53 1

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39, NO. 5 . MAY 1992 tensor [27]

(applying the usual summation convention over repeated indexes) where ARS are the components of the identity tensor. This symmetric tensor totally describes the surface deformation at any point. Its eigenvalues are the principal (maximum and minimum) strains and represent the major and minor epicardial segment length changes. Positive values imply extension and negative values imply compression. Their associated eigenvectors are the (orthogonal) directions in which they act. The derivatives with respect to X, in (A3) can be calculated from

[I21 D. Terzopoulos, “Regularization of inverse problems involving discontinuities,” IEEE Trans. Patt. Anal. Mach. Intell., vol. 8, pp. 413424, 1986. [I31 A. A. Young, “Epicardial deformation from coronary cintangiograms,’’ Ph.D. thesis, Univ. Auckland, Auckland, New Zealand, 1990. [I41 D. D. Streeter and W. T. Hanna, “Engineering mechanics for successive states in the canine left ventricle myocardium,” Circ. Res., vol. 33, pp. 639-664, 1973. [I51 L. D. R. Smith, G. P. Robinson, D. J. Stevenson, H. Sanghera, and P. Quarendon, “A semi-automatic computerized system for the production of quantitative moving three dimensional images of the heart and coronary arteries,” IEE Conf. Image Process. Appl., (IEE Conf. Pub. 265), pp. 72-76, 1986. [I61 D. L. Pope, D. E. Gustafson, and D. L. Parker, “Cine 3D reconstruction of moving coronary arteries from DSA images,” Comput. Cardiol., pp. 277-280, 1987. [I71 Q. X. Wu, P. J. Bones, and R. H. T. Bates, “Translational motion compensation for coronary angiogram sequences, IEEE Trans. Med. Imag., vol. 8, pp. 276-282, 1989. [I81 L. D. R. Smith and P. Quarendon, “Four dimensional cardiac imaging,” Med. Image Process., SPIE, vol. 593, pp. 74-77, 1985. [19] G. Coppini, M. Demi, G. D’Urso, A. L’Abbate, and G. Valli, Ten”

The derivatives of

is with respect to X,

are [13]:

_-

ax,

Kl L

where

‘I

sor Description of 3 0 Time Varying Surfaces Using Scattered Lendmarks: An Application to Heart Motion in Time-Varying Image Processing and Moving Object Recognition. V. Capellini, Ed. North

ACKNOWLEDGMENT The authors thank Dr. P. Brandt and Dr. J . Ormiston for the cinkangiograms, Dr. P. Nielsen and Dr. A. McCulloch for many helpful discussions and Aulbrey Mathias for constructing the telecine apparatus.

REFERENCES [l] M. V. Herman, R. A. Heinle, and M. D. Klein, “Localized disorders

in myocardial contraction: Asynergy and its role in congestive heart failure,” New Eng. J . Med., vol. 277, pp. 222-232, 1967. [2] G. D. Meier, M. C. Ziskin, W. P. Santamore, and A. A. Bove, “Kinematics of the beating heart,” IEEE Trans. Biomed. Eng., vol. 27, pp. 319-329, 1980. [3] L. K. Waldman, Y. C. Fung, and J. W. Covell, “Transmural myocardial deformation in the canine left ventricle,” Circ. Res., vol. 57, pp. 152-163, 1985. [4] S. Chauduri and S . Chattejee, “Estimation of motion parameters for a deformable object from range data,” IEEE Conf. Comput. Vision Pattern Recognition, pp. 291-295, 1989. [5] Y. Kong, J. J. Moms, and H. D. McIntosh, “Assessment of regional myocardial performance from biplane coronary angiograms,” Amer. J. Cardiol., vol. 27, pp. 529-537, 1971. [6] M. J. Potel, J. M. Rubin, S . A. MacKay, A. M. Aisen, J. AI-Sadir, and R. E. Sayre, “Methods for evaluating cardiac wall motion in three dimensions using bifurcation points of the coronary arterial tree,” Invest. Rad., vol. 18, pp. 47-57, 1983. [7] A. A. Young, P. J. Hunter, and B. H. Smaill, “Epicardial surface estimation from coronary ciniangiograms,” Comput. Vis. Graph. Image Proc., vol. 47, pp. 111-127, 1989. [8] M. Hashimoto and J. Sklansky, “Multiple-order derivatives for detecting local image characteristics,” Comput. Vis. Graph. Image Proc., vol. 39, pp. 28-55, 1987. [9] G. E. Sotak and K. L. Boyer, “The laplacian-of-gaussian kernel: A formal analysis and design procedure for fast, accurate full-frame output,” Comput. Vis. Graph. Image Proc., vol. 48, pp. 147-189, 1989. [lo] S . A. MacKay, M. J. Potel, and J. M. Rubin, “Graphics methods for tracking three-dimensional heart wall motion,” Comput. Biomed. Res., vol. 15, pp. 455-473, 1982. [ 111 P. J. Hunter and B. H. Smaill, “The analysis of the heart: A continuum approach,” Prog. Biophys. Molec. Biol., vol. 52, pp. 101-164, 1988.

