Vol. 273, Issue 6, H2910-H2918, December 1997
SPECIAL COMMUNICATION
CCD imaging in cryospectrophotometric determination of
microvascular oxyhemoglobin saturations
Kårstein
Måseide and
Einar K.
Rofstad
Institute for Cancer Research and The Norwegian Cancer Society, The
Norwegian Radium Hospital, Montebello, 0310 Oslo, Norway
 |
ABSTRACT |
A
microspectrophotometric imaging method has been developed for localized
measurements of intravascular oxyhemoglobin
(HbO2) saturations in
microvessels from sections of quick-frozen tissue. HbO2 saturation was calculated
from the absorption spectrum of red blood cells measured at five
selected wavelengths in the 520- to 570-nm range. We combined the use
of narrow-bandwidth interference filters and a CCD camera mounted on a
microscope to obtain one gray image of the sample at each wavelength.
Each pixel is a quantitative measure of transmitted light intensity
from the tissue sample at that location. A linear calibration curve for
blood frozen in vitro (humans and mice) and in vivo (mice) was obtained
using a multicomponent analysis. Oxy- and deoxyhemoglobin were assumed to be the only hemoglobin components present. A constant term compensates for light loss due to scattering on red blood cells and ice
crystals. The standard error in single measurements of HbO2 saturation was 5%. The
present method allows off-line analysis of the
HbO2 saturation distribution
within a microvessel network and offers new possibilities for
comparative morphological studies.
oxygen; optical spectroscopy; intravascular oxyhemoglobin
saturation; blood; oximetry
| |
INTRODUCTION |
SEVERAL TRANSMISSION spectrophotometric techniques have
been reported for the measurement of oxyhemoglobin
(HbO2) saturation. It is widely
accepted that the Lambert-Beer law of light absorption holds for
hemoglobin in solution (8, 29) but not for whole blood or red blood
cell (RBC) suspensions (11). The most prominent phenomenon in
particulate suspensions in contrast to solutions is the scattering of
light. A theoretical basis for separation of the absorption and
scattering terms for transmitted light through suspensions of
light-absorbing particles was formulated by Twersky (25-28).
Anderson and Sekelj (1) showed that for transmitted light the optical
density of an RBC suspension can be expressed as the sum of the
absorption and scattering terms as theorized by Twersky.
A three-wavelength transmitted light method for determination of
HbO2 saturation in whole blood was
developed by Pittman and Duling (16) and applied to microvessels in
vivo (17). By comparing optical densities from absorption spectra of
RBC suspensions and hemoglobin solutions at several wavelengths, they
showed that light scattering in the 510- to 575-nm range is wavelength
independent. Hence, the scattering term could be calculated from
optical density values at two isosbestic wavelengths. A third
nonisosbestic wavelength was used to calculate the
HbO2 saturation on the basis of
the scattering-corrected optical densities. Percent
HbO2 saturation measured by this
method is independent of optical path length and hematocrit. Sinha et
al. developed a similar three-wavelength method for determination of
HbO2 saturations in frozen blood
(21) and in small arteries and veins of quick-frozen tissue (22).
Grunewald and Lübbers (6) utilized the whole absorption spectrum
in the 520- to 620-nm range to determine
HbO2 saturation in capillaries of
a frozen organ sample. The measured absorption spectrum was
reconstructed from known basic oxygenated and deoxygenated hemoglobin
spectra by a least-squares curve fit, known as the weighted
multicomponent analysis of Lübbers and Wodick (12, 13). Light
scattering on frozen blood is attributed not only to the RBCs, but also
to ice crystals within the cells. Whereas Sinha et al. (21, 22) used a
constant scattering term, Grunewald and Lübbers used a polynomial
of first order to describe the light scattering. The best curve fit was
obtained when the spectrum of deoxygenated, dehydrated hemoglobin was
used as a component in addition to the basic oxygenated and
deoxygenated hemoglobin spectra.
In previous works, spectrophotometers with prism or grating
monochromators and photomultiplying tube (PMT) detectors were used for
spot measurements of HbO2
saturation in microvessels of frozen samples (3, 5, 6, 22). Zhu and
Weiss (31) used high-transmittance narrow-bandwidth interference
filters as monochromators for spectrophotometric determination of oxy- and carboxyhemoglobin in frozen microvessels. Pries et al. (18) combined the use of interference filters with a video system to obtain
an optical density image of the field of view for continuous and
off-line hematocrit determination in small vessels.
We have developed a new imaging technique for determination of
HbO2 saturations in single
microvessels of thin frozen tissue slices. Digital images of
transmitted light intensity distribution in frozen sections at five
selected wavelengths were recorded by using a black and white (B/W)
charge-coupled device (CCD) camera in combination with narrow-bandwidth
interference filters. Absorption spectra were constructed from gray
value measurements from selected areas inside vessel profiles, and
HbO2 saturations were calculated by use of the multicomponent analysis of Lübbers and Wodick (12, 13).
Our imaging technique allows rapid scanning of tissue sections, since
one image usually covers several vessels, the number of wavelengths is
small, and analysis is done off-line after image acquisition and
storage are completed. HbO2 values
of segmented areas can be evaluated, and the segmentation can be
repeated if desirable, in contrast to real-time spot measurements of
PMT systems. The combination of interference filters and a CCD camera
surpasses PMT systems as well as video systems with respect to light
sensitivity and linearity. The high spatial resolution of CCD images
gives new possibilities for relating
HbO2 saturation distribution to morphology by comparison with parallel histological sections. CCD
imaging has recently been used for
HbO2 saturation determination in
vivo (10, 23). Because of blood flow during the recording time of those
spectra, the in vivo technique gives a time-averaged saturation measure
from many RBCs, in contrast to our technique, which gives an
instantaneous saturation measure from only a few fixed RBCs in a
microvessel profile.
| |
MATERIALS AND METHODS |
Cryomicrospectrophotometer.
