| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
A/
measured by PET
Department of Anesthesia, Massachusetts General Hospital, and Harvard Medical School, Boston, Massachusetts 02114
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
ACKNOWLEDGEMENTS
FOOTNOTES
REFERENCES
ABSTRACT 
Treppo, Steven, Srboljub M. Mijailovich, and José G. Venegas. Contributions of pulmonary perfusion and ventilation to
heterogeneity in 
A/
measured by PET. J. Appl. Physiol. 82(4): 1163-1176, 1997. To estimate the contributions of the heterogeneity in regional
perfusion () and alveolar ventilation
(A) to that of ventilation-perfusion
ratio (A/), we have
refined positron emission tomography (PET) techniques to image local
distributions of and
A per unit of gas volume content
(s and sA,
respectively) and VA/ in
dogs. sA was assessed in two ways:
1) the washout of 13NN tracer after equilibration
by rebreathing (sAi), and
2) the ratio of an apneic image after a bolus intravenous
infusion of 13NN-saline solution to an image collected
during a steady-state intravenous infusion of the same solution
(sAp).
sAp was systematically higher than sAi in all
animals, and there was a high spatial correlation between
s and
sAp in both body positions
(mean correlation was 0.69 prone and 0.81 supine) suggesting that
ventilation to well-perfused units was higher than to those poorly
perfused. In the prone position, the spatial distributions of
s, sAp, and A/ were fairly
uniform with no significant gravitational gradients; however, in the
supine position, these variables were significantly more heterogeneous,
mostly because of significant gravitational gradients (15, 5.5, and
10%/cm, respectively) accounting for 73, 33, and 66% of the
corresponding coefficient of variation (CV)2 values. We
conclude that, in the prone position, gravitational forces in blood and
lung tissues are largely balanced out by dorsoventral differences in
lung structure. In the supine position, effects of gravity and
structure become additive, resulting in substantial gravitational
gradients in s and
sAp, with the higher
heterogeneity in
A/ caused by a
gravitational gradient in s, only partially compensated by that in sA.
positron emission tomography; body position; gas exchange; regional
ventilation-perfusion ratio; dog; pulmonary heterogeneity; functional
imaging
INTRODUCTION WILSON AND BECK (30) have recently outlined a
theoretical approach to assess the contributions of heterogeneities in
alveolar ventilation ( METHODS The experimental protocol for these animal experiments was approved by
the Massachusetts General Hospital Committee on Animal Care.
Experimental Setup
Animal Preparation
A) and regional
perfusion () to the heterogeneity of the
ventilation-perfusion ratio
(A/). Such an approach was illustrated with experimental data extrapolated from a compilation of reports by different investigators using different methodologies, leaving the quantitative validity of some of their conclusions in
question. A number of recent studies have measured and characterized the effect of body position on the spatial heterogeneity of
A (2, 15, 27) and (3,
11, 12), but the individual contributions of these variables to the
heterogeneity of A/ cannot be reliably assessed unless all variables are measured in the
same individual. We have supplemented the positron imaging technique to
measure A/ described by
Rhodes and co-workers (21, 22) with independent measurements of
and A per unit of gas
content (s and
sA, respectively) (17) using
13NN as the tracer gas. With the use of these techniques,
we have imaged the distributions of s,
sA, and
A/ and analyzed their
spatial correlation and heterogeneity, including the contribution of
gravitational gradients, in prone and supine dogs. The imaging data
were also analyzed to yield and characterize distributions of
A/ comparable to those
generated by the multiple inert-gas-elimination technique (MIGET). This
study provides the experimental data required by the method of Wilson
and Beck (30) to assess the contributions of heterogeneities in
A and to the
heterogeneity of A/.
Fig. 1.
Schematic representation of experimental apparatus. PET, positron
emission tomography; ACg, specific activity of labeled gas in rebreathing circuit; ACv, specific activity of systemic
blood; ACpa, specific activity of pulmonary artery blood;
ACi, specific activity of intravenous infusate.
