A fiber optic PCO2 sensor

Annals of BiomedicalEngineering, Vol. 11, pp. 499-510, 1983

0090-6964183 $3.00+ .00 Copyright 9 1984 Pergamon Press Ltd.

Printed in the USA. All rights reserved.

A FIBER OPTIC

PCO2 S E N S O R

G. G. Vurek P. J. Feustel J. W. Severinghaus Laboratory of Technical Development National Heart, Lung, and Blood Institute Bethesda, Maryland and Cardiovascular Research Institute University of California San Francisco, California

The theory, construction and performance o f a catheter tip optical PCO 2 probe is described. The sensor, called the Opticap, is made with plastic fiber optics. One fiber carries light to the sensitive tip which is a silicone rubber tube O.6 m m dia. • 1.0 m m long filled with a phenol red-KHCOssolution. Ambient PCO2controls the p H o f the solution which influences the optical transmittance o f the phenol red. A second fiber carries the transmitted signal to a receiver; the resulting electrical signal is linearly related to the PC02 over the range o f 2. 7 to 10. 7 kPa. The probe was tested as a tissue PC02 sensor on the cerebral cortex o f the cat and as an arterial PCO 2sensor. Drift over one day's use was O.6 KPa or less and individual probes have been used as long as 12 weeks.

Keywords - - Carbon dioxide, Fiber opt&, Optical, Blood gas measurement.

INTRODUCTION This paper describes a miniature PCO2 sensor which has a sensitive volume smaller than 0.5 microliter and can be used for blood and tissue PCO2 measurements. In common with other PCO2 measurement techniques we determine Address correspondence to G. G. Vurek, Laboratory of Technical Development, National Heart, Lung, and Blood Institute, Building 10, Room 5D-20, Bethesda, Maryland 20205. Acknowledgement--The authors thank Drs. J. I. Peterson and S. R. Goldstein of the National Institutes of Health for their advice and suggestions, and Ms. Mary Stafford of the University of California, San Francisco, for her technical assistance. 499

500

Vurek et al.

PCO2 by measuring the pH of a bicarbonate buffer solution in equilibrium with the local PCO2 but we use a colorimetric pH indicator. The optical approach avoids the problems of making miniature glass or liquid membrane PCO2electrodes (1) with adequate reference electrode stability and facilitates other electrometric measurements in the vicinity such as blood flow measurement by bare platinum hydrogen gas clearance electrodes (10). Liabbers and Opitz (4) described an optical approach to PCO2 measurement using pH sensitive fluorescent dyes and described sensors for gas and surface PCO2 measurement. Peterson and Goldstein (7) described a fiber optic pH sensor using phenol red dye covalently bound to acrylamide gel contained within ion and water permeable tubing. The present work is an extension of Peterson and Goldstein's work but a solution of indicator dye contained within a gas permeable, ion impermeable silicone rubber tube is used as the PCO2 sensor. THEORY In this section we show that the magnitude of the optical signal can be nearly linearly related to the PCO2. Three relationships are used: Eq. 1 the relation between pH and PCO2 in a solution containing bicarbonate (the HendersonHasselbalch relation); Eq. 2 the relation between the dissociation constant of the dye, Kd, the concentration of the dye base form, B, total dye concentration, D, and pH; and Eq. 3 the relation between the normalized optical transmittance signal, T, and the optical absorbance A, path length L, extinction, e, and B (the Beer-Lambert relation). Here the operator p is taken as the negative logarithm function, and the units of PCO2 are mm Hg (1 mm Hg = 0.133 kPa), and the units of concentration are millimoles/liter (mM). pH = 6.1 + log (HCO~-) - log (.03 PCO2) pKd = pH + log (D/B - 1) A = -logT=BLe

(1) (2) (3)

Manipulating Eq. 1 leads to Eq. 4, if (HCO~-) = 35 raM: pH

= 9.17-1ogPCO2

(4)

Rearranging Eq. 2 and using 7.5 for the value of pKd leads to Eq. 5: B

= D/(1 + 10exp(7.5 - pH))

