Pflfigers Archiv
PflfigersArch. 373, 273- 282 (1978)
EuropeanJournal of Physiology
9 by Springer-Verlag 1978
Instruments and Techniques A Fiberoptic Reflection Oximeter* M_ L. J. LANDSMAN,N. KNOP, G. KWANT,G. A. MOOK, and W. G. ZIJLSTRA Laboratory of Chemical Physiology,Universityof Groningen, Groningen,The Netherlands
Abstract. A catheter tip oximeter is described consisting of a cardiac catheter containing optical fibers, an incandescent light source, a light detection unit and a processing unit. Half of the optical fibers guide the light to the blood at the tip of the catheter, the other half the backscattered (reflected) light to the detection unit. The detection unit contains a dichroic mirror, transmitting most of the light with/l < 800 nm and reflecting most of the light with 2 > 900 nm, thus splitting the light into two beams. These pass through interference filters with nominal wavelengths of 640 and 920 nm respectively, and are focused on silicium barrier layer photocells. The photocell signals are amplified and fed into a divider giving the ratio of measuring (R ~4~ and compensating (R 92~ photocell output. The relationship between log R64~ 92~ and oxygen saturation is represented by a slightly curved line. The relation may be linearized by subtracting a constant voltage from the divider output before taking the logarithm. The slope of the calibration line is dependent on the total haemoglobin concentration. Nonetheless an average calibration line can be used between 70 and 100~ oxygen saturation. For 78 measurements of pig blood samples in this range (haemoglobin concentration between 96 and 161 g - 1-1), the standard deviation of the', difference between the fiberoptic oximeter and a Radiometer OSM1 oxygen saturation meter was 1 . 9 ~ saturation, for 152 samples over the entire saturation range the standard deviation of the difference was 3.1 ~ saturation. The influence of the flow velocity of blood on the light reflection depends on wavelength as well as on oxygen saturation. Therefore, complete compensation for the flow effect is not possible by simple means. * This work was supported in part by grants fromthe Netherlands Organization for the Advancement of Pare Research (Z.W.O.) received through the Foundation for Medical ScientificResearch (Fungo).
Key words: Oxygen saturation - Oximetry - Filter photometry - Fiberoptic oximeter - Cardiac catheterization.
INTRODUCTION A fiberoptic reflection oximeter measures oxygen saturation of blood at the tip of a cardiac catheter, utilizing the marked difference in light absorption between haemoglobin and oxyhaemoglobin (Fig. 1). The catheter contains 2 bundles of optical fibers, one of which guides light from an outside source into the blood; the other one guides the backscattered (reflected) l~ght to one or more photocells. This method has first been applied by Polanyi and Hehir in 1962 [16]. Later modifications of their technique have been described [1,4, 5, 8,15]. To compensate for non-specific effects such as changes of blood flow, haemoglobin concentration and erythrocyte shape, measurements are done in two wavebands, one in the red part of the spectrum (measuring wavelength), the other in the near infra-red (compensation wavelength). Presently most instruments employ light emitting diodes, turned on and off alternately at 2 0 0 - 300 Hz, in combination with a single photocell. Before light emitting diodes became available, chopped light had already been used [1,16]. Fifteen years ago the photomultiplier tube was the best choice for measuring very small light quantities. Using a light chopper only one photocell was needed, avoiding the problem of finding two photomultiplier tubes with identical dynamic characteristics. Also, compensation for dark current and ambient Sight could be easily attained. The present oximeter, like that described by Mook et al. [8], performs the light measurements simultaneously and continuously, using an incandescent lamp, interference filters and two photocells. There appeared
0031-6768/78/0373/0273/$ 02.00
274
Pfliigers Arch. 373 (1978)
1.2 f"HbO2' CHb Dig
A ] l'g
1.o
/
06
/
~o,/,-
/
/
/~-.~/ /',
I
-'--" ",
02
1.0
\
/ I
/
i
800
3
/.
02 -.. ,,~
900
2
RHb
/
0,4
,
B
/R(O/o)- 3.0
,
--'-~Z~o~l I ~ 1 700 600 ,
Fig. 2. Fiberoptic catheter. A catheter input; B catheter output; C junction disk; D catheter tip. The tip is shown in detail: 1 platinum tip; 2 catheter mantle; 3 fiber bundle; 4 electrode wire
0
Fig. 1. Absorption spectra of haemoglobin (Hb), oxyhaemoglobin (HbO2), indocyanine green (ig), and fiberoptic reflection spectra of oxygenated (Rnbo2) and deoxygenated (Rnb) blood. For Hb and Hb02 the extinction coefficients ~ are plotted against wavelength Z, for the dye the optical density D of a 5 mg 9 1-1 solution in plasma (light path length 1.0 cm). The fiberoptic reflection spectra have been drawn according to data of Mook et aL [8]
to be no dynamic imbalance between the two photocells. Replacement of the photomultiplier tubes by silicium barrier layer cells made the present instrument easier to handle and more stable. Mook et al. [8] studied the relation between light reflection measured through fiberoptics and oxygen saturation only briefly. The present paper fills up this gap and discusses the dependence of the oximeter output on haemoglobin concentration, blood flow velocity and indocyanine green concentration. DESCRIPTION OF THE OXIMETER The instrument consists of a fiberoptic catheter, a light source, a detection unit and a processing unit. The fiberoptic catheter (Fig. 2) has 2 proximal parts, one of which is the catheter input containing the efferent or illuminating fibers; the other is the catheter output containing the afferent or measuring fibers. The efferent and afferent fibers merge at the junction disk to form the distal part of the catheter. At the tip of the catheter efferent and afferent fibers are mixed at random. The standard F6 catheter 1 contains about 200 fibers of 50 gm diameter each. The catheter tip is finished off with a stainless steel ring. Upon the tip a tripod-like cage, 4 mm long, is mounted. This cage prevents direct contact of the glass fiber endings with 1 These catheters were built bY the American Optical Corporation for use with their first in vivo oximeter. This instrument has meanwhile been replaced by a new one employing a different type of catheter [14,15].