[View Larger Version of this Image (13K GIF file)]
Imaging Scans
Image collection during mechanical ventilation was gated by using an electrical signal from the mechanical ventilator at the start of inspiration. The gating scheme consisted of a collection of two consecutive images of equal duration during the breathing cycle. Because inspiratory time was set at 30% of the total breathing period, the first image included all of inspiration and the initial part of exhalation, whereas the second image included the latter part of exhalation, in which expiratory flow was small and the lungs remained almost stationary at a lung volume close to resting functional residual capacity (FRC). Five animals were studied in the prone position and five in the supine position. A transverse cross section of the lungs intersecting the apex of the heart was selected for imaging. Such a position was determined after equilibrating the lungs with inhaled 13NN-labeled gas and advancing the lungs into the field of view of the camera until the highest count rate was recorded by the camera. This position corresponded to the slice of the lungs having the greatest cross-sectional area and provided consistency among animals. After positioning, the following scan sequences were performed: inhaled tracer, bolus infusion, constant infusion, blood pool scan, and transmission and uniformity scans. These scan sequences are described in detail below, and the average counts per voxel of the resulting PET images are given in Table 1.
|
|||||||||||||||||||||||||||||||||
Blood pool scan. To estimate the contribution of counts originating from the pulmonary arterial blood per voxel during the constant-infusion (CI) imaging sequence, an additional blood pool scan was conducted. This was done by labeling the red blood cells with a temporary inhalation of 11C-labeled CO until the steady-state activity of the blood reached an adequate level. Two sequential images of 20-min duration were then collected. Blood samples were obtained at the beginning and the end of the imaging sequence to assess their respective specific activity needed to normalize the image and to correct for ventilatory losses of the CO tracer over the imaging time. Transmission and uniformity scans. To correct for gamma ray energy attenuation caused by the supporting structures and body tissues of the animal, a tubular ring, concentric to the PET camera's field, was filled with 18F-labeled water, and a gated transmission scan was collected during breathing. At the end of this scan, the animal and supporting structures were removed from the camera's field, and a final uniformity field scan was conducted.
Data Analysis
Image processing. PET images were initially corrected for camera sensitivity and for tissue attenuation. Image reconstruction was then performed with a convolution back-projection algorithm by using a Hanning filter yielding an effective spatial resolution of 10 mm determined from the width at one-half height of a point source image. This degradation of resolution length (from 4.5 mm of the camera) was needed to attenuate random noise to levels <5% of the measured coefficient of variation (CV)2 in most images processed. Resulting images consisted of an interpolated matrix of 159 × 159 voxels of 0.2 × 0.2 cm × 5 mm, or 57% of the resolution length. These reconstructed images of local counts per voxel were then processed following the methodology described in detail in the accompanying paper (17) and briefly discussed below, to yield functional images with voxel values in physical units (Table 2).
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
A/
IMAGE.
A steady-state CI image was created by adding, on a voxel-by-voxel
basis, the last three images of the protocol sequence. An initial mask
was created by thresholding the CI image to exclude areas outside of
the lung field. A threshold of 30% was used initially and then refined
in increments until no areas outside the lung field were included in
the mask (mean threshold used was 33 ± 6%). A second mask was then
created by thresholding the blood pool (Vb) image to define the heart
and largest vessels. (This was also done in an iterative process where
mean threshold used was 55 ± 5%.) This second mask was subtracted
from the first one to exclude heart and vessels from further analysis.
The masked CI image was then corrected for the contribution of
pulmonary arterial blood activity with the algorithm described by
Rhodes and co-workers (21, 22), where the volume of radiolabeled
arterial blood is assumed to be 40% of the total pulmonary blood
volume assessed from the 11CO scan. Thus a bloodcorrection
image was subtracted voxel-by-voxel from the CI image. A temporary
image was formed from the ratio of uncorrected to blood-corrected CI
images, and the mask was further refined to exclude from the analysis
additional areas with very high correction values. These areas,
covering ~20 voxels out of 2,000 voxels on average, were typically
located in proximity to the heart and large blood vessels and
corresponded to overcorrected voxels by partial volume effects from the
Vb scan. These final masks were then applied to all
functional images analyzed. Cardiac output
(T) was calculated by using a mass
balance from the infusion flow rate and the 13NN specific
activities of the infusate and the pulmonary arterial blood.
ALVEOLAR GAS CONTENT.
An alveolar gas content per voxel image (VA) was obtained
by decay-correcting the voxel values of the equilibrated inhaled tracer
scan, normalized by the specific activity of a gas sample to create an
image in units of milliliters of gas content per cubic centimeter of
voxel.
.