Holland: Elsevier Science, 1987, pp. 158-163. [20] -, “3D heart kinematics using biplane coronary arteriography,” Comput. Cardiol., pp. 711-713, 1986. [21] J. S. Rankin, P. A. McHale, C. E. Artentzen, D. Ling, J. C. Greenfield, and R. W. Anderson, “The three-dimensional dynamic geometry of the left ventricle in the conscious dog,” Circ. Res., vol. 39, pp. 304-313, 1976. [22] P. Bed, “Geometric modelling and computer vision,” Proc. IEEE, vol. 76, pp. 936-958, 1988. [23] K. S . Arun, T. S. Huang, and S. D. Blostein, “Least-squares fitting of two 3D point sets,” IEEE Trans. Patt. Anal. Mach. Intell., vol. 9, pp. 698-700, 1987. [24] L. Axel and L. Dougherty, “Heart wall motion: Improved method of spatial modulation of magnetization for MR imaging,” Radiol., vol. 172, pp. 349-350, 1989. [25] B. Shahraray and D. J. Anderson, “Optimal estimation of contour properties by cross-validated regularization, IEEE Trans. Patt. Anal. Mach. Intell., vol. 11, pp. 600-610, 1989. [26] A. R. Hashima, L. K. Waldman, and A. D. McCulloch, “Nonhomogeneous analysis of epicardial strain gradients in ischemic border zone,” Fed. Amer. Soc. Bio., 1990. [27] W. Fliigge, Tensor Analysis and Continuum Mechanics. Berlin: Springer, 1972. [28] F. J. Villareal, W. Y. W. Lew, L. K. Waldman, and J. W. Covell, “Transmural myocardial deformation in the ischemic left ventricle,” Circ. Res., vol. 68, pp. 368-381, 1991. ”

Development of a Medical Fiber-optic pH Sensor Based on Optical Absorption Roger Wolthuis, David McCrae, Elric Saaski, James Hartl, and Gordon Mitchell

Abstract-A new fiber-optic pH sensor system has been developed. The sensor uses an absorbtive indicator compound with a long wavelength absorbtion peak near 625 nm; change in absorbtion over the pH

Manuscript received June 28, 1991; revised October 31, 1991. R. A. Wolthuis was with MetriCor Inc., Woodinville, WA 98072. He is now with Metrilab Inc., 15430 N.E. 162nd St. Woodinville, WA 98072. D. McCrae and E. Saaski were with MetriCor Inc., Woodinville, WA 98072. They are now with Research International, Woodinville, WA 98072. J. C. Hartl and G. L. Mitchell are with Metricor, Inc., Woodinville, WA 98072. IEEE Log Number 9107174.