(5)

Combining Eqs. 3, 4 and 5 leads to Eq. 6: T

= 10exp(-DLe/(1 + 10exp(log P C O 2 - 1.667)))

(6)

Fiber Optic PC02 Sensor

501

Values for T were computed by inserting values of PCO: over the range from 0.1 to 16 kPa at values for DLe = 1,2, and 4 into Eq. 6 and normalized to unity at a PCO2 of 5.3 kPa. Figure 1 shows a plot of manually fitted curves of those values. E X P E R I M E N T A L APPARATUS We will outline the construction of the probe, light source, and detector. Probes have been hand-made and show some variation from unit to unit but once the manufacturing technique has been mastered probes are easily made and used providing that the user recognizes the need for calibrating each. First we will describe the general features of the system and then we will describe some of the specific features of our implementation of the concept. The CO: sensitive portion of the system is an isotonic saline/bicarbonate solution with the pH indicator dye, phenol red, enclosed in a silicone rubber tube. Two optical fibers, one from the light source and one to the photodetector are at

)Lr = 4

3.0 DL6=2

tj C '4"-

rE r

~2.0 DLc=I

N O

IE z 1.0

r DLr 7 T= io-Ll+lO(Jo~Pco=-I.ssT) ]

i

I

0 0

5.5 40

kPo mmHg PC02

10.6 80

J

16.0 120

FIGURE 1. Plot of the relation between t r a n s m i t t a n c e , T, and PCO 2 for three values of the parameter DLe. Curves are normalized to unity transmittance at a PC02 o f 5.3 kPa, and are calculated for (HCO~) 3 5 m M/L and pKd = 7 . 5 0 . Normalization factors for DLe = 1, 2, 4, are 1 / 3 . 4 4 6 , 1 / 1 1 . 8 7 , and 1 / 1 4 1 . 0 respectively.

502

Vurek et aL

one end of the tube and a diffuse reflector of silicone rubber seals the other end. Figure 2 illustrates the relation between the light source, detector and the two fibers. The length of the probe between the sensitive tip and the bifurcation of the two fibers is about 60 cm and there is an additional 18 cm between the bifurcation and source and detector. The light source is a 100 W tungsten-halogen lamp operated from an adjustable, stabilized source. Heat absorbing and visible wavelength selecting filters and lenses direct light onto the fiber which carries it to the sensor tip. The detector is a phototransistor operating as a photovoltaic diode into a current-to-voltage converter with a gain of 101~volts/amp.

Probes

The construction of a probe will now be described. Two 80 cm lengths of plastic optical fiber (duPont Crofon, 0.14 mm dia.) were threaded through a 60 cm piece of 0.75 mm o.d. by 0.33 mm i.d. black Teflon tubing which protected the fibers from mechanical damage and from room light pick-up. One end of the tubing was threaded into a short no. 18 needle (a strain relief for the bifurcation). Separate pieces of Teflon tubing were threaded over the fibers extending from the no. 18 needle hub. The fiber ends were put into No. 25 needles from the hub end and flared with gentle heat from a hot wire to polish the surface and to hold them mechanically in the needle tip. The needles were used to position the fibers at the focus of the light source and at the sensitive part of the detector. Figure 3 shows the details of the tip. The two fibers extending from the other end of the 60 cm piece of black tubing were sealed into it with epoxy and were cut with a sharp blade. An additional layer of epoxy was put over the cut ends to prevent the dye from binding to the fibers. A coating of clear silicone adhesive was put over the epoxy, except at the fiber tips. A piece of silicone rubber tubing, 0.6 mm o.d. by 0.3 mm i.d. by 1.2 mm long, and plugged at one end with white

LAMP SUPPLY Holder

101~ OUT -150

.j

~-

~.01

FIGURE2. Block diagram of the sensor system. See text for details.