The cryomicrospectrophotometer setup is shown schematically in Fig.
1. A Zeiss Axioscope with a 100-W halogen
lamp and a ×20 (NA 0.40) long-working-distance objective was
used. Illumination at the desired wavelengths was obtained by placing
narrow-bandwidth interference filters (bandwidth at half-maximum of 5 nm, Delta lights and optics) on top of the field diaphragm. Because
these filters have additional transmission bands in the infrared
region, a broad-bandwidth interference filter (480-590 nm) was
placed behind the objective to prevent infrared light from reaching the detector. A beam splitter on top of the microscope split the microscope image onto a B/W CCD camera (model 4710 CCIR, Cohu) and an RGB color
video CCD camera (model DXC-151P, Sony). The B/W camera was used to
acquire gray images of transmitted light intensity distribution in the
sample at the different wavelengths. After removal of the filters, a
color image of the sample was acquired with the RGB camera to
facilitate identification of vessel profiles. The CCD cameras were
connected to a personal computer (PC) equipped with a frame grabber
(Kontron Systems) and an image analysis program (KS300, Kontron
Systems). Live microscope images were shown on a separate monitor;
acquired and manipulated images were shown on the PC monitor.
Connection to a data network gave several possibilities for image
storage and printing.
View larger version (32K):
[in this window]
[in a new window]
|
Fig. 1.
Cryomicrospectrophotometer setup for imaging of frozen tissue sections
to measure hemoglobin absorption spectra in microvessels. The basic
components are a microscope with interference filters, charge-coupled
device (CCD) cameras connected to a personal computer (PC) with image
analysis software, and a cooling chamber with accessories. B/W, black
and white.
|
|
The tissue sections were kept cold during measurement in a biologic
cryostage (model BCS 196, Linkam). The sample was placed between two
cover glasses in a sample carrier on a temperature-regulated silver
block. Cooling was obtained by pumping cold
N2 vapor from a Dewar flask filled
with liquid N2 through the silver
block by use of a pump (model LNP 2, Linkam). The temperature was
measured and the silver block was heated by a built-in thermistor. The stage was connected to a PC via a computer interface (model CI 93, Linkam). LINK 2.0 software (Linkam) quickly regulated the temperature
to within 0.1°C by adjusting the pump speed and thermistor current.
Temperature profiles with cooling/heating rates up to 130°C/min
could be programmed. A condenser top lens built into the silver block
provided perfect optical conditions when used in combination with a
Zeiss condenser (NA 0.9). The sample carrier was moved in the
x,y-plane by manipulators, and dry
N2 gas was flushed through the
stage via special connections to avoid riming inside the stage. Our
stage was specially designed with sample carriers for square cover
glasses, extra-long tubing, and a bayonet-fitting lid for quick sample
loading inside a cryotome. Dry N2
gas was flushed onto the stage top and bottom windows to
avoid dew formation.
Determination of HbO2 saturations from
hemoglobin absorption spectra.
Several mathematical models have been utilized for calculating
HbO2 saturations from absorption
spectra of frozen blood samples. The simplest model assumes that
oxygenated (HbO2) and
deoxygenated (Hb) hemoglobin are the only hemoglobin components present
in the sample and that light scattering from RBCs and ice crystals is
constant (B0)
over a limited wavelength range (21). The
HbO2 saturation can be calculated
from optical density measurements at three selected wavelengths in this
range by solving three equations with three unknowns
(HbO2, Hb, and
B0). One extra
measuring wavelength is required for each additional component that is
taken into account. Grunewald and Lübbers (6) used a total of
five components by adding a first-order wavelength-dependent term
(B1) and the spectrum of deoxygenated, dehydrated hemoglobin. The multicomponent analysis of Lübbers and Wodick (12, 13) was applied to the continuous spectrum. By performing a least-squares curve fit, an
adjusted spectrum was constructed, from which the
HbO2 saturation was calculated.
We have modified the multicomponent analysis of Lübbers and
Wodick (12) to be applied for a discrete number of wavelengths by
noting that a least-squares curve fit can be done as long as the number
of measuring wavelengths (M) exceeds
the number of basic components (N).
We assume that the measured spectrum can be expressed as a sum of
N basic spectra. The optical density (D) at a certain wavelength (
) is given by
where
c
is the concentration of
component 
, 
() is the
millimolar extinction coefficient of component at wavelength ,
and d is the optical path length. A
least-squares fit of a combination of the basic spectra to the measured
spectrum f() at
M discrete measuring wavelengths is
given by the following minimization
where
(i) is a weight function.
Minimization is obtained when the partial derivatives with respect to
all c are zero, i.e.
These
N equations with
N unknowns
c can be solved by matrix
operations. When M = N, the method reduces to the exact solution instead of a curve fit for the
N unknown components. With this
general formalism, scattering terms are treated as component spectra,
which means that for a constant scattering term
B0 we set
() d = 1 and obtain
For a
first-order wavelength-dependent scattering term we set
() d = , which yields
In this study we were in principle interested in seven basic
components to describe the measured spectra of blood in frozen samples:
1) oxygenated hemoglobin at room
temperature (HbO2), 2) deoxygenated hemoglobin at room
temperature (Hb), 3) deoxygenated, dehydrated hemoglobin at room temperature (dHb),
4) oxygenated hemoglobin of frozen
blood at low temperature (HbOf2), 5) deoxygenated hemoglobin of frozen
blood at low temperature (Hbf),
6) a constant scattering term
(B0), and
7) a first-order
wavelength-dependent scattering term
(B1). We used a
total of five wavelength filters: three were close to isosbestic ( = 523, 549, and 569 nm), and two nonisosbestic measuring wavelengths were
chosen for maximal difference between oxygenated and deoxygenated
spectra ( = 538 and 560 nm). We tested several combinations of basic
spectra and wavelengths to obtain the best mathematical model for
HbO2 saturation determination.