0018-9294/92$03.00 0 1992 IEEE

532

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39, NO. 5, MAY 1992

range 6.8 to 7.8 is reasonably linear. The sensor is interrogated by a pulsed, red LED. Return light signal is split into short and long wavelength components with a dichroic mirror; the respective signals are detected by photodiodes, and their photocurrents are used to form a ratiometric output signal. In laboratory tests, the sensor system provided resolution of 0.01 pH, accuracy of kO.01 pH, and response time of 30-40 s. Following gamma sterilization, laboratory sensor testing with heparinized human blood yielded excellent agreement (e.g., r = 0.992 for n = 42) with a clinical blood gas analyzer. Excellent sensor performance and low cost, solid-state instrumentation are hallmarks of this sensor-system design.

INTRODUCTION The measurement of blood pH usually occurs in the clinical laboratory, and typically involves use of a glass pH electrode. While the glass pH electrode has been and remains a very important sensor product, its relatively large size, breakable (glass) construction, and instability over time have constrained its use for in vivo applications. Further, these limitations have served as a stimulus for the development of alternative pH sensor technologies. The evolution of transistor-based technology during the seventies led to the development of field effect transistors (FET’s), and subsequently to the development of ion sensitive field effect transistors (ISFET’s) . This solid-state technology offered small sensor size, low sensor cost, and rugged sensor packaging, all attributes desirable for use with in vivo catheter-based measurement systems. Unfortunately, pH sensors based on this technology did not provide required stability over time, due largely to the absence of a satisfactory reference electrode [ 11. Fiber-optic sensor technology became popular in the 1980’s, and two separate approaches to pH sensing evolved. One group of investigators elected to use luminescent indicators; i.e., indicators that are excited at one wavelength to emit light at another wavelength [2]-[7]. Unfortunately, these luminescent indicators required bright excitation sources, large optical fibers to capture the weak emitted signal, and complex return light analysis optics. Even under the best of circumstances, most luminescent indicator systems have relatively low signal to noise characteristics. Other investigators chose use of absorbtive indicators as the basis for their fiber optic pH sensors. Of these, only one identified an indicator that provided pH measurements over the physiologic range [8]; unfortunately, the supporting instrumentation was bulky and complex. The remaining investigators used absorbtive indicators that operated over very acidic or very basic pH ranges [9], [lo]. The continuing challenge has been the development of absorbtive indicators that respond within the physiologic range, and that are compatible with relatively simple, solid-state instrumentation systems. The present sensor program began in 1985, and had as its objective the development of a fiber-optic pH sensor system based on optical absorbtion, and one employing a solid-state instrumentation system. These objectives were considered essential in order to ensure high performance and low sensorlsystem cost. Development of the chemical indicator, design of supporting system hardware, and testing of the completed system were the responsibility of MetriCor and are described in this report. Specific medical product applications, related product engineering, and clinical evaluation were the responsibility of Baxter Edwards Laboratories, and these may be described in a later paper. INDICATOR DEVELOPMENT The initial requirements for a suitable pH indicator were identified as follows: change in absorbtion with change in pH.