TIP

Fiber Optic P C O 2 Sensor

503

silicone adhesive was filled with a solution of phenol red, KHCO3, and KC1. Phenol red has a PKd around 7.57; the exact value depends on the ionic strength of the solution, temperature and other variables. For best linearity, the solution pH should be adjusted to be equal to the dye PKd at the midpoint of the range of PCO2's to be measured. From the discussion of the previous section, the bicarbonate concentration for a nominal PCO2 of 5.3 kPa (40 mm Hg) is 35 mM. Enough salt was added to make the solution isotonic. The filled silicone tube was attached to the tip with silicone adhesive (General Electric RTV-112) which releases acetic acid during curing. In order to prevent the acid from diffusing into the bicarbonate solution and changing its composition, the filled tip was surrounded with a capillary filled with the solution during the initial 30 mi.n of the adhesive curing process. The sensor tip was stored under undyed bicarbonate solution except when being tested, calibrated, or used.

Light Source

We used an adjustable regulated AC power supply (Oriel Corp. Model 6329) to feed a 100 W tungsten-halogen lamp (12 V, 8.3A). The lamp housing has an adjustable focus lens which was used to image the filament at the focus of a pair of aspheric lenses with a filter set between them. A 5 mm thick heat absorbing filter was placed in the light path to protect the plastic fibers. The aspheric lenses (Edmond Scientific No. DCX 36 x 40, also heat absorbing) were spaced about 1 cm apart so that the light passing between them is roughly collimated. A twoposition sliding filter holder was placed in this space. One position held an interference filter (570 nm center wavelength, 18 nm half-width, 28% peak transmittance) and the other held a piece of red transmitting gelatin filter (Kodak Wratten No. 70) with a piece of fine brass screen to provide additional attenuation. A holder for the needle with the sending fiber kept the fiber tip at the focus of the filtered light. TEFLON SHEATH

BUFFER SPACE /

/ FROM LAMP TO DETECTOR

/

WHITE SILICONE ADHESIVE

FIGURE 3. Construction of sensor tip. See text for details.

SILICONE ,/TOBE

504

Vurek et al. Light Detector

A Fairchild FPT-100 phototransistor was modified by drilling a hole into the plastic lens over the silicon chip. A piece of needle tubing was epoxied into the hole so that the receiver needle would fit into it and force the end of the fiber against the bottom of the hole, quite close to the silicon chip. The transistor was operated in the photovoltaic mode, with the emitter floating, base grounded, and collector connected to an FET input amplifier with a gain of 101~volts/amp. The output offset voltage was less than 5 mV, and the dark noise was less than 2 mV peak-to-peak at a 0-1 Hz bandwidth. Typical signals from the green channel at a PCO2 of 4.5 kPa (34 mm Hg) ranged from 0.5 to 2 volts, and the red signals ranged from 2 to 6 volts, depending on the characteristics of the individual probe. The phototransistor and amplifier were mounted in a small metal box with the needle tube sticking out. TESTS

In Vitro

Probes were tested for their response to various concentrations of CO2, 02, and N2 as well as temperature sensitivity, and response time. A thermostatted water bath, usually at 37C, was used to heat a two-stage gas hydrator containing undyed buffer. The probe was inserted in the second stage, which was covered with black tape to prevent room light from entering the probe and introducing an error. Mass spectrometrically calibrated gases with CO2 content ranging from 3.03% CO2 to 15.2% CO2 were bubbled through the hydrator. Calibrations were done with five gases with CO2 content ranging from 3 % to 15 % (nominal); a typical set was: 3.03%, 4.84%, 7.13%, 9.84%, 14.94%. Corrections for barometric pressure and PH20 were made to calculate PCO2. At least ten min. were allowed for equilibrium to be achieved before making a measurement. First the red signal was obtained, and then the green. The filter was left in the green transmitting position except when a red measurement was being made. The ratio of the green to red signal was plotted versus PCO2. For in vivo use this procedure served as a calibration of the probe. The temperature response was determined by changing the setting of the waterbath thermostat. Separate calibration curves were made at 31C, 38C, and 42.5C to estimate the influence of temperature. Time responses were measured at room temperature (23C). Separate gas washing bottles were supplied with 3.05% CO2 and 9.91% CO2. The probe tip was mounted in an opaque tube that could be switched quickly from one bottle outlet to the other. Only the green signal was recorded on a potentiometric recorder.