Fig. 3. Light source and detection units. The top and side-walls as well as the electronic circuits of the detection unit have been taken away. A lamp housing; B microscope objective (20 x ); C holder for the fiberoptic catheter input with centering device; D as C, for the catheter output; E Abbe condenser; F dichroic mirror; G and G' interference filters; H and H' filter and lens h o l d e r s ; / a n d I' microscope objectives (20 x ); K and K' photocell holders
the vessel wall. In some of the catheters we replaced the stainless steel tip (with cage) by a platinum tip (without cage) to be used as an oxygen electrode [11,121. A 12 V/100 W iodine quartz lamp (Philips 7023) fed by a stabilized DC power supply (Delta Elektronika D10) is used as a light source (Fig.3). The lens system of the light source consists of a combination of two positive lenses and a 20 x microscope objective. In the detection unit (Fig. 3) the light coming out of the fiberoptic catheter passes through a 1.25 N.A. Abbe condenser to a dichroic mirror which transmits most of the light with 2 < 800 rim, whereas most of the light with 2 > 900 nm is reflected (Fig. 4, curve C). The transmitted and reflected beams pass through interference filters and are focused by a 20 dioptre lens and a 20 x microscope objective on a silicium
M. L. J. Landsman et al. : Fiberoptic Reflection Oximeter
275
outpa~
600
700
.800
900
1000
II00
Fig. 4. Spectral sensitivity of the combination of light source, fiberoptic catheter (glass fibers) and photocell (A Siemens BPY11 silicium photocell; B Motorola MRD500 photocell, see appendix) and transmission spectra of dichroic mirror (Grubb Parsons) under 45 ~ angle (C) and interference filters (Baird Atomic type B3), nominal wavelengths 640 (D), 800 (E) and 920 nm (/7). Measuring instrument: Beckman DU spectrophotometer
barrier layer photocell measuring I x 2mm (Siemens BPY 11). The spectral characteristics of the combination of light source, catheter and photocell, as well as the transmission spectra of the filters are shown in Figure 4. In the measuring and compensating channel, filters are used with nominal wawelengths of 640 and 920 nm. Before transmission to the processing unit, the photocell signals pass through preamplifiers in the detection unit. In the processing unit (Fig. 5) the outputs are adjusted for zero offset of the first amplification stage and/or cross-talk, i.e., direct transmission of light from illuminating to measuring fibers. With the catheter tip in Indian ink the output of the second amplification stage in the measuring and the compensating channels is made zero (Fig. 5, m and c, comp.). When there is no appreciable crossl;alk, this adjustment may be done with the light shut off. The gain of the photocell amplifiers can be adjusted at two levels. At the first level gains rn and c are used for setting the ratio of measuring and compensating photocellamplifier outputs, while the fiberoptic catheter tip faces a surface mirror serving as a reflection standard. At the second level the gain controls of the two channels are coupled in a double 10-turn potentiometer (Fig. 5, gain m, c), so that adjustment at this level
Fig. 5. Diagram of the circuits of the processing unit of the fiberoptic oximeter. The photocell signals which passed preamplifiers in the detection unit, are connected to the input. The first block contains the zero offset (comp.) and the first level gain controls. Gain rn, c is the coupled gain control which does not change the ratio of the measuring and the compensating signal. With selector switch (1) the functions zero control (too and co), calibration (adj.) of the recorder outputs Rm and Re, and operation (rec.) can be selected. The amplifiers AMP. serve as preamplifiers for the recorder, with a zero control, a stepwise attenuator (art.), and a continuous sensitivity control (sens.). The divider output passes an adjustable filter, and a linearizing circuit consisting of an adjustable bias and a logarithmic amplifier (LOG.AMP.). With selector switch (2) the divider outputs for 0 (/) and 100 ~ (h) oxygen saturation are connected to the linearizing system; with (2) at 1 the oxygen saturation meter (So2) or digital voltmeter (D VM) is adjusted to zero by zero suppression (l-adj.), with (2) at h the meter is adjusted to 100 with the gain control (h-adj.)
does not change the ratio of measuring and compensating signals. With the catheter in the blood stream the photocell-amplifier outputs can thus be adjusted to keep the signals below 10 V. The divider gives the ratio of the measuring and the compensating signals (Rm/R~). This is the actual, non-linear oximeter output. After electronic filtering the signal passes through a linearizing circuit so that after proper calibration the oxygen saturation can also be read from a digital voltmeter (DVM). The relation between oximeter output and oxygen saturation is made linear by subtracting a constant voltage (bias) from the divider output before taking the logarithm. The magnitude of the bias depends on position and shape of the calibration line. Calculation of the bias voltage can be done as follows. Between the divider output y and the oxygen saturation x the following relation is assumed: log(y-
c) = p . x + q
(1)
where c is the bias voltage and p and q are constants. Three corresponding values of x and y are read from
276
10
Pflfigers Arch. 373 (1978)
E~ ~Rm - (V)
2.0
Rm +1 (V) .k'LogEl~c
10
| i
So, (*/*) i 100
1.01
So, (%1 100
8 6
k.tog (El-~-c m -c)+1 (V)
b.[k.[og (E1R~- c)+1- a] (V)
2.0
4
CHb(g.[_, ) -11.8 "]90 0
100
Fig. 6. Graphic representation of steps in the linearizing system. R,, and R~ measuring and compensating photocell-amplifier outputs, E1 = scaling voltage of divider (= 10 V), k = 1 V c = bias voltage,a = divideroutput for 0 oxygensaturation ("low" in Fig. 5), b = gain factor (h-adj. in Fig. 5), So2 = oxygen saturation. Graphs 1, 2 and 3 show the steps followed by the system. Graph 2a shows the result when the logarithm of the divider output is taken directly
the calibration line yielding 3 equations with 3 unknowns (p, q and c) from which c can be solved. It is obvious to choose for x the values 100, 50 and 0. When ytoo, ys0 and yo are the corresponding y values, then c =
y5o 2 - ytoo 9yo
i
40
80
i
i
i
i
120
160
200
21,0
Fig. 7. Relationship between fiberoptic light reflection at 2 = 640 n m (R64~ 2 = 920 nm (R 9z~ as well as the ratio of light refections (R6r176176 and haemoglobin concentration (cnb). Measurements of flowing heparinized oxygenated pig blood. The So2 scale gives the apparent change in oxygen saturation
So=(%) 100
i
(2)
2 yso - yaoo - Yo Figure 6 shows the steps taken in the linearizing system. The divider outputs for 0 and 100% oxygen saturation are set at the low and high dials respectively (Fig. 5, high and low). The adjustment of the linearizing system is thus carried out with the actual oximeter output disconnected 9
PROPERTIES AND PERFORMANCE OF T H E O X I M E T E R Dependence o f Blood Reflection at ,~ = 640 and 920 nm on Total Haemoglobin Concentration
Fresh heparinized oxygenated pig blood was used to prepare samples with different haemoglobin concentrations by mixing cells and plasma in different pro-
portions. The samples were drawn through a black plastic cuvette using a H a r v a r d constant withdrawal pump. The tip of a fiberoptic catheter was situated inside the cuvette. The reflections at 2 = 640 nm (R 6r and 920 nm (R 92~ and the ratio R64~ 9z0 were recorded for all samples. The total haemoglobin concentration was measured as haemiglobincyanide [3] using a Vitatron H b F100. The relation between blood reflection and haemoglobin concentration (Fig. 7) is essentially the same as described by M o o k et al. [8]. The oximeter output R64~ 920 shows a plateau between 90 and 140 g . 1-1. Towards lower haemoglobin concentration the ratio decreases, to-' wards higher concentration it increases9 The result is that the apparent oxygen saturation is 92 and 104 % at 40 and 240 g . 1-1 respectively. Between 80 and 190 g - 1-1, however, the apparent oxygen saturation changes from 98 - 102 %. To study the effect of changes in oxygen saturation on the relation between haemoglobin concentration and ox~meter output, two portions of heparinized pig blood with different haemoglobin concentrations (99 and 181 g 9 1-1) were prepared. F r o m each portion 8 samples with different oxygen saturation were made. The oxygen saturation was determined with a Radiometer oxygen saturation meter (OSM1), which had previously been checked using a spectrophotometric two-wavelength method 9 Figure 8 shows that at lower oxygen saturation the dependence of the oximeter output on haemoglobin concentration increases9 At 9
/
M. L. J. Landsman eta]. : Fiberoptic Reflection Oximeter Log
277
10R6{~92~
0.~
/
0.E
A
"0 0A 190 %2{O/o)
So2(%) 0.~
,
20 i
,
40 i
J
60 t
,
B0 t
i
/
80 60 10
100 ,
Fig. 8. Relationship between fiberoptic oximeter output (R64~ R920) and oxygen saturation (So~) at haemoglobin concentrations of 181 g 91 -~ (dots) and 99 g. 1-~ (triangles). The logarithm of the oximeter output ( x 10) is plotted against oxygen saturation. Measurements of flowing heparinized pig blood. The factor 10 has been introduced to keep the figures positive
0 ll]R {V)
R~~
I00[$oP/o) about 9 5 ~ saturation variation in haemoglobin concentration does not change the oximeter output; at 5 0 ~ saturation, however, there is an apparent decrease in oxygen saturation of" about 1 2 ~ when haemoglobin concentration increases from 99 to 181 g 91 - I .