maximum sensitivity (A absorbtion/A pH) in the range pH 6.8 to 7.8. absorption change should occur at wavelengths longer than 600 nm to allow use of a bright, solid-state LED source. stable over time and during sterilization. Several classes of chemical compounds were surveyed for use as a pH indicator. The class of compounds containing the azo chromophore showed the most promise of meeting stated requirements. A satisfactory azo chromophore was subsequently synthesized and tested, and then modified with addends as shown in Fig. 1. Note that compound 5-91 has three functional parts: the azo chromophore, the reactive group, and the ‘modifier.’ The azo chromophore portion of indicator compound 5-91 has a weakly bound hydrogen ion; this hydrogen ion is readily lost or gained depending on the pH of the immediate environment. The loss or gain of a hydrogen ion causes a change in optical absorption near 600 nm. The reactive group portion of indicator compound 5-91 provides a means for attaching various chemical structures to the azo chromophore. The reactive group began with three chlorine atoms; one was lost when the azo chromophore was attached, another was lost when the ‘modifier’ was added. The third chlorine is lost when indicator compound 5-91 is attached to its substrate. The ‘modifier’ portion of this compound will be discussed later. Aqueous absorption spectra were obtained (Varian 100) for indicator compound 5-91 at each of several buffered pH conditions (Fig. 2, top); the long wavelength peak absorption is at 610 nm. Absorbtion data were subsequently replotted versus pH to obtain the classic sigmoid curve for pH indicators (Fig. 2, bottom). In this curve, the region with the greatest slope is the region with the greatest sensitivity to pH change. The midpoint of this region is termed the pKa, the point at which there are an equal number of indicator molecules with and without their weakly bound hydrogen ion. Note that in aqueous solution, compound 5-91 appears to be reasonably linear over the range pH 7-9. The optical stability of indicator compound 5-91 can be inferred from a study performed on an earlier indicator compound, 2-43, which contained the same reactive group and a similar azo chromophore. In this study, aqueous solutions of indicator compound 2-43 were buffered to a pH of 5 , 7 , 9 and 11, respectively, and then stored in cuvettes under room temperature, pressure and lighting conditions; the optical absorption of each cuvette was read (Varian 100) at varying intervals. As illustrated in Fig. 3, buffered aqueous indicator compound 2-43 remained optically stable for 40 days, and this stability was independent of pH environment. SUBSTRATE SELECTION A N D SUBSTRATE-INDICATOR TESTING In an absorbtive optical sensor system, the indicator is bound to a substrate so that indicator concentration within the light path remains constant over time. The requirements for a suitable substrate were as follows: less than 0.1 mm thick highly porous (to facilitate ion diffusion) relatively transparent in the 550-650 nm waveband hydroxyl groups available for covalent bonding with indicator Common packaging cellophane (cellulose, DuPont model 2 15PD) met these requirements and was subsequently chosen as the substrate material. The bonding of indicator compound 5-91 to cellophane is accomplished as follows. Cellophane is first cleaned in soap and alcohol solutions to remove softening agents (e.g., glycols). Clean cellophane is then placed in an aqueous solution of indicator com-

533

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39, NO. 5, MAY 1992

111

MODIFIER

REACTIVE

AZO

'\$

CHROMOPHORE

GROUP

Fig. 1. The structure of indicator compound 5-91 is shown. The three functional parts of this compound are defined and discussed in the text.

0.51 0.4

0,468

E

0

0.3

8

/

, I

@

8

E8

.....

0'

0.2 I

I

/ /

/

0.1

2

B

-

+ pH

11

pH 9

--e pH 7 -f

w

0

z

2 .4(2:

0 In m Q

.2 -

--

-

.--

-*

0

10

30

20


0.02 g indicator per 100 mL, however, all of the cellophane bonding sites are full and further indicator bonding is impossible. The spectral performance of 5-91 dyed cellophane is quite similar to the 5-91 aqueous absorbtion spectra shown earlier (see Fig. 2). The long wavelength peak absorption is now 628 nm (versus 610 nm for the aqueous spectra), suggesting that the cellophane environment has modified indicator performance slightly. The optical stability of 5-91 dyed cellophane was studied by placing pieces of 5-91 dyed cellophane in buffered solutions of pH 6 , 7, and 8, and then measuring optical absorbance (Varian 100) at varying intervals over the next 27 days. As shown in Fig. 5, there was no consistent change in optical absorbance for any of the cellophane samples; small data variations are likely due to inconsistent positioning of the dyed cellophane within the spectrophotometer light path. Since the final pH sensor product would require sterilization prior to use, the effect of sterilization on 5-91 dyed cellophane was studied. For gamma sterilization, pieces of 5-91 dyed cellophane were bathed in a buffered pH 6 , 7 , or 8 solution, and then tested in a spectrophotometer. Next, the dyed cellophane was air dried or placed in normal saline, and then subjected to 2.5 mrad gamma radiation. Following radiation, the dyed cellophane was placed back in its starting buffer solution and retested in the spectrophotometer. On average, the 5-91 dyed cellophane lost 21-26% of its starting absorption at 618 nm; this amplitude loss was independent of whether the dyed cellophane was wet or dry during gamma radiation. There were no other apparent changes in indicator spectral performance. In a steam sterilization study, 5-91 dyed cellophane was bathed in buffered pH 6.8 or 9.0 solution, and then tested in a spectro-