With the use of all five measuring wavelengths, it was possible to
perform exact calculations with five components or a least-squares
curve fit with three or four components. We also performed exact
calculations with three and four wavelengths.
Sample collection, preparation, and storage.
Pure hemoglobin solution with no scattering elements was used to
determine the extinction coefficients of the basic hemoglobin spectra
at room temperature. Hemoglobin solutions were made from fresh human
blood removed by venipuncture and tapped in Vacutainers containing
heparin as anticoagulant. Hemoglobin concentration of the blood was
determined spectrophotometrically on an automatic cell counter
(CELL-DYN 3500, Abbott). The blood was stored at 4°C until use.
Serum, white blood cells, and platelets were removed after
centrifugation at 2,500 rpm. The RBCs were washed three times in
isotonic saline solution and then lysed by addition of distilled water
to approximately three times the original blood volume. The hemoglobin
solution was purified by repeated centrifugation and separation of the
supernatant from the pellet of cell fragments. The supernatant was
filtered through 10-µm monofilament filters (Nytal) before each
centrifugation and finally filtered through 0.2-µm filters
(Microgon). Purity and concentration of the hemoglobin solution were
checked on the automatic cell counter. The hemoglobin solutions were
stored at 4°C until use.
Hemoglobin solution of ~0% saturation was obtained by equilibration
with 5% CO2-95%
N2 for 30 min in a closed
Vacutainer. The hemoglobin solution was pumped into flow cuvettes
(d = 0.5 mm, Helma Skandinavia) on the
microscope stage. After image acquisition, the hemoglobin solution was
pumped back into the Vacutainer and flushed with gas for a few more
minutes, and the procedure was repeated. To obtain hemoglobin solution
of ~100% at the same hemoglobin concentration, the hemoglobin
solution was pumped back into the Vacutainer and equilibrated with pure
O2 gas for ~5 min before new
measurements. The spectrum of deoxygenated, dehydrated hemoglobin was
measured on hemoglobin sheets formed after a drop of hemoglobin solution in the biologic cryostage was flushed with 5%
CO2-95% N2.
Frozen blood drops were used to determine the basic spectra for fully
oxygenated and deoxygenated blood at low temperature (
100°C). These spectra and spectra of blood drops of
intermediate saturations were used to calculate the calibration curve
of the cryomicrospectrophotometer. Fresh human blood was removed by
venipuncture and tapped in Vacutainers containing heparin as
anticoagulant. Fresh blood from anesthetized mice [1:1
fentanyl/fluanison (Hypnorm; Jansen)-midazolam (Dormicum; Roche),
0.050-0.075 ml/10 g sc] was drawn into
heparinized syringes by cardiac puncture. Hemoglobin concentration of
the blood was determined spectrophotometrically on the automatic cell
counter. The blood was stored at 4°C until use. Oxygenated blood
samples were obtained by equilibration with pure
O2 gas for 30 min in a closed
Vacutainer. Deoxygenated samples were obtained by equilibration with
5% CO2-95%
N2 for 30 min. Samples of
intermediate saturations were obtained by using shorter flushing times
and by alternating flushing with the two gas mixtures. Approximately
0.5 ml of the blood was drawn into a syringe, and a few drops were
rapidly frozen at 196°C by immediate immersion into liquid
N2. An aliquot of blood was drawn
from the closed Vacutainer into a syringe and within 15 min was
transferred to a CO-oximeter (model 855, Ciba Corning) for
determination of an HbO2
saturation reference value.
Blood frozen directly in the left and right ventricles of anesthetized
mice was used to obtain in vivo samples of high and low saturations for
the calibration curve. Immediately before freezing, an aliquot of blood
from the actual ventricle was drawn into a heparinized syringe for
HbO2 saturation determination by CO-oximetry. The heart was quickly frozen in situ by bringing a copper
block precooled in liquid N2 into
contact with the heart. The deep-frozen heart was cut out and immersed
in liquid N2, where it was stored
until it was used. In vivo samples of zero saturation were obtained by
immersing the liver and human melanoma xenografts in liquid
N2 5 min after mice were killed.
In vivo samples of quick-frozen normal and tumor tissue were obtained
by application of copper blocks precooled in liquid N2 to the tissue in situ after the
mice were anesthetized (20). Normal tissue samples were obtained from
the femoral muscle; tumor samples were obtained from human melanoma
xenografts (R-18) grown intradermally in athymic mice (19).
Sample mounting, sectioning, and transfer.
The same procedure was followed for all frozen samples with respect to
mounting, sectioning, and transfer of the sections to the biologic
cryostage. The samples were cut by a scalpel under liquid
N2 into pieces with external
dimensions of 2-10 mm. A specially designed stainless steel sample
holder covered with embedding medium (Tissue-Tek OCT compound, Miles)
was precooled by partial immersion in liquid
N2. One end of the sample was
lowered into the upper fluid layer of the embedding medium before the
whole assembly was quickly immersed in liquid
N2. By this procedure the sample
was fixed without being removed from liquid
N2 for >10 s.