Dependence of the Oximeter Output on Changes in Blood Flow Velocity The effect of changes in blood flow velocity was tested in vitro using heparinized dog blood samples with oxygen saturations 100 and 30 ~ . The tip of the fiberoptic catheter was placed inside a black plastic cuvette through which the blood was drawn by hand from a well stirred vessel using a 50 ml syringe. R 640, R 920 and R6r176176 w e r e recorded on an Elema mingograph EM81. The measuring and compensating signals were also displayed on a storage oscilloscope (Tektronix R564 B) used as x - y recorder. Differences in blood flow velocity of 0 - 3 0 cm - s -~ resulted in changes of the light reflection of about 50 ~ (Fig. 9). The ratio also showed variations corresponding to about 10 ~ oxygen saturation. The effect on the ratio appeared to be dependent on the oxygen saturation. With increasing reflection R64~ 92~ decreased at i00 ~/oxygen saturation and increased at 30 ~ saturation. The x - y recordings (Fig. 9, inserts) show the relation between R 6r176 and R 92~ at 100 and 30 ~ oxygen saturation. In vivo recordings generally show less flow
5)
.
Fig. 9. Effectof changes in blood flow on the fiberoptic oximeter output at 100~ (A) and 30~ (B) oxygen saturation. Heparinized dog blood. Tip of fiberoptic catheter inside a cuvette through which blood was drawn. R64~ and R9z~ = measuring and compensating photocetlamplifieroutputs9The divider output R64~ 92~is calibrated in percent oxygen saturation (So_0.The inserts are the x - y plots taken during the first part of the recording. The upper curve in the insert of B has been recorded after a 4-fold increase in gain of R64~
variations, though the variations may still be considerable. In the example of Figure 13 (measuring site in the aorta) the variations correspond to 6 ~ oxygen saturation.
Relation between Oximeter Output and Oxygen Saturation F o r the in vitro determination of the relation between the light reflections at 2 = 640 and 920 nm and the oxygen saturation, fresh heparinized pig blood was used. A portion of the native blood was oxygenated and an other portion de-oxygenated, then mixed in various proportions to obtain measurements over the whole range of 0 - 1 0 0 ~ o. The measurements were made with the fiberoptic catheter inside black plastic tubing, through which a constant flow of blood was maintained using a H a r v a r d constant infusion/with-
278
Pfl'~igers Arci~. 373 (1978)
tog10R6~0/ R920 t~110 R6z'O/R 9z0
R~~ 6~~ 0.8
0.8
x
N \
\\
\
~.(1
x
0,6 %" \ ~ ' / "
0.B
0.4 ~
3.0
"b. \ \
"~- ~ ~
9
s x
2.0
0.4
"-,-o.,.o.
-q
0'20
I
I
20
I
I
40
I
I
60
I
S%(%) 1.0 I [ 80 100
S@o/o) %
'
2;
'
4;
'
6;
'
'
Fig. 10. Relationship between the logarithm ofSberoptic oximeter output (R64~ 92~ and oxygen saturation (So2). Dots: native blood, partly oxygenated in an open flask. Crosses: equilibration with different 02, N2, CO2 mixtures in closed tonometers [17]. Open squares: partial de-oxygenation with Na2S204; for complete deoxygenation 0.4 mg Na2S204 was added per ml blood. Solid squares: reoxygenation after Na28204 treatment. The interrupted line with open circles is the result of plotting the reciprocal of the oximeter output (R92~ 64~ against oxygen saturation
Fig. l l . Relationship between the logarithm of uncompensated fiberoptic oximeter output (R64~ 92~ and oxygen saturation (So~). Data from the samples of Figure 10. For better comparison with Figure 10 R 64~ has been divided by a constant value, for which the mean of the R 92~ values of the samples has been taken
drawal pump. In this way stable readings of the blood reflection at the two wavelengths were obtained. The oxygen saturation of the samples was determined with a Radiometer oxygen saturation meter (OSM1). Figure 10 shows the logarithm of the 0ximeter output (log lOR64~ 920) as well as the reciprocal of the oximeter output (R92~ 64~ plotted against the oxygen saturation. All points fall on a slightly curved line irrespective of the way of preparation of the samples. It does appear, however, that some of the samples underwent considerable non-specific changes. This can be shown by plotting the uncompensated output R 64~ against the oxygen saturation (Fig. 11). The light reflection of the samples prepared by tonometry is lower than that of the samples to which Na2S20~ was added. From 14 different portions of pig blood, haemoglobin concentration ranging from 96-161 g. 1 t, a total of 152 samples with different oxygen saturations were prepared using Na2S204 to obtain low and zero saturation. Log 10R64~ 92~ has been related to the oxygen saturation, and the best fitting parabolic function has been calculated by the least squares method according to
Using this line as calibration line of the oximeter, for the 152 pig blood samples a standard deviation of the differences between fiberoptic oximeter and Radiometer oxygen saturation meter of 3.1 ~ saturation has been found. For 114 samples between 50 and 100 saturation the standard deviation was 2.7 ~o, and for 78 samples between 70 and 100~ it was 1.9~o saturation. In vivo measurements were done on 10 mongrel dogs in thialbarbital anaesthesia after premedication with acepromazine. In 5 experiments the animals were incubated and connected to a spirometer. Anaesthesia was maintained with intravenous thialbarbital and breathing was spontaneous. The oxygen was replaced stepwise by nitrogen to obtain low oxygen saturation. In 5 experiments the endotracheat tube was connected to a respirator and spontaneous breathing was suppressed with pancuronium bromide. The dog was given a nitrous oxide-oxygen mixture to breathe and oxygen saturation was varied by changing the N20/O2 ratio. The fiberoptic catheter was introduced into the left femoral artery and the tip was advanced into the descending aorta just above the diaphragm. An F8 Cournand catheter was introduced into the right femoral artery and the tip was positioned next to the tip of the fiberoptic catheter. Through this catheter blood samples were drawn simultaneously with the recording of the oximeter signal. The oxygen
S02 = ao + at log lOR64~ 92~ + a2 (log 10R64~176
2
(3)
where a0 = -120.0, a l = 361.7, and a2 = -130.6.
M. L. J. Landsman et al. : Fiberoptic Reflection Oximeter
saturation of the blood samples was determined with a Radiometer oxygen saturation meter (OSMI). From the results (Fig. 12) a standard deviation of 2.6 ~ oxygen saturation has been calculated.
279 tog10 R 6~'~ 920
0,~
Continuous Observation of Oxygen Saturation in the Dog In one of the experiments described above a continuous record was taken of the arterial oxygen saturation and of the change in oxygen tension (Po~). The fiberoptic catheter was provided with a platinum tip serving as a Poa-sensitive electrode [11, I2]. The dog breathed spontaneously. Oesophageal pressure was measured with a fluid-filled bird's eye catheter connected to a Statham P23Bb pressure transducer. Aortic pressure was recorded using the aortic sampling catheter connected to a Statham P23Db pressure transducer, and right atrial pressure via an F7 catheter and a P23Db transducer. Immediately before the record shown in Figure 13 a low oxygen gas mixture had been replaced by a mixture with 20 % Oa. During breathing the low oxygen gas mixture the arterial Pco~ became low due to hypoxic hyperpnoea. Following the restoration of a nearly normal Po~ in the respiratory gas mixture, apnoea occurred due to the disappearance of the hypoxic drive from the peripheral chemoreceptors. During the apnoea the arterial oxygen saturation decreased from 9 0 - 7 0 G. In the subsequent recovery period the oxygen saturation increased with each breath.