photometer. Next, the dyed cellophane was placed in distilled water or air dried, and then steam sterilized at 125°C for 30 min. Following steam sterilization, the dyed cellophane was replaced in its original buffer solution and retested in the spectrophotometer. Losses in absorbance at the lower pH averaged 3 1-47 % , and at the higher pH averaged 27-3 1 % . Further, dyed cellophane sterilized in the wet condition tended to loose less activity than cellophane sterilized dry. Importantly, loss of indicator compound activity during sterilization, from whatever cause, can be compensated for by loading the cellophane with additional indicator compound initially, or, by increasing the dynamic range of the supporting instrumentation system.

SENSORSYSTEMDEVELOPMENT Overall system optical design is illustrated in Fig. 6. The fused coupler is made from 100/140 pm multimode, step index, glass fiber. The source LED has a wavelength peak at 630 nm; this wavelength peak is achieved by selecting and then maintaining an appropriate LED operating temperature in the range 45-55°C. A dichroic ratio technique is used for retum light signal analysis. In this technique, retumed light is split (at or near the LED wavelength peak) by a dichroic mirror into short and long wavelength components; the corresponding detected photocurrents are used to form a ratiometric signal. This technique has been used successfully with other MetriCor fiber optic sensors, and has been described in detail earlier [ l l]. The dichroic ratio technique provides a system output relatively immune to changes in LED intensity, and, to the efficiency of lighthandling components within the optical system. In this technique, a change in light level affects both components of the ratio signal, leaving the ratio essentially unchanged. Sensor system performance using this technique is illustrated in Fig. 7. While this sigmoidal curve is quite similar to that shown earlier (see Fig. 2), the present data are based on two-pass light transmission (e.g., fibercellophane-reflector-cellophane-fiber) while Fig. 2 data were based on a single light pass and measured as optical absorbance. Hence, the shift of maximum sensitivity to a slightly lower pH and the apparent decrease in maximum sensitivity, both in Fig. 7, may be due to differences in data acquisition and data presentation. An early source of system instability/drift was traced to the high intensity LED-light emerging from the fiber tip was concentrated on a very small area of dyed cellophane, causing inactivation (photobleaching) of the pH indicator. This problem was resolved by

535

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39, NO. 5, MAY 1992

Source I

rn

LCU

h

r

Focusing A Lens

-k 3

’

I

%

I I

Fused Fiber

Dichroic

m -

’

I

To Optical

I

Sensor

I I

Receiving Photodiodes

‘ I Fig. 6. Overall sensor system optical design is illustrated. The LED is pulsed for 40 out of every 800 ms to prevent photobleaching of the indicator compound.

0 X=

(d re U

0 .0

I

0.55

s/

I

I

Quartz Base (200 um thick) Fiber Optic Coupler

I

I

U Fig. 8 . The in vitro pH sensor design configuration is shown. A disposable plastic dome (not shown) fits over the quartz base to form a sampling chamber. When hydrated, the dyed cellophane sensor button swells to fill the sensor cavity.

I+-.-” 4

5

6

7

8

Glass With Saw Slits

9 1 0 1 1

Buffered pH Fig. 7. Sensor system output (dichroic ratio) is shown as a function of pH. This sigmoidal curve is based on light passing through the indicator twice, and is measured in terms of optical transmission.