The biologic cryostage was flushed with dry
N2 gas and then sealed to avoid
riming inside the stage. It was cooled to 58°C and placed
inside a cryotome (Slee). The sample holder was mounted inside the
cryotome, where the sample was cut into 12- to 16-µm-thick sections
at 60°C. The sections were drawn by fine brushes onto a
precooled square cover glass and covered with a smaller precooled cover
glass within 30 s after sectioning. The lid of the biologic cryostage
was opened, and the sample-cover glass assembly was placed in the
sample carrier on the silver block before the lid was closed. By
keeping opening time short (<1 min) and keeping the stage temperature
slightly above the cryotome temperature, riming on the silver block was
minimized. The biologic cryostage was cooled to the measuring
temperature of 100°C within 1 min and mounted onto the
microscope. Total exposure time of the sample to 60°C in the
cryotome was 7-8 min. The sample holder was immersed and stored in
liquid N2 until further
sectioning.
Light intensity measurements.
For image acquisition, Köhler illumination was used to obtain
even and equal illumination of the specimen. The narrow-bandwidth interference filters were placed in specially designed filter holders
on the field diaphragm to keep all exchange of filters to the
illumination side of the specimen. Possible disturbances on this side
will not cause image displacement, as would be the case if there were
movements between the object and the CCD camera. The condenser aperture
diaphragm and the field stop were kept constant during measurements.
Focusing was performed at the central wavelength of 549 nm to minimize
the chromatic aberration at all wavelengths. The microscope lamp was
operated close to its maximal voltage to obtain sufficient intensity of
the transmitted light. Dry N2 gas
was flushed onto the windows of the biologic cryostage to avoid riming.
Before sample loading, one offset image representing the black current
of the CCD camera was acquired while the light path was blocked. Six
reference images were acquired in the following sequence: = 549, 560, 569, 523, 538, and 549 nm. By repeating the measurement at 549 nm,
possible drift in the measurements could be detected. Reference images
for hemoglobin solutions and frozen samples were acquired with a
cuvette filled with distilled water and with no sample in the biologic
cryostage, respectively. A gray filter in the light path was used for
the reference images to prevent the CCD camera from reaching its
saturation level. From each measuring site in the specimen, six sample
images ( = 549, 560, 569, 523, 538, and 549 nm) and one color image
were acquired. The number of measuring sites for one sample was two along the hemoglobin-filled cuvettes, four for frozen blood sections, and as many as necessary to cover all vessel profiles in the tissue sections. Total measuring time at one site was ~1 min. All images were stored for off-line image analysis.
Image analysis.
The reference and sample images were offset corrected by subtracting
the offset intensity image pixel by pixel. The offset-corrected sample
images were divided pixel by pixel with the offset-corrected reference
images for the corresponding wavelength to obtain a transmittance
image. If there were lateral displacements between the
transmittance images from the same site, they were manually translated until they visually overlapped. Segmentation of measuring areas was performed in one of three ways, depending on the sample type.
For the homogeneous hemoglobin solutions in cuvettes, a central
circular area covering roughly one-third of the image area was
segmented, and mean gray value was measured automatically. The frozen
blood sections showed some heterogeneity because of uneven distribution
and orientation of RBCs. Circular areas covering homogeneous optically
dense regions slightly less than the size of a single RBC were
segmented. The mean gray value in each of these regions was recorded
automatically. In tissue sections, vessel profiles were identified as
red regions in the color image and as optically dense regions in the
gray images. Mean gray values were measured on areas segmented by
setting gray value thresholds, then eroding to smooth the regions.
| |
RESULTS |
Determination of basic hemoglobin spectra.
Optical densities were measured on oxy- and deoxygenated hemoglobin
solutions in cuvettes under equal conditions, i.e., the same
concentration and optical path length. Thus the extinction coefficients
of the two spectra calculated from the Lambert-Beer law could be
compared quantitatively. The extinction coefficients were the slope of
the linear regression curves of D vs.
cd for repeated measurements on
hemoglobin solutions of various concentrations. The resulting spectra
for oxy- and deoxygenated hemoglobin are shown in Fig.
2A. The
intersections of the two spectra indicate that the three
"isosbestic" wavelength filters lie 1-2 nm above the true
isosbestic values. Single spectra of samples flushed with the same gas
mixture were highly parallel when plotted semilogarithmically (not
shown), indicating only small variations in saturations between different samples. The linear regressions of optical density vs. hemoglobin concentration at each wavelength showed that the
Lambert-Beer law was obeyed
(r2 > 0.991 for
all extinction coefficients; data not shown). The lack of scattering
elements in the solutions was confirmed by obtaining identical
absorption spectra on spectrophotometers with and without integrating
sphere (data not shown).
View larger version (12K):
[in this window]
[in a new window]
|
Fig. 2.
A: basic spectra of oxygenated
(HbO2), deoxygenated (Hb), and
deoxygenated, dehydrated hemoglobin (dHb) solutions at room
temperature. B: adjusted basic spectra
of oxygenated and deoxygenated frozen blood sections at low temperature
(100°C). , Wavelength; , extinction coefficient;
HbOf2,
HbO2 of frozen blood at low
temperature; Hbf, Hb of frozen blood at low
temperature.
|
|
Absorption spectra of deoxygenated, dehydrated hemoglobin were measured
on hemoglobin sheets formed by drying a drop of hemoglobin in an
O2-free atmosphere. Extinction
coefficients could not be calculated directly by the Lambert-Beer law,
since hemoglobin concentration and optical path length were not known.