Sensitivity of the Oximeter to Indocyanine Green The error in the oxygen saturation measurement due to the presence of indocyanine green has been determined from a series of measurements on heparinized pig blood samples, oxygen saturation ranging from 3 5 - 1 0 0 % . The samples were divided into two aliquots and kept in syringes. To one of each pair indocyanine green was added (anaerobically) at a concentration of 5 mg 91-a. Measurements were made with two fiberoptic oximeters, one containing the 640/920 filter combination, the other with a filter with a nominal wavelength of 800 nm instead of 920 nm. The calibration line of the first one was used to determine the oxygen saturation of the samples (without dye) and to calibrate the second oximeter. The lines relating oxygen saturation and oximeter output for the dye containing samples were determined for both instruments. From these lines the apparent change in oxygen saturation due to the presence of dye was determined at different levels of oxygen saturation. The results (Table 1) show that for the oximeter with the 640/920 combination the error caused by the
0.6
0.4
s%(%) '
2;
'
'
6;
'
'
Fig. 12. In vivo calibration line, based on measurements on 91 blood samples obtained in 10 dog experiments. The equation of the line is So~ = -- 74.35 + 281.5 log IOR,,/R~ - 99.05 (log IOR~/Rr ~
5~:0-.J
Fig. 13. Recording in the dog of electrocardiogram (ECG H), respiration (resp., inspiration upwards), oesophageal pressure (Poes), aortic pressure (P~o), right atrial pressure (Pr,), fiberoptic oximeter output (So2), the same with electronical filtering (0.5 Hz, 3 dB), change in oxygen tension (APoO, compensating (R 92~ and measuring (R 64~ photocell outputs of the fiberoptic oximeter
280
Pftfigers Arch. 373 (1978)
Table 1. Apparent oxygen saturation ($82) and change in So2 (ASo2) at different levels (So2) caused by the addition of 5 mg - 1-1 indocyanine green, measured with fiberoptic oximeters with 640/920 and 640/800 filter combinations
So=
SG
ASo2
(640/920) 100 90 80 70 60 50 40 30
98 88 78 68 58 48 38 28
$32
AN)e
(640/800) -2 -2 -2 - 2 -2 -2 -2 -2
109 101 93 85 77 68 59 50
+ 9 + tl +13 + 15 + 17 + 18 + 19 +20
presence of the dye is slight and constant over the entire saturation range. For the oximeter with the 640/800 combination the deviation is considerable. DISCUSSION On the basis of the work of Rodrigo [18], Polanyi and coworkers [14,16], and others [2,4] assume a linear relation between oxygen saturation and the ratio of compensating (oxygen saturation independent) over measuring (oxygen saturation dependent) photocell outi3ut Rc/R~. This is not borne out by our results (Fig. 10). The fact that 2 = 920 nm, the wavelength used for compensation, is not isobestic is irrelevant in this respect. Actually the wavelength used by Polanyi, 2 = 805 nm, is not isobestic either (Fig. 1), but this is not the cause of non-linearity, as he suggests [15]. In the Schwarzer oximeter the ratio RdRm is corrected to obtain linearity of oxygen saturation and output signal [15]. The fact is that the reciprocal of the blood reflection is not linear with oxygen saturation, and this non-linearity cannot be corrected for by the second measurement. In spectrophotometry of haemolyzed blood the situation is essentially different, because, when haemoglobin concentration is constant, optical density changes linearly with oxygen saturation. This relation is not affected when, for compensation of differences in haemoglobin concentration, the optical density is divided by the optical density measured at an isobestic wavelength. Also in the conventional reflection oximeters there is no linear relation between the reciprocal of the blood reflection or the ratio of reflections R d R m and oxygen saturation [9]. So, apparently, the theory which predicts a linear relation [14,18], is not a complete description of light reflection by blood as measured with conventional reflection oximeters or by fiberoptics. In the authors' laboratory it has been traditional to relate the loga-
rithm of the reflection obtained by the original onecolour instruments with the oxygen saturation [19]. Taking the logarithm of the ratio of measuring over compensating photocell output, when employing measurements in two colours, is equivalent to this. The relation between log R,,/Rc and oxygen saturation is not linear either, neither in conventional reflection oximetry [9] nor in fiberoptic reflection oximetry (Figs. 8, 10, and 12). Linearization can be achieved by subtracting a constant voltage c from the oximeter output Rm/Rc before taking the logarithm (Fig. 6); thus log (Rm/Rc - c) or log (Rm - c 9 Rc)/Rc is linearly related to oxygen saturation. When using the inverse ratio, R d ( R ~ + c . Rc) = 1/(R,,/R~ + c) can be made linear with oxygen saturation by properly choosing the value of c. This is the procedure applied in the Schwarzer oximeter. The influence of haemoglobin concentration on the calibration line is such that at higher saturation, above 70 ~, an average calibration line may be used. For accurate measurements below 70 ~ saturation, however, the calibration should be adapted to the haemoglobin concentration; the high and low adjustments (high and low, Fig. 5) are used to this end. The influence of non-specific effects other than the haemoglobin concentration is illustrated in Figures 9, 10 and 11. The changes in reflection of dog blood in vitro (Fig. 9) can be explained by formation and breaking up of rouleaux, and to a minor extent by a directional effect on the erythrocytes. The shape of the x-y plots shows that the influence of changes in blood flow cannot be compensated for by simple means. An additional complication is the oxygen saturation dependency of the flow effect. In vivo, however, the directional effect on the erythrocytes prevails. In the great blood vessels rouleaux formation does not occur. The nonetheless appreciable flow effect can be suppressed by electronic filtering. The dynamic response of the system is then determined by the electronic filter (Fig. 13). In Figure 11 the differences in reflection may be caused by changes of the erythrocyte surface due to prolonged (about I h) contact with the glass wall of the tonometer and/or by osmotic changes due to the addition of Na2S204. Compensation for these effects is practically complete (Fig. 10). Oximeters used in the cardiac catheterization laboratory or applied to monitoring in situations where cardiac output is measured by indocyanine green dilution, should be insensitive to this dye. Thus 2 - - 8 0 0 n m cannot be used as a compensation wavelength. Table I shows the large error in the oxygen saturation measurement when 2 = 800 nm is used for compensation. When 2 = 920 nm is used instead, the error is but slight. The oximeter does
M. L. J. Landsman et al. : Fiberoptic Reflection Oximeter
not become completely insensitive to dye, because at 2 = 640 nm the light absorption by indocyanine green is not zero (Fig. 1). Fiberoptic oximetry has two important advantages over catheter cuvette oximetry. Firstly, the fiberoptic oximeter is a catheter tip oximeter and as such capable to reliably record fast changes in oxygen saturation, which may occur in the heart or great vessels, especially in cases of shunts [1]. For any cuvette outside the body the dynamic response characteristics are limited by the catheter [10]. Secondly, there is no need to withdraw blood. This reduces blood loss and, still more important, it makes the method suited for continuous observations over long periods, e.g., for several hours to some days [6]. The present instrument has some advantages over commercially available fiberoptic oximeters [14,15]. The measurements as well as the ratio computation are continuous. Other instruments employ light emitting diodes, which are pulsed. The frequency used is 2 0 0 - 3 0 0 Hz. There is only one photocell, the signal of which thus contains the light reflections at the two wavelengths. These are separated electronically. Because the two measurements Rm and Rc are not synchronous, compensation for non-specific effects is dependent on the pulse frequency. The changes in reflection due to changes in blood flow may be so fast that it is least doubtful whether a pulse frequency of 300 Hz is enough for compensation of these changes. Continuous measurement with two photocells gives the best possible compensation and dynamic response. The dynamic responses of the two photocellamplifier assemblies do not differ so much as to be the cause of limitations in this respect [7, 8]. The main advantage of the present instrume.nt, however, is the use of light filters. In this way any wavelength and bandwidth may be chosen. Measurements can be done even without filter and thus in a band determined by the spectral characteristics of the lamp, the optical fibers, and the photocell. This is applied in the fiberoptic densitometer [7]. By using light filters it is also very simple to switch from one wavelength combination to another. The fiberoptic oximeter is preferably applied to the continuous recording of arterial or central (mixed) venous oxygen saturation. Figure ~3 shows an example obtained in a physiological experiment. With the simultaneous recording of oxygen tension (x-axis) and oxygen saturation (y-axis) in vivo haemoglobinoxygen equilibrium curves can be determined [13]. APPENDIX A more recent development of the fiberoptic reflectometer consists of a combined light source and detection
281
C
B D
/
E
8~
.
..