changing from continuous to pulsed mode LED operation; the LED is turned on for 40 ms each 800 ms, and indicator absorption is measured toward the end of the “on” period. Using pulsed mode operation, pH sensors buffered at physiologic ionic strength and capacity drifted only 0.015 pH units (range 0.012-0.021 pH) over 18 h; most of this drift was attributed to the prototype instrumentation used at the time. Sensor configurations were developed for in vitro applications. The in vitro sensor configuration involved a disposable plastic dome (27 mm diameter X 7 mm tall) with a sample chamber machined in the bottom. The sample chamber had horizontal inflow and outflow ports, and was sealed at the bottom with a removable quartz base attached with U V curing adhesive (Fig. 8). The quartz base (14 mm long x 8 mm wide x 0.28 mm thick) contained a series of chemically etched cavities, of varying diameter and depth, designed to hold various optical sensors then under development (i.e., pH, CO2, 0 2 ,Pressure, Temperature). For the pH sensor, a 300 pm O.D. sensor button was punched from a sheet of 5-91 dyed cellophane. The sensor button was placed in the proper quartz cavity, and the cavity was partially covered with a reflectorized retainer bar, held in place with U V curing adhesive. Over 50 of these in vitro sensors were fabricated, providing a majority of the performance data obtained during this development program. An in vivo pH sensor design configuration was also designed and tested (in virro). This in vivo sensor design configuration employed a 20 mm X 150 pm I . D . capillary tube as the sensor housing. At

Optical

Optical Fiber

Fig. 9 . The in vivo pH sensor design configuration is shown. This design was used for in vitro studies relating to the effect of dimensional features on sensor performance.

the middle of its length, the capillary tube was partially sawn through from opposite sides (Fig. 9). The interrogating 100/140 pm optical fiber was inserted into one end of the capillary tube and the tip positioned just short of the saw slot. A 150 pm O.D. sensor button was then punched from a sheet of 5-91 dyed cellophane, inserted through the open end of the capillary tube, and positioned even with the saw slot. Finally, a 100/140 pm optical fiber was cleaved, coated with gold (by evaporation) to form a reflective surface, inserted into the remaining end of the capillary tube, and positioned as shown. The optical fiber parts were held in place with U V curing adhesive. The saw slots provided a means for test solutions to access the sensor button. Since these assemblies were quite fragile, relatively few were fabricated and tested during the development program.

SENSORSYSTEMEVALUATION Performance data were obtained from over 50 in vitro sensors during the three year development program. A number of the per-

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39, NO. 5, MAY 1992

536

formance parameters are independent of sensor configuration, and data for these parameters are summarized as follows: Specification

Requirement

Performance

Range, linear Range, maximum Resolution Accuracy Sterilization

6.8 to 7.8 pH none 0.01 pH fO.O1 pH Gamma, 2.5 mrad

6.75 to 8.0 pH 6.0 to 9.0 pH 0.01 pH fO.O1 pH Passed

In Vitro Testing of the In Vitro Sensor Configuration Sensor system linearity is determined by several factors. First, recall that the relationship between indicator compound absorbance and pH is curvilinear (see Fig. 7), and that maximum, useable indicator sensitivity for indicator compound 5-91 occurs between pH 6-9. Within this span, a reasonably linear segment of 1.0 pH unit can be defined. Second, this reasonably linear segment can be shifted toward higher or lower pH with the addition of a ‘modifier’ to the indicator compound (see Fig. 1). This effect is illustrated by data from two separate in vitro sensors (Fig. 10). Note that indicator compound 3-97 (without the ‘modifier’) appears reasonably linear over the range pH < 6 . 5 to 7.3. With addition of the ‘modifier,’ however, the entire sigmoidal curve is shifted to the right by -0.25 pH units, and indicator compound 5-91 has a reasonably linear region positioned over the required range pH 6.8 to 7.8. Sensor system response time is constrained by a number of design variables, one of which is the diffusion path length between indicator compound molecules within the dyed cellophane, and, the outside test solution. For in vitro sensors at 37”C, the response time (90%) to a step change in pH (7.05 to 6.5) ranged from 30 to 40 s; for a step change from pH 6.5 to 7.05, response times were 10 s longer. This difference in response times was surprising, and may be associated with the cation exchange between dyed cellophane and the test solution. When test solution pH becomes more basic, protons leave the dyed cellophane and are replaced by metal cations from the test solution [12]. When test solution pH becomes more acid, cation movement is reversed. This cation exchange may exhibit diffusional differences which could explain the noted directional differences in response time. In vitro sensors were generally evaluated in the laboratory, using standard buffer solutions maintained at 37°C. Toward the end of the development program, however, gamma sterilized in vitro sensor assemblies were evaluated using heparinized human blood. As illustrated in Fig. 11, the MetriCor pH sensor showed excellent agreement when compared with a clinical laboratory blood gas instrument (ABL).