Extinction coefficients were estimated by averaging single spectra and
multiplying the optical density values by a constant so that the
spectrum fitted the same range of ordinate values as the two other room temperature spectra. The resulting spectrum showed a sharper absorption maximum at ~560 nm and was elsewhere flattened compared with the spectrum of deoxygenated hemoglobin (Fig.
2A).
The spectra of oxygenated and deoxygenated frozen human blood showed
10-fold higher optical densities than the corresponding spectra of
hemoglobin solutions at room temperature. The shapes of the spectra
were also changed. Spectra of single measurements were highly parallel,
indicating only small variations in saturation values. Extinction
coefficients could not be calculated exactly because of unknown
hemoglobin concentration and optical path length at the measuring
spots. Single, parallel spectra of spots showing high saturation were
averaged, as were deoxygenated spectra. By assuming a
hemoglobin concentration similar to that of whole blood and an optical
path length similar to the thickness setting on the cryotome, estimated
extinction coefficients were calculated. By adjusting the estimated
scattering term, the spectra were provided to fit the same range of
ordinate values as the hemoglobin room temperature spectra (Fig.
2B).
Selection of mathematical model for HbO2
saturation calculations.
The calibration curve for the cryomicrospectrophotometer was based on
spectra recorded from sections of human blood frozen in vitro, blood
from mice frozen in vitro, and blood from mice frozen in vivo. The in
vitro samples were equilibrated with different gas mixtures to obtain
oxyhemoglobin saturations ranging from ~0% to ~100%. In vivo
samples of high and low saturations were obtained from mice by freezing
the left and right ventricles after having drawn reference samples.
Reference saturation values were measured by CO-oximetry. For in vivo
samples from normal and tumor tissue of mice that had been killed, all
O2 was assumed to be consumed
after 5 min, and the reference values were put to zero.
Saturations were calculated for 19 different combinations of selected
wavelengths and component spectra by the unweighted multicomponent
analysis. The extinction coefficients used for HbO2 saturation calculations were
taken from the component spectra in Fig. 2. Linear regressions for
calculated saturations of the frozen samples vs. reference saturations
were performed. The ordinate value of each data point represented mean
saturation from 
20 spots on a frozen section. The abscissa values
were means of two or three CO-oximeter readings. The results of the
regression analyses are shown in Table 1,
indicating that only combinations 2, 4, 15, and 17 have
ordinate intersection
(b0) close to
0, slope (b1) close to 1, and
r2 close to 1. All four combinations are based on the low-temperature spectra plus
scattering terms only, and they include the wavelengths 523 and 549 nm.
Best correlation was obtained when all five wavelengths were exploited
for a least-squares curve fit. A first-order wavelength-dependent scattering term
(B1) gave no
improvement compared with the constant scattering term
(B0) only. The
addition of the spectrum of dehydrated, deoxygenated hemoglobin gave no
good model for saturation determinations on frozen samples when used in
combination with the room temperature spectra or the low-temperature
spectra. In general, adding more component spectra to the three
components used in model 2 resulted in
aggravation of the regression curves.
View this table:
[in this window]
[in a new window]
|
Table 1.
Linear regression data for HbO2 saturations of frozen blood
sections measured by cryomicrospectrophotometry vs. blood at room
temperature measured by CO-oximetry
|
|
On the basis of the regression data in Table 1, method
2 was selected for subsequent calculations. The
corresponding linear regression curve (Fig.
3) shows that all points fit close to the regression line and that there is no significant difference between human and mouse blood samples, nor is there a significant difference between in vitro and in vivo samples. The standard errors for the
ordinate intersection and the slope were 1.4 and 0.03, respectively. The standard error for the HbO2
saturation determination in a single measurement was <5% over the
whole range of saturation values.
View larger version (25K):
[in this window]
[in a new window]
|
Fig. 3.
Linear regression curve for HbO2
saturations of human blood frozen in vitro, mouse blood frozen in
vitro, and mouse blood frozen in vivo determined by
cryomicrospectrophotometry (CMSP) vs.
HbO2 saturations of corresponding
ex vivo reference samples determined by CO-oximetry.
|
|
Application of the method to tissue samples.
Relative frequency distributions of the
HbO2 saturations in single vessels
of normal and tumor tissue are shown in Fig.
4, A and
B, respectively. Each count represents
the mean HbO2 saturation in a
microvessel profile in the tissue section. Sections from the same
sample were taken at least 50 µm apart. For vessels lying in the
section plane, measurements were performed on several sites only when
the sites were >50 µm apart. The normal tissue showed a symmetrical
distribution from 40% to 110% with a peak at ~70-80%. In
addition, there were two vessels, probably venules, with saturations slightly <20%. The tumor tissue saturations were spread over the whole range, but saturations were <20% in 67% of the vessels. A
Mann-Whitney rank-sum test showed that the
HbO2 saturation was significantly
lower in tumor tissue than in normal tissue
(P < 0.0001).
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 4.
Relative frequency distribution of microvascular
HbO2 saturations in normal femoral
muscle tissue (A) and human melanoma
xenografts (R-18) grown intradermally in athymic mice
(B).
|
|
| |
DISCUSSION |
We have described a cryomicrospectrophotometric method for measuring
HbO2 saturations in single
microvessels by imaging tissue sections at five different illumination
wavelengths.
Spectral analysis.