Fiber Optic P C O 2 Sensor

505 ]n Vivo

One of the primary purposes in developing this sensor was to estimate the PCO2 of the tissue fluid in the vicinity of the respiratory control center in the cat medulla. We simulated this measurement by comparing the response of a wellcharacterized surface PCO2 sensor (8) placed against the exposed cortex of a cat with the response of an Opticap probe placed between the surface sensor and the cortex. In four separate tests cats (approx. 3 Kg) were anesthetized with phenobarbital and paralyzed with Flaxedil (Davis-Geck). Respiration was controlled with a pneumatic ventilator and monitored with a mass spectrometer connected to the tracheal tube. Information from the mass spectrometer was collected and displayed by the laboratory's computer system. A transcutaneous CO2 sensor temperature was set to 37~ and it was calibrated before and after the test. The sensor was placed on the exposed cortex of the animal. After calibration, the Opticap probe was placed directly under the center of the transcutaneous sensor; this was verified by observation of the light through the glass of the sensor. Both probe and sensor tensions were measured on samples withdrawn from a catheter in the aorta via a femoral artery. Inspired gas compositions were changed and the response of the Opticap and transcutaneous sensor were compared. Another potential application of the Opticap is intravascular monitoring. Another set of four anesthetized cats, with controlled and monitored respiration as above, were prepared with a blood sampling catheter in the descending aorta via the femoral artery and an Opticap probe in the aorta via a carotid artery. Arterial gas composition was changed and, after steady state was achieved, as indicated by end-tidal PCO2 measurements, the output of the Opticap was recorded and three mL blood samples were taken for PCO2 measurement on a calibrated blood-gas analyzer.

RESULTS

As was expected from the theoretic analysis above and from the variation among the hand-made probes, the relation between the probe response and PCO2 was not exactly linear. But over the range of calibration gases used, 2.7 to 14 kPa, the departure of the response from linearity was less than 10 percent and over the range from 3.5 to 14 kPa, less than five percent. A typical calibration curve had a slope of 0.252 green/red units per kPa, intercept of 0.011 green/red units, and a correlation coefficient of r = .9995 (n = 5). The time response to a step change in PCO2 can be approximated by a double exponential, with the fast component corresponding to a time constant of 20 sec and the second to one of 2.5 min. That is, 95% of the response to step change is achieved in one min. and the final response is achieved in less than 10 min.

506

Vurek et al.

The temperature response, measured over the range of 31 to 42.5C on one probe was not a simple function; both slope and intercept changed with temperature. Increasing temperatures gave decreasing slope and increasing intercept. The mean slope, estimated from linear fits to the calibration curves made at the three temperatures over the range of 3% CO2 to 7% CO2 changed from 5.1 x 10-1 green/red units per Pa at 31C to 3.5 x 10-1 at 42.5C for an average change of 3.7% K - ~.The intercept changed from 0.01 units to 0.022 units over the same range. The calibration curve for each probe was measured before and after use for probe performance tests or experiments on the respiratory control center. There was some drift over periods of several hours, usually less than one percent/hr or 0.6 kPa/6 hours. Well made probes could be used for at least two months, although the calibration curves did shift toward increasing transmission during that time. The shift was principally in the green signal; this could have been due to several causes including loss of bicarbonate due to microbial activity or inward diffusion of organic acids through the slightly permeable membrane but the most likely cause seemed to be adsorption of the dye on inadequately epoxy coated fiber tips. Inspection of probes after they were no longer usable showed dye bound to the fiber tips. Because the light source shifts its spectral characteristics with applied power, the green energy increasing relatively more than the red with increasing lamp voltage, the effect of lamp voltage fluctuation was measured. At a mean lamp voltage of 11 volts, the red signal increased about 2%/% change in voltage and the green increased about 3 %/%; the ratio increased about 1%/%. The particular supply we used (Oriel Model 6329) drifted less than 1/2% after a one hour warm-up. The transmittance of the red filter was temperature sensitive. This manifested itself as a decrease in the red signal as the lamp housing warmed up. We did not determine whether this effect was due to a slight mechanical shift or to a direct effect on the filter properties. Probably it was a combination of both, because a piece of the filter was observed to darken when placed at the focus of the first lens of the optical train. We wanted to compare the response of the Opticap to a surface (transcutaneous) sensor because the latter had been used to estimate brain tissue PCO2 in other studies. The surface probe was too bulky for studies with the medullary respiratory control center and that was one of the reasons the Opticap was developed. Figure 4 shows a correlogram between the Opticap response and the surface sensor placed together on the surface of a cat's cortex. The two sensors gave readings within 0.6 kPa, but because the surface sensor measures over a larger area than the Opticap, it was not possible to accurately attribute the differences to either systemic or random error. The Opticap seemed to indicate tissue PCO2 with the same precision as the surface sensor. A better assessment of the in vivo response was obtained by comparing the response of the probe inserted in a cat's aorta to the PCO2 of blood samples taken from the same place and at the same time. Figure 5 shows the correlogram