Aj
6/-.0~ , ~F 8 0 0 ~
920 920 - nf
/ E
i
PflfigersArch. 373, 273- 282 (1978)
EuropeanJournal of Physiology
9 by Springer-Verlag 1978
Instruments and Techniques A Fiberoptic Reflection Oximeter* M_ L. J. LANDSMAN,N. KNOP, G. KWANT,G. A. MOOK, and W. G. ZIJLSTRA Laboratory of Chemical Physiology,Universityof Groningen, Groningen,The Netherlands
Abstract. A catheter tip oximeter is described consisting of a cardiac catheter containing optical fibers, an incandescent light source, a light detection unit and a processing unit. Half of the optical fibers guide the light to the blood at the tip of the catheter, the other half the backscattered (reflected) light to the detection unit. The detection unit contains a dichroic mirror, transmitting most of the light with/l < 800 nm and reflecting most of the light with 2 > 900 nm, thus splitting the light into two beams. These pass through interference filters with nominal wavelengths of 640 and 920 nm respectively, and are focused on silicium barrier layer photocells. The photocell signals are amplified and fed into a divider giving the ratio of measuring (R ~4~ and compensating (R 92~ photocell output. The relationship between log R64~ 92~ and oxygen saturation is represented by a slightly curved line. The relation may be linearized by subtracting a constant voltage from the divider output before taking the logarithm. The slope of the calibration line is dependent on the total haemoglobin concentration. Nonetheless an average calibration line can be used between 70 and 100~ oxygen saturation. For 78 measurements of pig blood samples in this range (haemoglobin concentration between 96 and 161 g - 1-1), the standard deviation of the', difference between the fiberoptic oximeter and a Radiometer OSM1 oxygen saturation meter was 1 . 9 ~ saturation, for 152 samples over the entire saturation range the standard deviation of the difference was 3.1 ~ saturation. The influence of the flow velocity of blood on the light reflection depends on wavelength as well as on oxygen saturation. Therefore, complete compensation for the flow effect is not possible by simple means. * This work was supported in part by grants fromthe Netherlands Organization for the Advancement of Pare Research (Z.W.O.) received through the Foundation for Medical ScientificResearch (Fungo).
Key words: Oxygen saturation - Oximetry - Filter photometry - Fiberoptic oximeter - Cardiac catheterization.
INTRODUCTION A fiberoptic reflection oximeter measures oxygen saturation of blood at the tip of a cardiac catheter, utilizing the marked difference in light absorption between haemoglobin and oxyhaemoglobin (Fig. 1). The catheter contains 2 bundles of optical fibers, one of which guides light from an outside source into the blood; the other one guides the backscattered (reflected) l~ght to one or more photocells. This method has first been applied by Polanyi and Hehir in 1962 [16]. Later modifications of their technique have been described [1,4, 5, 8,15]. To compensate for non-specific effects such as changes of blood flow, haemoglobin concentration and erythrocyte shape, measurements are done in two wavebands, one in the red part of the spectrum (measuring wavelength), the other in the near infra-red (compensation wavelength). Presently most instruments employ light emitting diodes, turned on and off alternately at 2 0 0 - 300 Hz, in combination with a single photocell. Before light emitting diodes became available, chopped light had already been used [1,16]. Fifteen years ago the photomultiplier tube was the best choice for measuring very small light quantities. Using a light chopper only one photocell was needed, avoiding the problem of finding two photomultiplier tubes with identical dynamic characteristics. Also, compensation for dark current and ambient Sight could be easily attained. The present oximeter, like that described by Mook et al. [8], performs the light measurements simultaneously and continuously, using an incandescent lamp, interference filters and two photocells. There appeared
0031-6768/78/0373/0273/$ 02.00
274
Pfliigers Arch. 373 (1978)
1.2 f"HbO2' CHb Dig
A ] l'g
1.o
/
06
/
~o,/,-
/
/
/~-.~/ /',
I
-'--" ",
02
1.0
\
/ I
/
i
800
3
/.
02 -.. ,,~
900
2
RHb
/
0,4
,
B
/R(O/o)- 3.0
,
--'-~Z~o~l I ~ 1 700 600 ,
Fig. 2. Fiberoptic catheter. A catheter input; B catheter output; C junction disk; D catheter tip. The tip is shown in detail: 1 platinum tip; 2 catheter mantle; 3 fiber bundle; 4 electrode wire
0
Fig. 1. Absorption spectra of haemoglobin (Hb), oxyhaemoglobin (HbO2), indocyanine green (ig), and fiberoptic reflection spectra of oxygenated (Rnbo2) and deoxygenated (Rnb) blood. For Hb and Hb02 the extinction coefficients ~ are plotted against wavelength Z, for the dye the optical density D of a 5 mg 9 1-1 solution in plasma (light path length 1.0 cm). The fiberoptic reflection spectra have been drawn according to data of Mook et aL [8]
to be no dynamic imbalance between the two photocells. Replacement of the photomultiplier tubes by silicium barrier layer cells made the present instrument easier to handle and more stable. Mook et al. [8] studied the relation between light reflection measured through fiberoptics and oxygen saturation only briefly. The present paper fills up this gap and discusses the dependence of the oximeter output on haemoglobin concentration, blood flow velocity and indocyanine green concentration. DESCRIPTION OF THE OXIMETER The instrument consists of a fiberoptic catheter, a light source, a detection unit and a processing unit. The fiberoptic catheter (Fig. 2) has 2 proximal parts, one of which is the catheter input containing the efferent or illuminating fibers; the other is the catheter output containing the afferent or measuring fibers. The efferent and afferent fibers merge at the junction disk to form the distal part of the catheter. At the tip of the catheter efferent and afferent fibers are mixed at random. The standard F6 catheter 1 contains about 200 fibers of 50 gm diameter each. The catheter tip is finished off with a stainless steel ring. Upon the tip a tripod-like cage, 4 mm long, is mounted. This cage prevents direct contact of the glass fiber endings with 1 These catheters were built bY the American Optical Corporation for use with their first in vivo oximeter. This instrument has meanwhile been replaced by a new one employing a different type of catheter [14,15].