The 500- to 600-nm range of the spectrum gives optimal optical density
values for microvessels with respect to the sensitivity of the CCD
camera (14, 29). Narrow-bandwidth interference filters were selected in
this range at three isosbestic wavelengths reported in the literature
and at two intermediate nonisosbestic measuring wavelengths (15, 21,
29). The measured spectra of hemoglobin solutions at room temperature
indicate that the isosbestic filters lie 1-2 nm above the true
isosbestic values (Fig. 2A). The
observed parallelism between single spectra and the good linear
regressions of D vs. cd indicate that
the saturation was the same in all samples. Such a stability can only
be expected close to 0% and 100% saturations, where the
O2 dissociation curve is
flattened. In contrast to some previous works (6, 29), we did not use
sodium dithionite
(Na2S2O4)
for deoxygenated samples, since it is very difficult to find the right
concentrations. Too little dithionite will not provide complete
desaturation; too much can cause light scattering due to aggregation
and eventually cause chemical side reactions.
The spectra of frozen blood samples had significantly higher optical
densities than the room temperature spectra because of light
scattering. In addition, there were changes in the shape of the spectra
that cannot be explained by a wavelength-dependent scattering term or
by the presence of deoxygenated, dehydrated hemoglobin (Table 1, Fig.
2B). The absorption maxima at ~538 and 560 nm seem to be narrower and higher than for hemoglobin solutions, and the marked splitting of the deoxygenated and oxygenated spectrum at 569 nm indicates a leftward shift of the isosbestic wavelength. This would be in accordance with observations of Grunewald and Lübbers (6), but more wavelengths are needed to examine the
detailed changes in the spectra. Even if Table 1 shows no good
regression data when deoxygenated, dehydrated hemoglobin is used as one
of the component spectra, it does not necessarily mean that this
component is not present in the frozen samples. All samples will
probably be dehydrated to some degree during the freezing process, but
this will also be the case for samples used for determination of the
basic low-temperature spectra in Fig.
2B.
Absolute values for extinction coefficients could not be determined on
frozen samples because of lack of knowledge about the hemoglobin
concentration, optical path length, and scattering term. Single spectra
from different sections cut at the same thickness setting showed a
vertical shift relative to each other. The same was the case for
different measurement spots from the same frozen section. The
parallelism of the spectra was still good. Even if we know that the
main contribution to large vertical shifts is light scattering, it is
not possible to distinguish between these shifts and small differences
due to different concentration/optical path length or saturation.
Different scattering terms will cause vertical displacements of the
spectra but will have no influence on the calculated saturation values.
On the contrary, higher concentration/optical path length of the
oxygenated basic spectrum relative to the deoxygenated spectrum will
give the oxygenated spectrum a lower component in the curve fit
calculation of the multicomponent analysis, resulting in saturation
values that are too low. The calibration curve will then be curved and
intersect the "true" calibration curve only at the end points. We
picked only the most extreme, yet parallel spectra at the two end
points to obtain spectra with saturations close to the extreme values.
Averaging the spectra was justified by their parallelism. After
correction for estimated optical path length and concentration, a
suitable scattering term was subtracted, such that the spectra fit the
same range as the room temperature spectra. This last operation is
justified by not having any influence on the calculated saturation
values. The correctness of the estimated concentration/optical path
length was shown by the straight calibration curve obtained (Fig. 3).
The division of the spectrum of deoxygenated, dehydrated hemoglobin by
a constant to fit the same range as the other room temperature spectra
was done under the assumption that there were no scattering elements in
the sample.
In contrast to previous works using a small number of wavelengths, we
treated all wavelengths as nonisosbestic to calculate the saturations.
This was done because of the above-mentioned problems with determining
the concentrations/optical path lengths and the scattering terms. Only
small vertical shifts of the spectra in Fig. 2 relative to each other
will lead to quite different intersection points. Knowledge about the
intersections is not required for application of the nonisosbestic
theory. When three nonisosbestic wavelengths and the three basic
spectra of oxyhemoglobin, deoxyhemoglobin, and a constant scattering
term are used, this method is equivalent to the isosbestic wavelength
method used by others (16, 21, 22), except the scattering term no
longer can be calculated independently of saturation. As long as the wavelengths are close to those used for the isosbestic theory, the
precision in the saturation calculations should be the same for these
two methods. The precision in our saturation determinations was found
by statistical analysis to be 5%, which is comparable to 3-4%
reported by Zhu and Weiss (31) using the isosbestic theory.
Sample handling.
All sample handling was performed at lowest possible temperatures to
minimize O2 diffusion. There is no
easy way to measure the O2
diffusivity in such complex matter as frozen tissue. Even if the
fractional water contents of tissues are high [75-76% as measured in rat muscle (2) and 77-86% as measured in melanoma xenografts in mice (24)] and the
O2 diffusion in solid crystalline ice is essentially zero (7), there will probably be some
O2 diffusion through small cracks
formed during the freezing and sectioning processes. Figure 3 shows
that the regression curve intersects close to the origin, with no
deviation from the straight line at low saturations, as was the case in
the work of Grunewald and Lübbers (6). This indicates that there
is no significant oxygenation of the samples during preparation,
mounting, sectioning, and measurements with the exposure times and
temperatures applied. Our sectioning and measurement temperatures are
close to those chosen by Grunewald and Lübbers but lower than
those used by Zhu and Weiss (31). To further test for possible
O2 diffusion during sectioning, we
performed separate experiments where blood drops were sectioned at
higher temperatures. At 40°C and with the same exposure
times, we still found regions of the blood samples that were
deoxygenated (data not shown), supporting the assumption that there is
no significant O2 diffusion during
our sectioning at 60°C.
Image analysis.
Although filters were exchanged on the illumination side, some
measurement series showed displacement between images of different wavelengths. All image series were checked for possible displacements by visual inspection and eventually corrected for it by image translation in steps of one pixel. Thus the corrected images should in
principle not be displaced by more than one-half pixel in the x- or the
y-direction from any other image in
the same series. Image displacements will result in measurements at
slightly different locations for the different wavelengths. This may
cause errors if the measurement area includes heterogeneous parts of
the blood vessel or nonperfused areas. By segmenting areas not
significantly smaller than an RBC but still well inside the vessel
walls, this type of measurement error was minimized.
