Journal of Applied Physiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

J Appl Physiol 101: 1451-1465, 2006. First published July 6, 2006; doi:10.1152/japplphysiol.01131.2005
8750-7587/06 $8.00
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow All Versions of this Article:
101/5/1451    most recent
01131.2005v1
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chon, D.
Right arrow Articles by Hoffman, E. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chon, D.
Right arrow Articles by Hoffman, E. A.

Regional pulmonary blood flow in dogs by 4D-X-ray CT

Deokiee Chon,1,2 Kenneth C. Beck,1 Ranae L. Larsen,3 Hidenori Shikata,1 and Eric A. Hoffman1,2

1Departments of Radiology and 2Biomedical Engineering, University of Iowa, Iowa City, Iowa; and 3Departments of Medicine, Division of Cardiology, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania


   ABSTRACT
TOP
 ABSTRACT
METHODS
 RESULTS
 DISCUSSION
 GRANTS
 DISCLOSURE
 ACKNOWLEDGMENTS
 REFERENCES
 
ECG-triggered computed tomography (CT) was used during passage of iodinated contrast to determine regional pulmonary blood flow (PBF) in anesthetized prone/supine dogs. PBF was evaluated as a function of height within the lung (supine and prone) as a function of various normalization methods: raw unit volume data (PBFraw) or PBF normalized to regional fraction air (PBFair), fractional non-air (PBFgm), or relative number of alveoli (PBFalv). The coefficient of variation of PBFraw, PBFair, PBFalv, and PBFgm ranged between 30 and 50% in both lungs and both body postures. The position of maximal flow along the height of the lung (MFP) was calculated for PBFraw, PBFair, PBFalv, and PBFgm. Only PBFgm showed a significantly different MFP height supine vs. prone (whole lung: 2.60 ± 1.08 cm supine vs. 5.08 ± 1.61 cm prone, P < 0.01). Mean slopes (ml/min/gm water content/cm) of PBFgm were steeper supine vs. prone in the right (RL) but not left lung (LL) (RL: –0.65 ± 0.29 supine vs. –0.26 ± 0.25 prone, P < 0.02; LL: –0.47 ± 0.21 supine vs. –0.32 ± 0.26 prone, P > 0.10). Mean slopes of PBFgm vs. vertical lung height were not different prone vs. supine above this vertical height of MFP (VMFP), but PBFgm slopes were steeper in the supine position below the VMFP in the RL. We conclude that PBFgm distribution was posture dependent in RL but not LL. Support of the heart may play a role. We demonstrate that normalization factors can lead to differing attributions of gravitational effects on PBF heterogeneity.

computed tomography; regional perfusion; physiological imaging; four-dimensional computed tomography; multidetector row computed tomography

REGIONAL PULMONARY BLOOD FLOW (PBF) has been evaluated by a variety of methods and imaging modalities. Distributions of PBF quantified by lodging of microspheres into the microvascular bed (5, 12, 13, 16) have been the classic method, but this requires lung excision and post mortem processing. Methods employing external projection imaging (44) have been unable to provide detailed spatial distributions of pulmonary blood flow because of inherent problems with superimposition and foreshortening. Nuclear tomographic techniques such as single photon emission computed tomography (SPECT) (17, 18, 30) and positron emission tomography (PET; Refs. 33, 36, 40), although very powerful techniques with their own strengths, have limited ability to correlate function with detailed lung structure. Furthermore, the resolution of these modalities is degraded if scanning is done during respiration, and both SPECT and PET scans are usually ungated and must span many respiratory and cardiac cycles to allow for the needed count statistics. Recently, magnetic resonance methods have been employed to image regional pulmonary perfusion (31, 38, 42). However, quantitation is complicated by the difficulty in imaging lung parenchyma and thus in the identification of an appropriate normalization factor to express perfusion in traditional metrics of milliliter per minute per gram. Recent clinical computed tomography (CT) methods have been presented that provide images of the pattern of lung blood pool enhancement to identify gross defects such as emboli (19). However, these methods are not quantitative and provide snapshots in time of contrast enhancement of pulmonary blood volume rather than metrics of perfusion that require time series imaging to measure first-pass blood flow kinetics. Inherent to all of these methods is the need to identify an appropriate normalization factor, expressing regional pulmonary blood flow as a function of a regional characteristic of the lung (tissue content, air content, alveoli, etc). Because of the exquisite anatomic detail offered by CT, such normalization factors can be interchangeably evaluated.