Because of diffusion to neighboring regions and/or
readsorption of the tracer into capillary blood, the voxel tracer
concentration may have changed during the apneic period following the
bolus intravenous infusion of 13NN-labeled saline. An
estimation of, and correction for, these tracer kinetics effects was
conducted on a voxel-by-voxel basis by assessing the differences in
tracer content between the last two 30-s images of the sequence and
then extrapolating the activity level expected at the time of the
tracer's arrival to the alveoli (17). The tracer-kinetics-corrected
image was normalized by the ratio of total infused activity to
I to yield an image of local
in units of milliliter per minute of blood flow per cubic centimeter of
voxel volume. Finally, the image was divided in a
voxel-by-voxel manner by the VA image to yield an image of s (in units of s
1).
SA.
We derived images of regional sA by
using two independent methods. One method directly assessed the
kinetics of inhaled NN2 tracer during a washout maneuver
following an equilibration scan
(sAi), as described in the
accompanying paper (17). A second method indirectly assessed
ventilation (sAp) as the
ratio of the local concentration of the NN2 tracer, infused
during apnea (distribution of ), divided by the
local concentration of the tracer during CI of the tracer in saline
solution in steady-state breathing, after subtraction of activity from
the pulmonary arterial blood (distribution of
/sA) (17). The
resulting sAp image represented exclusively sA of
perfused units, since unperfused units would not receive tracer during
either of the two imaging protocols.
Assessment of spatial heterogeneity.
The spatial heterogeneity of the functional images was assessed from
the CV of the voxel data within the lung field defined as the SD
normalized by the mean value of the data
|
or VA), the contribution of noise
to the CV2 was calculated as the random noise
(g2) caused by the finite counts
from the image. Thus, the noise-corrected CV (CVcr)
was
|
) of the original PET image or
|
A/,
sA, or s), the CV
of the ratio image [CV2(x/y)] was
corrected by the sum of random-noise contributions to the original
images, yielding
|
and vs.
s; and s vs.
sAi and vs.
sAp.
In those cases where the pairs of functional images were originally
derived by using a common image, i.e., VA vs.
s, Rs was corrected to
eliminate the pseudocorrelation caused by imaging noise in the common
image VA.1
FRACTIONAL DISTRIBUTIONS.
Mean-normalized distribution histograms for VA,
, s,
sA, and
A/ and their
corresponding log-transformed versions were generated for each of these
functional images. Regional data were grouped by either fraction of
total voxels (lung fraction), VA,
A and/or
fractions. These distributions were then characterized by evaluating
the corresponding Pearson coefficient of skewness (Skx) and coefficient of kurtosis (
).
BIVARIATE DISTRIBUTIONS.
Mean-normalized bivariate-distribution histograms for log-transformed
mean-normalized s vs.
sA were also generated, in which
sA was calculated from either inhaled
or perfused tracer. These bivariate distributions were plotted as
three-dimensional surfaces with the z-axis corresponding to the
fraction of total voxels having the corresponding relative values of
sA and s. Distributions were then averaged within each group of animals.
Shunt Fraction
Because of the low solubility of nitrogen in blood and tissues, an index of overall lung venous admixture can be calculated from the fraction of tracer recirculating back into the lungs during the steady-state period of CI. Such a recirculation fraction (FR) can be estimated from the ratio between the peripheral CV and Cpa simultaneously measured during the steady-state part of the CI protocol
|
Statistical Analysis
Comparisons between supine and prone positions were made by using single-tailed Student's t-test for independent samples. Comparisons between average and CV values of sAi and
sAp were made using
multivariate analysis of variance with body position and method as
factors. Statistical significance was taken at P < 0.05 level.
RESULTS
Mean ± SE values for supine and prone positions of the average voxel
value within the imaged section for ,
s, VA,
sAp, sAi, and
A/ are presented in
Table 2. Of these parameters, only and
s were significantly greater in the supine compared with the prone position.
Measurement of Heterogeneity
The contributions from different factors to heterogeneity, such as noise or vertical gradients, are additive only when expressed in terms of CV2; thus the findings of this study are presented in Table 2 and Fig. 3 as CV2. Because the definition of CV (SD/mean) gives a more intuitive impression of the degree of heterogeneity, in the text we present the results in terms of the CV.
),
ventilation-perfusion ratio
(A/), and specific
alveolar ventilation measured from kinetics of perfused tracer
(sAp) images in 5 supine
(
) and 5 prone animals (
). Note that CV2 values of
supine dogs are higher that those in prone dogs and that in all animals
CV2 of A/
is lower than that of s.
A/.