The linear regression curve (Fig. 3) shows good agreement between the
cryomicrospectrophotometer and the CO-oximeter measurements, with only
a slight deviation from the ideal 1:1 regression curve. This deviation
can be attributed to the accuracy of the CO-oximeter measurements,
different treatments of the samples for the two types of measurements,
and the uncertainty of the cryomicrospectrophotometer measurements. The
frequency histograms demonstrate the applicability of the method for
discriminating between well-oxygenated normal tissue (Fig.
4A) and tumor tissue with low
HbO2 saturation (Fig. 4B). This is in agreement with
similar histograms reported in the literature for intracapillary
HbO2 saturations (4) and polarographically measured tissue PO2
(9). Vaupel et al. (30) showed that tissue
PO2 histograms calculated from
intracapillary HbO2 saturations
measured by cryospectrophotometry correspond well with tissue
PO2 histograms measured
polarographically. Because of the mapping of the intracapillary
HbO2 saturations by our imaging
technique, it surpasses polarographic
PO2 measurements with respect to
relating tumor oxygenation to histology at the microregional level.
In summary, the cryomicrospectrophotometer offers rapid scanning and
imaging of tissue sections for
HbO2 saturation determination in
RBCs in microvessel profiles. This is an important tool for measuring
O2 supply to tumor tissue. The
standard error in a single measurement is 5%. The technique gives new
possibilities for comparative measurements on parallel histological
sections, as well as for off-line analysis.
| |
ACKNOWLEDGEMENTS |
The skillful assistance of Jørn Iversen and his staff at the
instrument workshop, The Norwegian Radium Hospital, is gratefully acknowledged.
| |
FOOTNOTES |
Financial support was received from The Norwegian Cancer Society.
Address for reprint requests: K. Måseide, Dept. of Biophysics,
Institute for Cancer Research, The Norwegian Radium Hospital,
Montebello, 0310 Oslo, Norway.
Received 24 April 1997; accepted in final form 7 August 1997.
| |
REFERENCES |
1.
Anderson, N. M.,
and
P. Sekelj.
Reflection and transmission of light by thin films of nonhaemolyzed blood.
Phys. Med. Biol.
12:
185-192,
1967[Medline].
2.
Cameron, I. L.,
V. A. Ord,
and
G. D. Fullerton.
Characterization of proton NMR relaxation times in normal and pathological tissues by correlation with other tissue parameters.
Magn. Reson. Imaging
2:
97-106,
1984[Medline].
3.
Fenton, B. M.,
and
T. E. J. Gayeski.
Determination of microvascular oxyhemoglobin saturations using cryospectrophotometry.
Am. J. Physiol.
259 (Heart Circ. Physiol. 28):
H1912-H1920,
1990[Abstract/Free Full Text].
4.
Fenton, B. M.,
E. K. Rofstad,
F. L. Degner,
and
R. M. Sutherland.
Cryospectrophotometric determination of tumor intravascular oxyhemoglobin saturations: dependence on vascular geometry and tumor growth.
J. Natl. Cancer Inst.
80:
1612-1619,
1988[Abstract/Free Full Text].
5.
Gayseki, T. E. J.,
and
C. R. Honig.
Myoglobin saturation and calculated PO2 in single cells of resting gracilis muscles.
In: Oxygen Transport to Tissue, edited by I. A. Silver,
M. Erecinska,
and H. I. Bicher. New York: Plenum, 1978, vol. III, p. 77-84.
6.
Grunewald, W. A.,
and
D. W. Lübbers.
Die Bestimmung der intracapillären HbO2-Sättigung mit einer kryo-mikrofotometrischen Methode angewandt am Myokard des Kaninchens.
Pflügers Arch.
353:
255-273,
1975[Medline].
7.
Hemmingsen, E.
Permeation of gases through ice.
Tellus
11:
355-359,
1959.
8.
Horecker, B. L.
The absorption spectra of hemoglobin and its derivatives in the visible and near infra-red regions.
J. Biol. Chem.
148:
173-183,
1943[Free Full Text].
9.
Kallinowski, F.,
R. Zander,
M. Hoeckel,
and
P. Vaupel.
Tumor tissue oxygenation as evaluated by computerized-PO2-histography.
Int. J. Radiat. Oncol. Biol. Phys.
19:
953-961,
1990[Medline].
10.
Kobayashi, H.,
and
N. Takizawa.
Oxygen saturation and pH changes in cremaster microvessels of the rat.
Am. J. Physiol.
270 (Heart Circ. Physiol. 39):
H1453-H1461,
1996[Abstract/Free Full Text].
11.
Kramer, K.,
J. O. Elam,
G. A. Saxton,
and
W. N. Elam, Jr.
Influence of oxygen saturation, erythrocyte concentration and optical depth upon the red and near-infrared light transmittance of whole blood.
Am. J. Physiol.
165:
229-246,
1951.
12.
Lübbers, D. W.,
and
R. Wodick.
The examination of multicomponent systems in biological materials by means of a rapid scanning photometer.
Appl. Optics
8:
1055-1062,
1969.
13.
Lübbers, D. W.,
and
R. Wodick.
Schnelle Photometrie komplizierter biochemischer Mehrkomponentensysteme.
Z. Anal. Chem.
261:
271-280,
1972.
14.
Pittman, R. N.
In vivo photometric analysis of hemoglobin.
Ann. Biomed. Eng.
