-skin interface temperature and signal level, when the interface temperature was maintained at or above 35 C, usually there was no problem in detection of signal from the cheek. Since the head is exposed to the anesthesiologist during surgery, application of the sensor is easier in comparison to the fingertip. Also, since the circulation to the cheek or forehead is related to the circulation to the head, the 251
100,---------------------------~
x
95 N
oCIS
III
x 90
x
x x
x xx x
xxx
x
x
x 85 N=38 Y =0.9424 X+ 4.22 r =0.926 OO~~~,-~~,-~~,-~_r,-~~
00
85
90
IL-282 Sa02
95
100
(%)
Fig. 6. Correlation plot between the Sa02 by the reflectance pulse oximeter and those by the IL-282 CO-Oximeter in critically ill patients. monitoring from the head area may be indicative of the circulation to the vital organ. The reflectance pulse oximeter can be a powerful alternative to the transmission oximeter in the operating room and intensive care unit. ACKNOWLEDGEMENT This research was partially supported by a grant in aid from the Colin Electronics Inc., Komaki, Japan. REFERENCES Cohen, A. and Longini, R.L., 1971, Theoretical determination of the blood's relative oxygen saturation in vivo. Med BioI Eng 9:61-69. Mendelson, Y., Cheung, P.W., Neuman, M.R., Fleming, D.G. and Cahn, S.D., 1983, Spectrophotometric investigation of pulsatile blood flow for transcutaneous reflectance oximetry. Oxygen Transport to Tissue VI:93-102. Mendelson, Y., Kent, J.C., Yocum, B.L. and Birle, J., 1988, Design and evaluation of a new reflectance pulse oximeter sensor. Medical Instrumentation 22(4):167-173. Mendelson, Y. and Ochs, B.D., 1988, Noninvasive pulse oximetry utilizing skin reflectance photoplethysmography. IEEE Trans Biomed Engr 35(10):798-805. Pologe, J.A., 1987, Pulse oximetry; Technical aspects of machine design. Int Anesthesiol Clin 25:137-153. severinghaus, J.W. and Naifeh, K.H., 1987, Accuracy of response of six pulse oximeters to profound hypoxia. Anesthesiology 67:551-558. 252
Reynolds, 1.0., Johnson, C.C., and Ishimaru, A., 1976, Diffuse reflectance from a finite blood medium: Application to the modeling of fiber optic catheters. Appl Opt 15:2059-2067. Takatani, S. and Graham, M.D., 1979, Theoretical analysis of diffuse reflectance from a two-layer tissue model. IEEE Trans Biomed Engr 26(12):656-664. Takatani, S., 1989, Toward absolute reflectance oximetry: I. Theoretical consideration for noninvasive tissue reflectance oximetry. Oxygen Transport to Tissue XI:91-102. Yelderman, M. and New W., 1983, Evaluation of pulse oximetry. Anesthesiology 59:349-352.
253