Although there were substantial differences in spatial heterogeneity
among the different dogs (Fig. 3), the heterogeneity of
A/ was significantly
lower in the prone animals (CV = 0.14) compared with the supine
animals (CV = 0.34) (Table 2). This difference in heterogeneity was
partially accounted for by the presence of a systematic vertical
gradient in the supine position that accounted for 66% of the
CV2, whereby
A/ decreased by 10%/cm
distance in the direction of gravity. In contrast, the vertical
gradient in the prone position was not significantly different from
zero (0.63%/cm). After we removed the vertical gradient from the
images by linear regression, the differences in residual heterogeneity
between supine and prone positions were still statistically significant
but of a much lesser magnitude [CVr = 0.13 for
prone and 0.17 for supine positions (Table 2)].
.
T, measured from the specific
tracer activities of the pulmonary artery and the saline infusate
(ACpa and ACi) was 1.01 ± 0.180 l/min for prone and 1.27 ± 0.340 l/min for supine
animals. Mean values of average regional
were 0.031 ml · s1 · cm3 for
prone and 0.061 ml · s1 · cm3 for
supine dogs (Table 2). Regional distributions of
either normalized by voxel volume or by alveloar gas
volume s, were more heterogeneous in the supine than
in the prone position. CV and
CVs were 0.41 and 0.46 in supine and 0.25 and
0.18 in prone position, respectively (Table 2). The greater
heterogeneity of and s in the
supine position was due, in part, to consistent gravitational
gradients, whereby the respective variables increased by 9.5 and
15.1%/cm distance in the direction of gravity. These gradients
contributed to 38 and 73% of the total
CV 2 and
CVs2 , respectively. In contrast, there were no consistent vertical gradients
in the prone position.
When the regional distributions of s and
for each position were compared,
s was significantly less heterogeneous than in the prone position while the contrary was true in
the supine position (Table 2). Differences in the width of the
corresponding fractional distribution are consistent with this finding
(Fig. 4, B and C, right) by
showing that the average distribution of s in the
prone position is wider than the corresponding distribution of
, with the opposite happening in the supine position.
; B), and
(s; C) in prone (left) and supine (middle) positions. Height of surfaces over the x-y
plane represents relative value of corresponding variable. Plots on
right are fractional distributions of corresponding variables
averaged for all dogs studied. Overbars designate average values.
VA. As illustrated by a wider fractional distribution of VA (Fig. 4A, right), the spatial heterogeneity of local gas content was greater in the supine position compared with the prone position (CV was 0.28 for supine and 0.21 for prone; Table 2). The higher heterogeneity of the supine position was partially accounted for by a systematic vertical gradient, whereby gas content decreased by 5.1%/cm distance in the direction of gravity. This gradient contributed to 23.7% of the total CV2. In contrast, the prone position had smaller and not significant vertical gradients (
0.370%/cm), without significant contributions to the total
CV2 (1.2%).
sA.
sA was directly assessed with the
13NN tracer delivered by inhalation
(sAi), and, indirectly,
from the ratio of images with the 13NN delivered by
intravenous infusion
(sAp). Two-way analysis of
variance with repeated measures on the values of average
(
) for both
methods, including the fixed effect of body position, showed a
significant effect of method but not of body position. Student's
t-tests confirmed no statistical differences for
between supine (0.042 s1; Table 2) and prone (0.042 s1) positions and for
between supine (0.086 s1) and prone (0.070 s1) positions but demonstrated
to
be significantly greater than
for each position. There was, however, a significant correlation between the individual values of
and
,
and a linear fit between the two independent estimates of
sA had a slope of 1.66 and
R2 = 0.59 (Fig. 5).
plotted against corresponding
for each animal studied.
was consistently lower than
.
Local values of s
Ap on a
voxel-by-voxel basis were poorly correlated with, and consistently
higher than, those of sAi (Fig. 6). CV of
sAi of prone
(0.18 ± 0.09) was not significantly different from that of supine
position (0.22 ± 0.1), whereas the CV of
sAp of prone animals (0.16)
was significantly lower than that of supine dogs (0.25). Variations in
the width of the corresponding fractional distribution histograms of
sAp and
sAi illustrate these findings
(Fig. 7).
Ap vs.
sAi for a representative dog
in supine (A) and one in prone (B) position.
Ai
(A) and sAp
(B) in prone (left) and supine positions
(middle). Height of surfaces over the x-y plane
represents relative value of corresponding variable. Plots on
right are fractional distributions of corresponding variables averaged for all dogs studied.