14:
119-137,
1986[Medline].
15.
Pittman, R. N.
Microvessel blood oxygen measurement techniques.
In: Microcirculatory Technology, edited by C. H. Baker,
and W. L. Nastuk. Orlando, FL: Academic, 1986, p. 367-389.
16.
Pittman, R. N.,
and
B. R. Duling.
A new method for the measurement of percent oxyhemoglobin.
J. Appl. Physiol.
38:
315-320,
1975[Abstract/Free Full Text].
17.
Pittman, R. N.,
and
B. R. Duling.
Measurement of percent oxyhemoglobin in the microvasculature.
J. Appl. Physiol.
38:
321-327,
1975[Abstract/Free Full Text].
18.
Pries, A. R.,
G. Kanzow,
and
P. Gaehtgens.
Microphotometric determination of hematocrit in small vessels.
Am. J. Physiol.
245 (Heart Circ. Physiol. 14):
H167-H177,
1983.
19.
Rofstad, E. K.
Orthotopic human melanoma xenograft model systems for studies of tumour angiogenesis, pathophysiology, treatment sensitivity and metastatic pattern.
Br. J. Cancer
70:
804-812,
1994[Medline].
20.
Rofstad, E. K.,
P. DeMuth,
B. M. Fenton,
and
R. M. Sutherland.
31P nuclear magnetic resonance spectroscopy studies of tumor energy metabolism and its relationship to intracapillary oxyhemoglobin saturation status and tumor hypoxia.
Cancer Res.
48:
5440-5446,
1988[Abstract/Free Full Text].
21.
Sinha, A. K.,
J. A. Neubauer,
J. A. Lipp,
and
H. R. Weiss.
Oxygen saturation determination in frozen blood.
Microvasc. Res.
10:
312-321,
1975[Medline].
22.
Sinha, A. K.,
J. A. Neubauer,
J. A. Lipp,
and
H. R. Weiss.
Blood O2 saturation determination in frozen tissue.
Microvasc. Res.
14:
133-144,
1977[Medline].
23.
Tateishi, N.,
N. Maeda,
and
T. Shiga.
A method for measuring the rate of oxygen release from single microvessels.
Circ. Res.
70:
812-819,
1992[Abstract/Free Full Text].
24.
Tufto, I.,
and
E. K. Rofstad.
Interstitial fluid pressure in human melanoma xenografts.
Acta Oncol.
34:
361-365,
1995[Medline].
25.
Twersky, V.
Multiple scattering of waves and optical phenomena.
J. Opt. Soc. Am.
52:
145-171,
1962.
26.
Twersky, V.
On propagation in random media of discrete scatterers.
Am. Math. Soc.
16:
84-116,
1964.
27.
Twersky, V.
Interface effects in multiple scattering by large, low-refracting, absorbing particles.
J. Opt. Soc. Am.
60:
908-914,
1970.
28.
Twersky, V.
Absorption and multiple scattering by biological suspensions.
J. Opt. Soc. Am.
60:
1084-1093,
1970.
29.
Van Assendelft, O. W.
Spectrophotometry of Haemoglobin Derivatives. Assen, The Netherlands: Thomas, 1970, p. 8-73.
30.
Vaupel, P.,
W. A. Grunewald,
R. Manz,
and
W. Sowa.
Intracapillary HbO2 saturation in tumor tissue of DS-carcinosarcoma during normoxia.
Adv. Exp. Med. Biol.
94:
367-375,
1978.
31.
Zhu, N.,
and
H. R. Weiss.
Oxy- and carboxyhemoglobin saturation determination in frozen small vessels.
Am. J. Physiol.
260 (Heart Circ. Physiol. 29):
H626-H631,
1991[Abstract/Free Full Text].
AJP Heart Circ Physiol 273(6):H2910-H2918
0363-6135/97 $5.00
Copyright © 1997 the American Physiological Society
Top
Abstract
Introduction
![D(&lgr;) = <LIM><OP>∑</OP><LL>&ngr; = 1</LL><UL><IT>N</IT></UL></LIM> [c<SUB>&ngr;</SUB>&phgr;<SUB>&ngr;</SUB>(&lgr;)] <IT>d</IT>](/content/vol273/issue6/fulltext/H2910/img001.gif)
![<LIM><OP>∑</OP><LL><IT>i</IT> = 1</LL><UL><IT>M</IT></UL></LIM> {[ <IT>f</IT> (&lgr;<SUB><IT>i</IT></SUB>) − D(&lgr;<SUB><IT>i</IT></SUB>)]&rgr;(&lgr;<SUB><IT>i</IT></SUB>)}<SUP>2</SUP> = min](/content/vol273/issue6/fulltext/H2910/img002.gif)
![<LIM><OP>∑</OP><LL><IT>i</IT> = 1</LL><UL><IT>M</IT></UL></LIM><FENCE><FENCE><IT>f</IT> (&lgr;<SUB><IT>i</IT></SUB>) − <LIM><OP>∑</OP><LL>&ngr; = 1</LL><UL><IT>N</IT></UL></LIM> [c<SUB>&ngr;</SUB>&phgr;<SUB>&ngr;</SUB>(&lgr;<SUB><IT>i</IT></SUB>)] <IT>d</IT></FENCE>&rgr;<SUP>2</SUP>(&lgr;<SUB><IT>i</IT></SUB>)</FENCE>&phgr;<SUB>&ngr;</SUB>(&lgr;<SUB><IT>i</IT></SUB>)<IT>d</IT> = 0 &ngr; = 1, … , <IT>N</IT>](/content/vol273/issue6/fulltext/H2910/img003.gif)