Dynamic X-ray CT obtained in conjunction with injection of contrast material has been used for regional tissue blood flow in the brain (3), heart (7, 50), kidney (6, 27), and lungs (28, 48, 51). These dynamic X-ray CT perfusion studies of the lungs (28, 48, 51) have been restricted to measurement of large regions of interest (ROI); i.e., dividing lung slices into thirds: anterior, middle, and posterior), and they reported flow per unit total volume of the ROI. This normalization does not match with most microsphere studies, which typically report values per gram of lung dry weight. Despite the difficulties of identifying appropriate normalization factors, a number of studies have used microspheres (50) and implanted perivascular electromagnetic flow probes (6) to validate the basic mathematical principles used in this study to evaluate regional tissue perfusion via CT scanning of the heart (50, 51), kidney (6, 26, 27), and lungs (28, 37, 49, 51).

In the current study we used an early version high-speed multidetector-row computed tomography (MDCT) with sharp bolus contrast injection into the right ventricle to quantify regional pulmonary blood flow with high spatial resolution, and we report data in small lung ROIs in terms of blood flow per unit total volume, per unit of regional air volume, per unit relative number of alveoli, and per unit of regional tissue volume to explore the effects of normalization on reported gradients and heterogeneity of flow. The techniques presented here are machine independent and have been implemented on modern MDCT scanners (22). In the analysis reported here, we provide detailed evaluations of regional pulmonary perfusion throughout the lung, providing a link between structure and function. As MDCT rapidly expands the number of slices obtainable in a single axial scan, these studies are best characterized as 4DCT (3 spatial dimensions plus time). It is the temporal dimension that provides perfusion from first pass kinetics and the expanded z-axis coverage that provides the spatial coverage of the perfusion measures.


   METHODS
TOP
 ABSTRACT
 METHODS
RESULTS
 DISCUSSION
 GRANTS
 DISCLOSURE
 ACKNOWLEDGMENTS
 REFERENCES
 
Dynamic and volumetric X-ray CT imaging.   We used an early version MDCT scanner known as an electron beam computed tomography (EBCT) scanner (Imatron), which has high temporal and spatial resolutions. To obtain a time series of images after injection of a bolus of radiopaque contrast agent, the scanner is aimed at a fixed section of lung (see below) and image acquisition is triggered from the peak of the QRS complex from the ECG signal. At each trigger, four pairs of 8-mm-thick contiguous slices are obtained with a 4-mm-gap between each slice pair. Each pair of images is obtained in the sweep time of 50 ms with an 8-ms intersweep delay. Thus the eight slices covering 7.6 cm of the body are achieved in 224 ms. A maximum of 80 slices can be acquired before scanning must be halted to transfer the data out of the scanner memory, allowing 10 time points to be acquired to define passage of a dye bolus through the lungs. At the beginning of each study, a higher spatial resolution volumetric scanning mode was used to obtain 40 parallel 3-mm-thick contiguous slices during a breath hold of 52 s. This image was used to locate the region used for the blood flow studies.

For a pulmonary blood flow study, contrast agent was injected after one trigger to allow for an eight-slice set for precontrast baseline images. The scan sequence was then adjusted such that image acquisition occurred at each heart beat for the first four or five heartbeats (to capture the peak) and then at every second or third heartbeat thereafter to define the curve tail. Depending on the animal's heart rate, a pulmonary blood flow scanning sequence was accomplished within a 20-s or less breath hold.

With a 15-cm field of view and matrix of 256 by 256 picture elements (pixels), each pixel of the reconstructed volume and flow images had a dimension of 0.59 mm on a side. The 3-mm-thick slices of the high-resolution volume images thus had a volume of 0.001 ml and the 8-mm-thick time series images had a volume of 0.003 ml/voxel. To reduce noise in the time series images, square regions of 5 x 5 voxels were analyzed encompassing 0.070 ml. Each voxel in the image data set has an associated gray scale attenuation proportional to the local characteristic tissue X-ray attenuation coefficient [Hounsfield units (HU)], with HU calibrated such that air equals –1,024, water equals 0, and bone and other solid tissue elements are >0.