A consistent difference between the spatial distributions of s
Ap and
sAi occurred in the supine
position where sAp presented a vertical gradient of 5.5%/cm length in the direction of gravity accounting for 33.2% of the total CV2, whereas
sAi had no significant
gradient (Table 2).
Spatial Correlations
was found to have a high and significant
spatial correlation with VA in the prone position
(RS = 0.74 ± 0.05) but none in the supine
position (RS = 0.23 ± 0.01) (Table
3). s was not spatially
correlated with sAi in either
position (RS = 0.036 ± 0.13 for prone, and
RS = 0.11 ± 0.23 for supine positions). In
contrast, s was highly correlated to
sAp in both prone and supine
positions (RS = 0.69 and 0.81, respectively).
This correlation is illustrated in the bivariate distributions (Fig. 8
and Fig. 9),
where most voxels contain combinations of
), where 
is average regional
s, and
log(sAp/)
that fall along a 45° projection (constant
A/).
|
|||||||||||||||||||||||
vs. log(s/)
averaged over all dogs studied in prone position. Height of surface
over the x-y plane in surface plot (bottom left)
represents average fraction of voxels containing respective combination
of relative sAp and
s. A contour plot (top right) illustrates
correlation between these variables (direction of constant
A/ isopleths is
45°).
vs. )
averaged over all dogs studied in supine position. Height of surface
over the x-y plane in surface plot (bottom left)
represents average fraction of voxels containing respective combination
of relative sAp and
s. A contour plot (top right) illustrates correlation between these variables (direction of constant
A/ isopleths is
45°).
Fractional Distributions
Mean-normalized fractional distributions of functional images originating from the ratio of two PET images (sAp, s,
and A/) were all
skewed to the right, as shown by the mean Skx significantly greater than zero at both body positions (Table 4). Logarithmic transformation of the data
results in unskewed fractional distributions with mean
Skx values not different from zero. The
log-transformed distributions were also mesokurtic, i.e., they had
coefficient of kurtosis that does not appreciably deviate from 3. This suggests that the regional distributions closely resemble the
shape of a log-normal distribution.
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
In contrast, fractional distributions of functional images originating
from single PET images (VA, , and
A) were not skewed (Skx not different from zero) and were mesokurtic
while logarithmic transformation of the voxel data made the
distributions significantly skewed to the left
(Skx < 0).
Shunt Fraction
Reflecting the higher degree of venous admixture expected from a less uniformA/
distribution, the FR of the infused tracer recirculating back into the lungs during the CI protocol was
significantly higher in the supine position
(FR = 0.018 ± 0.005) compared with the prone
position (FR = 0.007 ± 0.005).
DISCUSSION
The most significant findings of this study were as follows.
1) The regional distribution of
A/ was more
heterogeneous in the supine position compared with the prone position.
The higher heterogeneity supine was due to a significant vertical
gradient that contributed more than one-half of the total
CV2. 2) Differences in the heterogeneities between
supine and prone in both and
A contributed to the differences in
A/ heterogeneity. 3) In both body positions, there was a high spatial correlation between s and
sAp while there was no
correlation between s and
sAi. Regional
and VA were positively correlated in the prone position but not in the supine position.
Methodology
Most issues related to the PET imaging methodology have either been discussed by the original proponents of the CI technique (7, 21, 22) or have been discussed in detail in the accompanying paper (17). A major methodological departure from the original method to assessA/ was in the
estimation of VA, where we imaged the lungs after
equilibration with inhaled 13NN gas instead of deriving
VA from the transmission scan as described by Rhodes et al.
(22). This modification improved the quality and signal-to-noise ratio
of the A/ images
because it largely canceled out systematic imaging artifacts and
because transmission scans with an equal number of events have greater
inherent noise than emission scans. As pointed out by Brudin et al.
(7), this improvement in image quality is only realized when imaging
normal lungs, since delayed equilibration in areas of very low
ventilation introduces errors in VA with the inhaled
13NN technique. Given the small size of our animals, it was
crucial to use this modification to obtain the highest possible spatial resolution of our instrument, compatible with an appropriate
signal-to-noise ratio.