Fig. 3. Light source and detection units. The top and side-walls as well as the electronic circuits of the detection unit have been taken away. A lamp housing; B microscope objective (20 x ); C holder for the fiberoptic catheter input with centering device; D as C, for the catheter output; E Abbe condenser; F dichroic mirror; G and G' interference filters; H and H' filter and lens h o l d e r s ; / a n d I' microscope objectives (20 x ); K and K' photocell holders
the vessel wall. In some of the catheters we replaced the stainless steel tip (with cage) by a platinum tip (without cage) to be used as an oxygen electrode [11,121. A 12 V/100 W iodine quartz lamp (Philips 7023) fed by a stabilized DC power supply (Delta Elektronika D10) is used as a light source (Fig.3). The lens system of the light source consists of a combination of two positive lenses and a 20 x microscope objective. In the detection unit (Fig. 3) the light coming out of the fiberoptic catheter passes through a 1.25 N.A. Abbe condenser to a dichroic mirror which transmits most of the light with 2 < 800 rim, whereas most of the light with 2 > 900 nm is reflected (Fig. 4, curve C). The transmitted and reflected beams pass through interference filters and are focused by a 20 dioptre lens and a 20 x microscope objective on a silicium
M. L. J. Landsman et al. : Fiberoptic Reflection Oximeter
275
outpa~
600
700
.800
900
1000
II00
Fig. 4. Spectral sensitivity of the combination of light source, fiberoptic catheter (glass fibers) and photocell (A Siemens BPY11 silicium photocell; B Motorola MRD500 photocell, see appendix) and transmission spectra of dichroic mirror (Grubb Parsons) under 45 ~ angle (C) and interference filters (Baird Atomic type B3), nominal wavelengths 640 (D), 800 (E) and 920 nm (/7). Measuring instrument: Beckman DU spectrophotometer
barrier layer photocell measuring I x 2mm (Siemens BPY 11). The spectral characteristics of the combination of light source, catheter and photocell, as well as the transmission spectra of the filters are shown in Figure 4. In the measuring and compensating channel, filters are used with nominal wawelengths of 640 and 920 nm. Before transmission to the processing unit, the photocell signals pass through preamplifiers in the detection unit. In the processing unit (Fig. 5) the outputs are adjusted for zero offset of the first amplification stage and/or cross-talk, i.e., direct transmission of light from illuminating to measuring fibers. With the catheter tip in Indian ink the output of the second amplification stage in the measuring and the compensating channels is made zero (Fig. 5, m and c, comp.). When there is no appreciable crossl;alk, this adjustment may be done with the light shut off. The gain of the photocell amplifiers can be adjusted at two levels. At the first level gains rn and c are used for setting the ratio of measuring and compensating photocellamplifier outputs, while the fiberoptic catheter tip faces a surface mirror serving as a reflection standard. At the second level the gain controls of the two channels are coupled in a double 10-turn potentiometer (Fig. 5, gain m, c), so that adjustment at this level
Fig. 5. Diagram of the circuits of the processing unit of the fiberoptic oximeter. The photocell signals which passed preamplifiers in the detection unit, are connected to the input. The first block contains the zero offset (comp.) and the first level gain controls. Gain rn, c is the coupled gain control which does not change the ratio of the measuring and the compensating signal. With selector switch (1) the functions zero control (too and co), calibration (adj.) of the recorder outputs Rm and Re, and operation (rec.) can be selected. The amplifiers AMP. serve as preamplifiers for the recorder, with a zero control, a stepwise attenuator (art.), and a continuous sensitivity control (sens.). The divider output passes an adjustable filter, and a linearizing circuit consisting of an adjustable bias and a logarithmic amplifier (LOG.AMP.). With selector switch (2) the divider outputs for 0 (/) and 100 ~ (h) oxygen saturation are connected to the linearizing system; with (2) at 1 the oxygen saturation meter (So2) or digital voltmeter (D VM) is adjusted to zero by zero suppression (l-adj.), with (2) at h the meter is adjusted to 100 with the gain control (h-adj.)
does not change the ratio of measuring and compensating signals. With the catheter in the blood stream the photocell-amplifier outputs can thus be adjusted to keep the signals below 10 V. The divider gives the ratio of the measuring and the compensating signals (Rm/R~). This is the actual, non-linear oximeter output. After electronic filtering the signal passes through a linearizing circuit so that after proper calibration the oxygen saturation can also be read from a digital voltmeter (DVM). The relation between oximeter output and oxygen saturation is made linear by subtracting a constant voltage (bias) from the divider output before taking the logarithm. The magnitude of the bias depends on position and shape of the calibration line. Calculation of the bias voltage can be done as follows. Between the divider output y and the oxygen saturation x the following relation is assumed: log(y-
c) = p . x + q
(1)
where c is the bias voltage and p and q are constants. Three corresponding values of x and y are read from
276
10
Pflfigers Arch. 373 (1978)
E~ ~Rm - (V)
2.0
Rm +1 (V) .k'LogEl~c
10
| i
So, (*/*) i 100
1.01
So, (%1 100
8 6
k.tog (El-~-c m -c)+1 (V)
b.[k.[og (E1R~- c)+1- a] (V)
2.0
4
CHb(g.[_, ) -11.8 "]90 0
100
Fig. 6. Graphic representation of steps in the linearizing system. R,, and R~ measuring and compensating photocell-amplifier outputs, E1 = scaling voltage of divider (= 10 V), k = 1 V c = bias voltage,a = divideroutput for 0 oxygensaturation ("low" in Fig. 5), b = gain factor (h-adj. in Fig. 5), So2 = oxygen saturation. Graphs 1, 2 and 3 show the steps followed by the system. Graph 2a shows the result when the logarithm of the divider output is taken directly
the calibration line yielding 3 equations with 3 unknowns (p, q and c) from which c can be solved. It is obvious to choose for x the values 100, 50 and 0. When ytoo, ys0 and yo are the corresponding y values, then c =
y5o 2 - ytoo 9yo
i
40
80
i
i
i
i
120
160
200
21,0
Fig. 7. Relationship between fiberoptic light reflection at 2 = 640 n m (R64~ 2 = 920 nm (R 9z~ as well as the ratio of light refections (R6r176176 and haemoglobin concentration (cnb). Measurements of flowing heparinized oxygenated pig blood. The So2 scale gives the apparent change in oxygen saturation
So=(%) 100
i
(2)
2 yso - yaoo - Yo Figure 6 shows the steps taken in the linearizing system. The divider outputs for 0 and 100% oxygen saturation are set at the low and high dials respectively (Fig. 5, high and low). The adjustment of the linearizing system is thus carried out with the actual oximeter output disconnected 9
PROPERTIES AND PERFORMANCE OF T H E O X I M E T E R Dependence o f Blood Reflection at ,~ = 640 and 920 nm on Total Haemoglobin Concentration
Fresh heparinized oxygenated pig blood was used to prepare samples with different haemoglobin concentrations by mixing cells and plasma in different pro-
portions. The samples were drawn through a black plastic cuvette using a H a r v a r d constant withdrawal pump. The tip of a fiberoptic catheter was situated inside the cuvette. The reflections at 2 = 640 nm (R 6r and 920 nm (R 92~ and the ratio R64~ 9z0 were recorded for all samples. The total haemoglobin concentration was measured as haemiglobincyanide [3] using a Vitatron H b F100. The relation between blood reflection and haemoglobin concentration (Fig. 7) is essentially the same as described by M o o k et al. [8]. The oximeter output R64~ 920 shows a plateau between 90 and 140 g . 1-1. Towards lower haemoglobin concentration the ratio decreases, to-' wards higher concentration it increases9 The result is that the apparent oxygen saturation is 92 and 104 % at 40 and 240 g . 1-1 respectively. Between 80 and 190 g - 1-1, however, the apparent oxygen saturation changes from 98 - 102 %. To study the effect of changes in oxygen saturation on the relation between haemoglobin concentration and ox~meter output, two portions of heparinized pig blood with different haemoglobin concentrations (99 and 181 g 9 1-1) were prepared. F r o m each portion 8 samples with different oxygen saturation were made. The oxygen saturation was determined with a Radiometer oxygen saturation meter (OSM1), which had previously been checked using a spectrophotometric two-wavelength method 9 Figure 8 shows that at lower oxygen saturation the dependence of the oximeter output on haemoglobin concentration increases9 At 9
/
M. L. J. Landsman eta]. : Fiberoptic Reflection Oximeter Log
277
10R6{~92~
0.~
/
0.E
A
"0 0A 190 %2{O/o)
So2(%) 0.~
,
20 i
,
40 i
J
60 t
,
B0 t
i
/
80 60 10
100 ,
Fig. 8. Relationship between fiberoptic oximeter output (R64~ R920) and oxygen saturation (So~) at haemoglobin concentrations of 181 g 91 -~ (dots) and 99 g. 1-~ (triangles). The logarithm of the oximeter output ( x 10) is plotted against oxygen saturation. Measurements of flowing heparinized pig blood. The factor 10 has been introduced to keep the figures positive
0 ll]R {V)
R~~
I00[$oP/o) about 9 5 ~ saturation variation in haemoglobin concentration does not change the oximeter output; at 5 0 ~ saturation, however, there is an apparent decrease in oxygen saturation of" about 1 2 ~ when haemoglobin concentration increases from 99 to 181 g 91 - I .