Animal preparation.   All animal studies were performed within guidelines for animal care of the American Physiological Society, and the animal use protocol was approved by the University of Pennsylvania Institutional Animal Care and Use Committee (where scanning was performed). These animals are the same as used in the data analysis presented in Ref. 49. Seven dogs (12–25 kg, 4 beagles and 3 mongrels) were anesthetized with either fentanyl (0.05 mg/kg) or pentobarbital sodium (25 mg/kg), intubated, and mechanically ventilated. Catheters were placed in the right ventricular outflow tract (inserted via jugular vein) for contrast agent delivery and in the descending aorta (inserted via either carotid or femoral artery) and pulmonary trunk for pressure monitoring. The animal's ECG was monitored via extremity leads. A urinary catheter was also placed and intravenous saline was given at the rate of urine flow. Anesthesia was maintained throughout the entire experiment with supplemental pentobarbital or fentanyl every hour. At the conclusion of scanning, animals were overdosed with KCl and death was verified via ECG and pressure tracings.

Scanning and experimental protocol.   Each animal was studied in both the supine and prone positions on a table within the scanner gantry. The ECG signal was fed to a scanner using a computerized ECG monitoring system. Respiration was maintained by a computer-programmable mechanical ventilator (CT9000, CWE, Ardmore, PA). During all scanning procedures, left and right ventricular pressures, ECG, and airway pressure were recorded using data-acquisition software running on a Macintosh computer. The ventilator was preprogrammed to give five breaths to 25 cmH2O airway pressure just before scanning and to then hold airway pressure at 0 cmH2O (FRC) during scanning. A localization or scout scan was obtained without contrast injection to determine the level of the lung apices and base. A 40-level high-resolution scan (3-mm-thick contiguous slices) was obtained by triggering the scanner from the R-wave of the ECG signal during a single breath hold for an evaluation of pulmonary anatomic detail and determination of the level of the carina and main pulmonary artery.

For the pulmonary blood flow studies, after five hyperinflation maneuvers and during a breath hold at 0 cmH2O as above, an 8-level, 10-time-point flow study was carried out with the most cranial position at the level of the carina. Just before the second heartbeat after initiation of scanning, a radiopaque contrast agent [either Iohexol (nonionic) or Omnipaque (ionic)] was administered by a bolus injection of 1 ml/kg over 2 s. With injection of a contrast agent, the reconstructed attenuation coefficients change in proportion to the local concentration (mass) of the contrast, and care was taken to keep the blood concentration of contrast agent low enough to avoid voxel clipping at +1,024 HU (upper limit of the Imatron scanner dynamic range).

After each blood flow scanning protocol, images were reconstructed and contrast dilution curves were measured in the main pulmonary arteries and in the lung parenchyma to assure proper timing of the injection relative to scanning (at least a single time point existed before contrast arrival (injection) for use as a baseline, and the curve peak and tail were clearly delineated). After a repeat flow scan, the animal was repositioned in the opposite body posture (prone or supine) and the full scanning protocol was repeated. Animals were randomly chosen to be scanned in the supine or prone position first. Image data were stored on nine-track magnetic tape for later analysis.

Data analysis.   Both the reconstructed flow study images and the high-resolution, thin-slice volumetric image data sets were analyzed using custom computer analysis software. Regional blood flows were determined by measuring temporal changes in HU due to the influx and efflux of radiopaque contrast material sampled in ROIs placed over the lung field of the reconstructed temporal slice sequences and comparing this with a similar measurements made within an ROI placed over the ipsilateral central branch of the pulmonary artery. Blood flow in a region of lung parenchyma is obtained by quantifying the accumulation of contrast agent in the region (47, 48). The fundamental mass balance relationship for this technique is given by

Formula 1(1)
where Fi and Fo are input and output flow to the tissue; Ci and Co are incoming and outgoing concentration of indicator; and Ap is accumulated amount of indicator (e.g., mg) in parenchymal tissue. Assuming that Fi = Fo and that the bolus injection is sharp enough such that the amount of contrast leaving a sample is minimal before the full bolus arrival into the sample [Co(t) = 0], then Eq. 1 can be simplified and integrated to give

Formula 2(2)
Using subscript p to represent pulmonary parenchyma and V representing the volume of parenchyma under consideration, then by definition

Formula 3(3)
Substituting Eq. 3 into Eq. 2 and rearranging

Formula 4(4)
Injecting a contrast agent in conjunction with dynamic X-ray CT imaging causes the reconstructed X-ray voxel gray scale value to change in direct proportion to the local concentration of the contrast (Ap).

Formula 5(5)
where HUpa is the HU of contrast agent in the feeding pulmonary artery, HUpa,base is the HU measured before the arrival of contrast into the pulmonary artery, HUt,pk is the HU measured at the peak of the parenchymal dilution curve, and HUt,base is the HU measured before the arrival of contrast into the ROI. The ratio of the peak HU in an ROI placed over lung parenchyma to area of the arterial input curve thus gives flow per volume of the region, and raw blood flow (PBFraw, ml/min) was obtained from this ratio multiplied by volume of the voxel.