Aside from the contributions of random noise to image heterogeneity,
there are systematic distortions introduced by PET imaging. For
example, the filtering involved in the convolution-back-projection algorithm not only affects the spatial resolution of the camera but
also results in smearing of the lung edges that artifactually increases
the estimated heterogeneity of an image that includes them. Voxel size
of our images was 0.2 cm × 0.2 cm × 5 mm with the average
amount of 37 ml of lung studied for the 10 animals (from an average of
2,000 voxels/animal). Maximal resolution of the instrument was 4.5 mm × 4.5 mm × 5 mm. The resolution length, defined as the width at
half maximum of a point source, was increased to 1 cm after filtering.
Other effects, such as imperfect uniformity calibration and
transmission corrections, also contribute to an increase in the
CV2 of images derived from single PET scans. Fortunately,
these systematic distortions cancel out in images obtained from the
voxel-by-voxel ratio of two independently acquired PET images such as
in sA, s, and
A/ (25). The
contribution of systematic imaging artifacts to the CV2 of
single PET images such as VA or
was also estimated by imaging a uniform
lung-like phantom and reconstructing, masking, and thresholding the
resulting images in the same way as our images. The CV2
measured from those images was 0.015, which corresponds to 34 and 25%
of the CV2 in the prone position and to 18 and 12% in the
supine position, measured for VA and ,
respectively.
Heterogeneity of A/
A/ was found to be
higher in supine (CV = 0.34) than in prone animals (CV = 0.14). In
the supine position, 66.0% of the CV2 was attributed to a
significant vertical gradient (10.3%/cm). The prone position, in
contrast, had no systematic vertical gradients in
A/, although there was
substantial interanimal variability with gradients ranging from 3.3
to +2.0 %/cm. The residual heterogeneity of the supine position, after
subtraction of the vertical effect, was still significantly greater
than that of the prone position. This might be attributed to the
inadequacy of the linear-regression model in describing nonlinear
vertical gradients. The CI technique to assess the distribution of
A/ was originally
described by Rhodes et al. (21). Although the authors did not study the effect of body position, they reported a CV = 0.21 for supine, spontaneously breathing normal humans. More recently, the same group of
investigators (5) reported a very modest vertical gradient in
A/, whereby
A/ decreased, on
average, by 2%/cm in the ventral-to-dorsal direction, with this
gradient explaining only 20% of the CV2. Remarkably, the
single patient studied in the prone position showed a more substantial
gradient in the direction of gravity than the group of supine patients.
The lower CV2 and gradients in humans compared with our
dogs could be due to a difference in distribution of ventilation
between spontaneously breathing and mechanically ventilated subjects
(20). Also, partial volume effects, exaggerated by a lack of
respiratory gating and poorer spatial resolution of their PET
instrument (1.7 cm) compared with ours (1 cm), must have also
accounted for lower CV in comparison with our study.
Using MIGET, Beck and co-workers (3) reported values of lnSD for the
main A/ distribution
peak of 0.45 and 0.35 for the supine and prone positions, respectively.
For narrow distributions, the lnSD approximates the CV (30), and
the results from MIGET appear somewhat greater than our direct
measurement of CV for A/. Intrinsic
differences between PET and MIGET need to be considered before
discussing these results. MIGET distributions are understood to reflect
the overall heterogeneity of
A/ of the whole lung
and at all physiologically relevant length scales. In normal
experimental conditions, however, the capability of MIGET to resolve
narrow distributions of
A/ is limited to distributions with SDlog 0.2 to 0.3 (19, 29). In contrast, the
heterogeneity in A/
measured by PET in this study is based on actual topographical
distribution but is limited to detect heterogeneities with length
scales greater than the spatial resolution of our PET imaging method (1 cm) and to sampling a single transverse cross section of the lung.
Thus part of the higher CV seen by MIGET, compared with PET, could possibly be attributed to the limited sampling and spatial resolution of our study. Although for the supine position the CV values measured by MIGET (0.45) and PET (0.34) do not appear to be too different, for the prone position the CV for MIGET (0.35) was more than double that measured by PET (0.14). Thus, at first glance, one could attribute the greater CV recovered from MIGET to a limitation of the technique to resolve narrow distributions
![ln[(<A><AC>V</AC><AC>˙</AC></A><SC>a</SC>/<A><AC>Q</AC><AC>˙</AC></A>)/(<OVL><A><AC>V</AC><AC>˙</AC></A><SC>a</SC></OVL>/<OVL><A><AC>Q</AC><AC>˙</AC></A></OVL>)]](/content/vol82/issue4/fulltext/1163/img028.gif)