Dependence of the Oximeter Output on Changes in Blood Flow Velocity The effect of changes in blood flow velocity was tested in vitro using heparinized dog blood samples with oxygen saturations 100 and 30 ~ . The tip of the fiberoptic catheter was placed inside a black plastic cuvette through which the blood was drawn by hand from a well stirred vessel using a 50 ml syringe. R 640, R 920 and R6r176176 w e r e recorded on an Elema mingograph EM81. The measuring and compensating signals were also displayed on a storage oscilloscope (Tektronix R564 B) used as x - y recorder. Differences in blood flow velocity of 0 - 3 0 cm - s -~ resulted in changes of the light reflection of about 50 ~ (Fig. 9). The ratio also showed variations corresponding to about 10 ~ oxygen saturation. The effect on the ratio appeared to be dependent on the oxygen saturation. With increasing reflection R64~ 92~ decreased at i00 ~/oxygen saturation and increased at 30 ~ saturation. The x - y recordings (Fig. 9, inserts) show the relation between R 6r176 and R 92~ at 100 and 30 ~ oxygen saturation. In vivo recordings generally show less flow
5)
.
Fig. 9. Effectof changes in blood flow on the fiberoptic oximeter output at 100~ (A) and 30~ (B) oxygen saturation. Heparinized dog blood. Tip of fiberoptic catheter inside a cuvette through which blood was drawn. R64~ and R9z~ = measuring and compensating photocetlamplifieroutputs9The divider output R64~ 92~is calibrated in percent oxygen saturation (So_0.The inserts are the x - y plots taken during the first part of the recording. The upper curve in the insert of B has been recorded after a 4-fold increase in gain of R64~
variations, though the variations may still be considerable. In the example of Figure 13 (measuring site in the aorta) the variations correspond to 6 ~ oxygen saturation.
Relation between Oximeter Output and Oxygen Saturation F o r the in vitro determination of the relation between the light reflections at 2 = 640 and 920 nm and the oxygen saturation, fresh heparinized pig blood was used. A portion of the native blood was oxygenated and an other portion de-oxygenated, then mixed in various proportions to obtain measurements over the whole range of 0 - 1 0 0 ~ o. The measurements were made with the fiberoptic catheter inside black plastic tubing, through which a constant flow of blood was maintained using a H a r v a r d constant infusion/with-
278
Pfl'~igers Arci~. 373 (1978)
tog10R6~0/ R920 t~110 R6z'O/R 9z0
R~~ 6~~ 0.8
0.8
x
N \
\\
\
~.(1
x
0,6 %" \ ~ ' / "
0.B
0.4 ~
3.0
"b. \ \
"~- ~ ~
9
s x
2.0
0.4
"-,-o.,.o.
-q
0'20
I
I
20
I
I
40
I
I
60
I
S%(%) 1.0 I [ 80 100
S@o/o) %
'
2;
'
4;
'
6;
'
'
Fig. 10. Relationship between the logarithm ofSberoptic oximeter output (R64~ 92~ and oxygen saturation (So2). Dots: native blood, partly oxygenated in an open flask. Crosses: equilibration with different 02, N2, CO2 mixtures in closed tonometers [17]. Open squares: partial de-oxygenation with Na2S204; for complete deoxygenation 0.4 mg Na2S204 was added per ml blood. Solid squares: reoxygenation after Na28204 treatment. The interrupted line with open circles is the result of plotting the reciprocal of the oximeter output (R92~ 64~ against oxygen saturation
Fig. l l . Relationship between the logarithm of uncompensated fiberoptic oximeter output (R64~ 92~ and oxygen saturation (So~). Data from the samples of Figure 10. For better comparison with Figure 10 R 64~ has been divided by a constant value, for which the mean of the R 92~ values of the samples has been taken
drawal pump. In this way stable readings of the blood reflection at the two wavelengths were obtained. The oxygen saturation of the samples was determined with a Radiometer oxygen saturation meter (OSM1). Figure 10 shows the logarithm of the 0ximeter output (log lOR64~ 920) as well as the reciprocal of the oximeter output (R92~ 64~ plotted against the oxygen saturation. All points fall on a slightly curved line irrespective of the way of preparation of the samples. It does appear, however, that some of the samples underwent considerable non-specific changes. This can be shown by plotting the uncompensated output R 64~ against the oxygen saturation (Fig. 11). The light reflection of the samples prepared by tonometry is lower than that of the samples to which Na2S20~ was added. From 14 different portions of pig blood, haemoglobin concentration ranging from 96-161 g. 1 t, a total of 152 samples with different oxygen saturations were prepared using Na2S204 to obtain low and zero saturation. Log 10R64~ 92~ has been related to the oxygen saturation, and the best fitting parabolic function has been calculated by the least squares method according to
Using this line as calibration line of the oximeter, for the 152 pig blood samples a standard deviation of the differences between fiberoptic oximeter and Radiometer oxygen saturation meter of 3.1 ~ saturation has been found. For 114 samples between 50 and 100 saturation the standard deviation was 2.7 ~o, and for 78 samples between 70 and 100~ it was 1.9~o saturation. In vivo measurements were done on 10 mongrel dogs in thialbarbital anaesthesia after premedication with acepromazine. In 5 experiments the animals were incubated and connected to a spirometer. Anaesthesia was maintained with intravenous thialbarbital and breathing was spontaneous. The oxygen was replaced stepwise by nitrogen to obtain low oxygen saturation. In 5 experiments the endotracheat tube was connected to a respirator and spontaneous breathing was suppressed with pancuronium bromide. The dog was given a nitrous oxide-oxygen mixture to breathe and oxygen saturation was varied by changing the N20/O2 ratio. The fiberoptic catheter was introduced into the left femoral artery and the tip was advanced into the descending aorta just above the diaphragm. An F8 Cournand catheter was introduced into the right femoral artery and the tip was positioned next to the tip of the fiberoptic catheter. Through this catheter blood samples were drawn simultaneously with the recording of the oximeter signal. The oxygen
S02 = ao + at log lOR64~ 92~ + a2 (log 10R64~176
2
(3)
where a0 = -120.0, a l = 361.7, and a2 = -130.6.
M. L. J. Landsman et al. : Fiberoptic Reflection Oximeter
saturation of the blood samples was determined with a Radiometer oxygen saturation meter (OSMI). From the results (Fig. 12) a standard deviation of 2.6 ~ oxygen saturation has been calculated.
279 tog10 R 6~'~ 920
0,~
Continuous Observation of Oxygen Saturation in the Dog In one of the experiments described above a continuous record was taken of the arterial oxygen saturation and of the change in oxygen tension (Po~). The fiberoptic catheter was provided with a platinum tip serving as a Poa-sensitive electrode [11, I2]. The dog breathed spontaneously. Oesophageal pressure was measured with a fluid-filled bird's eye catheter connected to a Statham P23Bb pressure transducer. Aortic pressure was recorded using the aortic sampling catheter connected to a Statham P23Db pressure transducer, and right atrial pressure via an F7 catheter and a P23Db transducer. Immediately before the record shown in Figure 13 a low oxygen gas mixture had been replaced by a mixture with 20 % Oa. During breathing the low oxygen gas mixture the arterial Pco~ became low due to hypoxic hyperpnoea. Following the restoration of a nearly normal Po~ in the respiratory gas mixture, apnoea occurred due to the disappearance of the hypoxic drive from the peripheral chemoreceptors. During the apnoea the arterial oxygen saturation decreased from 9 0 - 7 0 G. In the subsequent recovery period the oxygen saturation increased with each breath.