Fiber Optic PC02 Sensor

507

-1-

80

- 10.7

60 8.0I I

g_ (3-

<

40

5.3

F-a-

,

o

I 20

0-

2.7

0" 7 ~ ~ -

2.7

5.3

8.0

10.7

kPa I

1

I

I

l

0

20

40

60

80

mmHg S U R F A C E Pco 2

FIGURE 4. Comparison of cortical surface PCO 2 measured with an Opticap and with a heated surface electrode. The dotted line is the line of identity; the solid line is a linear least squares fit with the equation y -- 1 . 0 3 7 x - 0 , 1 3 7 .

between measurements with the Opticap and with a calibrated blood gas analyzer. Opticap estimated PCO2 correlated well with the machine measured blood PCO2, having a correlation coefficient of the data pairs of 0.99 (n = 31). The slope and intercept of the linear least squares fitted regression curve were not significantly different from unity and zero, respectively.

DISCUSSION

The system presented here offers a way to measure PCO2 in tissue or in intravascular locations with several advantages. First, the sensor is completely electrically isolated from the subject so that the hazards of electrical connections and the hassle of getting adequate common mode rejection required for high impedance pH electrode based sensors are avoided. Sensors with built-in impedance converters as described by Moss, et al. (6), among others, offer a solution to the latter nuisance, but maintenance of the integrity of the insulation is required for long-term operation and safety. The directly linear relation between the

508

Vurek et al.

/ 6O

PopticapCO~-0.9+ Paco2(1.06+02)

re g ' ~

9 V/

PopticapCOz

identity

(mmHcj) 40

O / 9 9/

20

.y"

0

// ~~I

I

20

9 9

I

I

40 Poco 2 (rnmHcj)

I

I

60

I

FIGURE5. Correlation

between arterial blood PCOz as measuredby an Opticap and by conventional blood gas measurements. The straight line is a linear least squares fit of the data. The equation of fit is indicated (1 mm Hg = 0.133 kPa).

optical signal and PCO2 is a greater convenience than the log-log relation suggested in an earlier report of this device (9). The use of a computer to fit a second or third degree polynomial to the calibration data reduces the need for the reproducible fabrication techniques required to obtain the limited linear range described above. The sensitive volume is less than 0.5 ~zl so that the sensor measures local PCO2, and the response time is a minute or less. Sensor stability is of some concern, of course. There are two major sources of instability: solution composition changes and geometric changes. Failure of the seal can cause direct loss of the contents, but this can be prevented by proper fabrication techniques. The silicone rubber tube is permeable to water vapor so that the sensor will dry out if it is exposed to room air for a while. The osmolarity of the solution used in the sensors described here was significantly hypotonic. While this should have caused the sensor to imbibe water during the time it was exposed to normal osmolar fluids and thus dilute the contents and shift the calibration curve, there was no observable effect on the performance of sensors used over the course of experiments lasting eight or more hours. However, sensors observed over a period of several weeks including one followed for 12 weeks, showed a gradual shift in the calibration curve toward steeper slopes and greater intercepts. Inspection of the probes showed that there was dye concentrated around the plastic fibers, suggesting that it was binding to them through a defect in the epoxy protective coating. The calibration curves remained reasonably linear over the course of this change, but daily curves were required for