-

7

6.5

a

7.5

Buffered pH Fig. 10. The linearity of an in vitro pH sensor design containing 3-97 dyed cellophane (no ‘modifier’) and one containing 5-91 dyed cellophane (with ‘modifier’) is illustrated. Use of an appropriate ‘modifier’ shifts the sigmoidal curve toward higher pH and provides a reasonably linear response across the desired range pH 6.8-7.8. 7.6

2E

~

74-

5

; c 772-

0

5

e

-

/

x

-I

--6-4-

64

66

68

7 72 pH measured bf ABL2 Mood gas machm

14

76

Fig. 11. Paired measurements of human blood samples were made with a sterilized MetriCor in vitro sensor design and with a common clinical laboratory blood gas instrument (ABL).

0.55

0.5 0.45

In Vitro Testing of the In Vivo Sensor Configuration Sensor system linearity is illustrated once again in Fig. 12; note the essentially linear response over the range pH 6.75-8.0. Because of this linearity, sensor calibration (slope and intercept) can be accomplished with the use of two known pH solutions. Further, if optical absorption of the dyed cellophane sensor is the same from sensor to sensor, the slope response (absorption versus pH) will be the same for each sensor, and a single point calibration is all that is needed to define the intercept value. As the development program neared completion, another series of tests were conducted to characterize sensor system stability. In these tests, a sensor was calibrated and then placed in a buffered 7.09 pH solution. Sensor system output (rounded at two decimals) was recorded by a computer at 10 min intervals. As shown by the data in Fig. 13, sensor system output vacillated between pH 7.08

7

0.4

0.35 6

6.5

7

7.5

a

Buffered pH Fig. 12. Linearity of a MetriCor in vivo pH sensor design is illustrated.

and 7.09 during the three day test. This degree of stability is impressive when compared with conventional electrochemical pH sensor performance, and indicates that the optical sensor system may not require recalibration in applications lasting up to 72 h.

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39, NO. 5, MAY 1992

7 I

!

I

1

531

proved stability, etc.), these advantages can only be realized by controlling other variables that do not affect the glass electrode, namely, ionic strength and buffer capacity. Hence, it seems likely that both technologies will coexist for some time to come, with each serving applications that fit respective strengths best.

ACKNOWLEDGMENT

.L_ I711

The authors express appreciation to Baxter Edwards Critical Care Laboratories for their generous support of this development program, to J. Dove of Baxter Edwards for permission to use the data presented in Fig. 11, and to K. Garcin, D. Lawrence, J. Barger, D. Koop, G. Canny, and B. Willard of MetriCor for their many and valued contributions. n

1. y U

_ - rq-, .,. Titnr,

.>- ‘I.IN.,. ~

t~,,,,

’;I -,inil

ii~iiti.

REFERENCES

Fig. 13. Stability of a MetriCor in vivo pH sensor design is shown. The instrument rounded its output at two decimal places, with output ranging between pH 7.08 and 7.09 during the three-day test.