Sensitivity of the Oximeter to Indocyanine Green The error in the oxygen saturation measurement due to the presence of indocyanine green has been determined from a series of measurements on heparinized pig blood samples, oxygen saturation ranging from 3 5 - 1 0 0 % . The samples were divided into two aliquots and kept in syringes. To one of each pair indocyanine green was added (anaerobically) at a concentration of 5 mg 91-a. Measurements were made with two fiberoptic oximeters, one containing the 640/920 filter combination, the other with a filter with a nominal wavelength of 800 nm instead of 920 nm. The calibration line of the first one was used to determine the oxygen saturation of the samples (without dye) and to calibrate the second oximeter. The lines relating oxygen saturation and oximeter output for the dye containing samples were determined for both instruments. From these lines the apparent change in oxygen saturation due to the presence of dye was determined at different levels of oxygen saturation. The results (Table 1) show that for the oximeter with the 640/920 combination the error caused by the
0.6
0.4
s%(%) '
2;
'
'
6;
'
'
Fig. 12. In vivo calibration line, based on measurements on 91 blood samples obtained in 10 dog experiments. The equation of the line is So~ = -- 74.35 + 281.5 log IOR,,/R~ - 99.05 (log IOR~/Rr ~
5~:0-.J
Fig. 13. Recording in the dog of electrocardiogram (ECG H), respiration (resp., inspiration upwards), oesophageal pressure (Poes), aortic pressure (P~o), right atrial pressure (Pr,), fiberoptic oximeter output (So2), the same with electronical filtering (0.5 Hz, 3 dB), change in oxygen tension (APoO, compensating (R 92~ and measuring (R 64~ photocell outputs of the fiberoptic oximeter
280
Pftfigers Arch. 373 (1978)
Table 1. Apparent oxygen saturation ($82) and change in So2 (ASo2) at different levels (So2) caused by the addition of 5 mg - 1-1 indocyanine green, measured with fiberoptic oximeters with 640/920 and 640/800 filter combinations
So=
SG
ASo2
(640/920) 100 90 80 70 60 50 40 30
98 88 78 68 58 48 38 28
$32
AN)e
(640/800) -2 -2 -2 - 2 -2 -2 -2 -2
109 101 93 85 77 68 59 50
+ 9 + tl +13 + 15 + 17 + 18 + 19 +20
presence of the dye is slight and constant over the entire saturation range. For the oximeter with the 640/800 combination the deviation is considerable. DISCUSSION On the basis of the work of Rodrigo [18], Polanyi and coworkers [14,16], and others [2,4] assume a linear relation between oxygen saturation and the ratio of compensating (oxygen saturation independent) over measuring (oxygen saturation dependent) photocell outi3ut Rc/R~. This is not borne out by our results (Fig. 10). The fact that 2 = 920 nm, the wavelength used for compensation, is not isobestic is irrelevant in this respect. Actually the wavelength used by Polanyi, 2 = 805 nm, is not isobestic either (Fig. 1), but this is not the cause of non-linearity, as he suggests [15]. In the Schwarzer oximeter the ratio RdRm is corrected to obtain linearity of oxygen saturation and output signal [15]. The fact is that the reciprocal of the blood reflection is not linear with oxygen saturation, and this non-linearity cannot be corrected for by the second measurement. In spectrophotometry of haemolyzed blood the situation is essentially different, because, when haemoglobin concentration is constant, optical density changes linearly with oxygen saturation. This relation is not affected when, for compensation of differences in haemoglobin concentration, the optical density is divided by the optical density measured at an isobestic wavelength. Also in the conventional reflection oximeters there is no linear relation between the reciprocal of the blood reflection or the ratio of reflections R d R m and oxygen saturation [9]. So, apparently, the theory which predicts a linear relation [14,18], is not a complete description of light reflection by blood as measured with conventional reflection oximeters or by fiberoptics. In the authors' laboratory it has been traditional to relate the loga-
rithm of the reflection obtained by the original onecolour instruments with the oxygen saturation [19]. Taking the logarithm of the ratio of measuring over compensating photocell output, when employing measurements in two colours, is equivalent to this. The relation between log R,,/Rc and oxygen saturation is not linear either, neither in conventional reflection oximetry [9] nor in fiberoptic reflection oximetry (Figs. 8, 10, and 12). Linearization can be achieved by subtracting a constant voltage c from the oximeter output Rm/Rc before taking the logarithm (Fig. 6); thus log (Rm/Rc - c) or log (Rm - c 9 Rc)/Rc is linearly related to oxygen saturation. When using the inverse ratio, R d ( R ~ + c . Rc) = 1/(R,,/R~ + c) can be made linear with oxygen saturation by properly choosing the value of c. This is the procedure applied in the Schwarzer oximeter. The influence of haemoglobin concentration on the calibration line is such that at higher saturation, above 70 ~, an average calibration line may be used. For accurate measurements below 70 ~ saturation, however, the calibration should be adapted to the haemoglobin concentration; the high and low adjustments (high and low, Fig. 5) are used to this end. The influence of non-specific effects other than the haemoglobin concentration is illustrated in Figures 9, 10 and 11. The changes in reflection of dog blood in vitro (Fig. 9) can be explained by formation and breaking up of rouleaux, and to a minor extent by a directional effect on the erythrocytes. The shape of the x-y plots shows that the influence of changes in blood flow cannot be compensated for by simple means. An additional complication is the oxygen saturation dependency of the flow effect. In vivo, however, the directional effect on the erythrocytes prevails. In the great blood vessels rouleaux formation does not occur. The nonetheless appreciable flow effect can be suppressed by electronic filtering. The dynamic response of the system is then determined by the electronic filter (Fig. 13). In Figure 11 the differences in reflection may be caused by changes of the erythrocyte surface due to prolonged (about I h) contact with the glass wall of the tonometer and/or by osmotic changes due to the addition of Na2S204. Compensation for these effects is practically complete (Fig. 10). Oximeters used in the cardiac catheterization laboratory or applied to monitoring in situations where cardiac output is measured by indocyanine green dilution, should be insensitive to this dye. Thus 2 - - 8 0 0 n m cannot be used as a compensation wavelength. Table I shows the large error in the oxygen saturation measurement when 2 = 800 nm is used for compensation. When 2 = 920 nm is used instead, the error is but slight. The oximeter does
M. L. J. Landsman et al. : Fiberoptic Reflection Oximeter
not become completely insensitive to dye, because at 2 = 640 nm the light absorption by indocyanine green is not zero (Fig. 1). Fiberoptic oximetry has two important advantages over catheter cuvette oximetry. Firstly, the fiberoptic oximeter is a catheter tip oximeter and as such capable to reliably record fast changes in oxygen saturation, which may occur in the heart or great vessels, especially in cases of shunts [1]. For any cuvette outside the body the dynamic response characteristics are limited by the catheter [10]. Secondly, there is no need to withdraw blood. This reduces blood loss and, still more important, it makes the method suited for continuous observations over long periods, e.g., for several hours to some days [6]. The present instrument has some advantages over commercially available fiberoptic oximeters [14,15]. The measurements as well as the ratio computation are continuous. Other instruments employ light emitting diodes, which are pulsed. The frequency used is 2 0 0 - 3 0 0 Hz. There is only one photocell, the signal of which thus contains the light reflections at the two wavelengths. These are separated electronically. Because the two measurements Rm and Rc are not synchronous, compensation for non-specific effects is dependent on the pulse frequency. The changes in reflection due to changes in blood flow may be so fast that it is least doubtful whether a pulse frequency of 300 Hz is enough for compensation of these changes. Continuous measurement with two photocells gives the best possible compensation and dynamic response. The dynamic responses of the two photocellamplifier assemblies do not differ so much as to be the cause of limitations in this respect [7, 8]. The main advantage of the present instrume.nt, however, is the use of light filters. In this way any wavelength and bandwidth may be chosen. Measurements can be done even without filter and thus in a band determined by the spectral characteristics of the lamp, the optical fibers, and the photocell. This is applied in the fiberoptic densitometer [7]. By using light filters it is also very simple to switch from one wavelength combination to another. The fiberoptic oximeter is preferably applied to the continuous recording of arterial or central (mixed) venous oxygen saturation. Figure ~3 shows an example obtained in a physiological experiment. With the simultaneous recording of oxygen tension (x-axis) and oxygen saturation (y-axis) in vivo haemoglobinoxygen equilibrium curves can be determined [13]. APPENDIX A more recent development of the fiberoptic reflectometer consists of a combined light source and detection
281
C
B D
/
E
8~
.
..
Aj
6/-.0~ , ~F 8 0 0 ~
920 920 - nf
/ E
i