Fiber Optic PCO: Sensor

509

correct results during in vivo experiments. Drifts over the course of several hours were less than 0.6 kPa. Better sealing may prevent this sort of drift and make the probes useful for chronic studies. Other changes in solution composition could occur by diffusion of colored or pH active materials through the silicone barrier. Recent reports of acid and alkaline interference with the performance of carbon dioxide and ammonia sensors suggest that long-term exposure to the complex milieu of biological fluids may result in unexpected shift (2,3). The second source of shift in the calibration curve was due to deformation of the sensor tip during placement or removal from the experimental preparation. The flexible nature of the silicone tip offers little resistance to deformation. While the signal measured with the red filter offers some indication that a shift in geometry has occurred, simple ratioing will not provide adequate correction. That approach is all right for small changes in geometry but large shifts in optical path cannot be corrected this way. A better approach would be to make the probe more rigid so that deformation will not occur, using a geometry similar to that now used by Markle, et al. (5) for the optical pH probe. Alternately, one could make a measurement at the isobestic wavelength between the acid and base forms of the dye, obtain an estimate of the effective value for the DL term of the transmittance/PCO2 relation and thus derive a correction factor. This would also compensate for dilution or dye loss. The mechanical approach may cause the time response to increase because it might require some area now open to CO2 diffusion to be obstructed by rigid impermeable material. CONCLUSION The fiber optic PCO2 sensor described here offers the advantage of complete electrical isolation, small sensor volume, and fast response time. Simple theoretical analysis shows that the optical approach offers good linearity; experimental evidence supports the theory. REFERENCES 1. Coon, R.L., N.C.J. Lai, and J.P. Kampine. Evaluation of a dual function pH and PCO 2 in vivo sensor. J. AppL Physiol. 40:625-629, 1976. 2. Kobos, R.K., S.J. Parks, M.E. Meyerhoff. Selectivity characteristics of potentiometric carbon dioxide sensors with various gas membrane materials. Anal Chem. 54:1976-1980, 1982. 3. Lopez, M.E. and G.A. Rechnitz. Selectivity of the potentiometric ammonia gas sensing electrode. Anal Chem. 54:2085-2089, 1982. 4. LObbers, D.W. and N. Optiz. The PCOJPO 2 Optode: A new probe for measurement of PCO 2 or PO 2 in fluid and gases. Z. Naturforsch. 30C:532-533, 1975. 5. Markle, D.R., D.A. McGuire, S.R. Goldstein, R.E. Patterson, and R.M. Watson. A pH measurement system for use in tissue and blood, employing miniature fiber optic probes. Adv. Bioeng. Am. Soc. Mech. Eng. In press. 6. Moss, S.D., C.C. Johnson, J. Janata. Hydrogen, calcium and potassium ion sensitive transducers: A preliminary report. IEEE Trans. Biomed. Eng. 25:49-59, 1978.

510

Vurek et al.

7. Peterson, J.I., S.R. Goldstein, R.V. Fitzgerald, and D. K. Buckhold. Fiber optic pH probe for physiologic use. Anal. Chem. 52:864-869, 1980. 8. Severinghaus, J.W.A combined transcutaneous PO2-PCO2 electrode with electrochemical HCO; stabilization. J. Appl. Physiol. 51:1027-1032, 1981. 9. Vurek, G.G., J.I. Peterson, S.R. Goldstein, and J.W. Severinghaus. Fiber Optic PCO2 probe. Fed. Proc. Fed. Am. Soc. Exp. Biol. 41:1483, 1982. 10. Young, W.H2 clearance measurementsof blood flow: A review of techniques and polarographic principles. Stroke 11:552-564, 1980.