Other Issues Care should be exercised in the use of any sensor system, and pH is no exception. By definition, pH is the electrochemical measurement of hydrogen ion activity with respect to some reference. Optical methods equate chemical concentrations with hydrogen ion activity by means of activity coefficients. These coefficients are affected by ionic strength, solvent, and surface interactions, as noted earlier by Janata [13]. Hence, optical pH sensors must be used under carefully defined and calibrated conditions to insure accuracy of the resulting measurements. On a related matter, an interesting observation was made in connection with testing the pH sensor in blood, e.g., sensor calibrations performed with various buffer solutions were always offset from sensor calibrations performed with blood, even though both gave the same reading with an electrochemical pH sensor. After considerable effort and the careful control of all known variables, it was discovered that solution buffer capacity was the culprit, i.e., when the buffer capacity of the buffer standard was adjusted to that of blood, equivalent readings were obtained with the optical pH sensor system. Woods et al. [ 141 noted a similar effect while using an optical pH sensor in conjuction with flow injection analysis work. A possible explanation for both observations is that the sensor acts as an ion exchange resin, and the extent of this effect is mitigated by the buffer capacity of the solution.

CONCLUSIONS Program objectives called for development of an absorbtive pH sensor and supporting solid state instrumentation system. These objectives were met successfully. Interestingly, while these fiber optic sensors have a number of potential advantages over the traditional glass pH electrode (i.e., smaller size, less fragile, im-

U

[l] S. Yee, H. Jin, and L. Lam, “Miniature liquid junction reference electrode with micromachined silicone cavity,” Sensors and Actuators, vol. 15, pp. 337-345, 1988. [2] L. Saan and W. Seitz, “pH sensor based on immobilized fluoresceinamine,” Anal. Chem., vol. 54, pp. 821-823, 1982. [3] N. Opitz and D. Lubbers, “New fluorescence photochemical techniques for simultaneous measurements of ionic strength and hydrogen ion activities,” Sensors and Actuators, vol. 4, pp. 473-479, 1983. [4] Z. Zhujun and W . Seitz, “A fluorescence sensor for quantifying pH in the range from 6.5 to 8.5,” Analytica Chimica, vol. 160, pp. 4755, 1984. [ 5 ] 0. Wolfbeis, H. Posch, and H. Kronis, “Fiber optical fluorosensor for determination of halothane and/or oxygen,” Anal. Chem., vol. 57, pp. 2556-2561, 1985. [6] J. Gehrich, D. Lubbers, N . Opitz, D. Hansmann, W. Miller, J. Tusa, and M. Yafuso, “Optical fluorescence and its applicaton to an intravascular blood gas monitoring system,” ZEEE Trans. Biomed. Eng., vol. BME-33, pp. 117-132, 1986. [7] D. Jordan, D. Walt, and F. Milanovitch, “Physiological pH fiberoptic chemical sensor based on energy transfer,” Anal. Chem., vol. 59, pp. 437-439, 1987. [8] J. Peterson, S. Goldstein, R. Fitzgerald, and D. Buckhold, “Fiber optic pH probe for physiological use,” Anal. Chem., vol. 52, pp. 864-869, 1980. [9] G. Kirkbright, R. Narayanaswamy, and N. Welti, “Fibre-optic pH probe based on the use of an immobilised colorimetric indicator,” Analyst., vol. 109, pp. 1025-1028, 1984. [lo] T. Jones and M. Porter, “Optical pH sensor based on the chemical modification of a porous polymer film,” Anal. Chem., vol. 60, pp. 404-406, 1988. [ I l l R. Wolthuis, G. Mitchell, E. Saaski, J. Hartl, and M. Afromowitz, “Development of medical pressure and temperature sensors employing optical spectrum modulation,” IEEE Trans. Biomed. Eng., vol. 38, pp. 974-981, 1991. [12] B. A. Woods, J. Ruzicka, G. D. Christian, and N . J. Rose, “Measurement of rainwater pH by optosensing flow injection analysis,” Analyst., vol. 113, pp. 301-306, 1988. [I31 J. Janata, “Do optical sensors really measure pH?,” Anal. Chem., vol. 59, pp. 1351-1356, 1987. [14] B. A. Woods, J. Ruzicka, G. D. Christian, and R. J. Chartson, “Measurement of pH in solutions of low buffering capacity and low ionic strength by optosensing flow injection analysis,” Anal. Chem., vol. 58, pp. 2496-2502, 1986.