To further normalize blood flow, the fractions of non-air (tissue + blood) and air can be obtained from the images obtained before injection of contrast

Formula 6(6)

Formula 7(7)
where HUROI is the mean HU value in the parenchymal ROI under consideration. In this paper, we report values for PBF normalized per gram of non-air tissue (PBFgm) assuming the average tissue density is 1.05 g tissue/volume.

Formula 8(8)
Furthermore, blood flow is normalized by air fraction and relative number of alveoli.

Formula 9(9)

Formula 10(10)
Where Fmaxair is the maximum value of the parenchymal air fraction from all four slices. As alveoli decrease in size, there are larger numbers of alveoli within a region and the region has a smaller air fraction. This normalization attempts to duplicate studies that measure blood flow by microsphere methods using lung samples taken from lungs dried at total lung capacity where the lung is uniformly expanded. The fractional blood volume in a parenchymal ROI was calculated from the ratio of the area under the parenchymal time-intensity curve to the area under the arterial curve, where both areas were obtained by integrating the fitted curves (see below). Fractional regional blood volume was used as a criterion to accept ROIs for inclusion in the data set (Table 1). If there is too much blood in a voxel, this was considered to be contaminated by blood vessels, and if too little blood, this was considered to be contaminated by an airway.


View this table:
[in this window]
[in a new window]
  		Table 1. Inclusion criteria used to accept regions of pulmonary parenchyma from the image data sets

 
To perform the required curve integrations, the regional and arterial input time-intensity curves were fit to a gamma variate function {y = c·[t – t0]a·exp[–(t – t0)/b] + d} using the Levenberg-Marquardt least-squares minimization routine. The input data were obtained from the HU values in ROIs of the time series of images but excluding all data points after the time following the peak where HU – HUbase dropped to below 50% of the peak value. In each region, the {chi}2 value (a measure of goodness of fit), the peak value, time of the peak, mean transit time, and area under the curve were obtained from the gamma variate function.

Lung segmentation.   To avoid including non-parenchyma in the automated lung sampling routine, the outline of the lungs was drawn using a mouse-controlled video cursor to follow near the visually determined lung-chest wall, lung-cardiac, and lung-diaphragm borders. The outline was then smoothed with a three-point low-pass filter.

Criteria for sample inclusion.   The analysis program sampled the lung parenchyma in a regular lattice of square regions of 5 x 5 voxels within the lung boundaries outlined as above (sample volume = 0.070 cm3). Slice position represented the z coordinate of the samples, x represented the lateral to medial axis and y was the vertical (gravitation) direction. Criteria for accepting a given sample to include in the data set are given in Table 1. If the gamma variate curve-fitting routine (above) failed or gave high {chi}2 value because of a poorly defined baseline, lack of a distinct peak, or poorly defined tail, the ROI was excluded. When this occurred, it was most often in the most non-dependent lung regions and was likely due to extremely low blood flow to that region. Thus the lowest flow regions (<2 ml·min–1·ml parenchyma–1) are not represented in the data reported here. Samples with partial volume problems associated with blood vessels, larger airways, and regions that partially included mediastinal or chest wall borders and myocardium were eliminated by setting a range of acceptable percent air content from 40 to 90% (Table 1). This range was picked on the basis of our previous work, where at FRC normal lung density for a supine or prone animal remains between ~35 and 90% air contents (20, 21, 23, 24). We also eliminated all samples where the gamma variate curve fit to the dilution curve gave a mean transit time and/or arrival time that was greater or <2 SD beyond the mean of all the ROIs sampled. After the elimination process, depending on the quality of the data set, there were anywhere between 500 and 1,500 samples per right or left lung, which corresponds to between 60 and 75% of the total samples. This elimination process sought to exclude regions dominated by veins, arteries, or airways. We did not seek to eliminate parenchymal regions. Visual display of the included regions served to verify that, indeed, we were eliminating the targeted regions.


   RESULTS
TOP
 ABSTRACT
 METHODS
 RESULTS
DISCUSSION
 GRANTS
 DISCLOSURE
 ACKNOWLEDGMENTS
 REFERENCES
 
Figure 1 shows curves representing the transit of dye through the pulmonary artery and selected regions of parenchyma in one animal in the supine position. Note the difference in baseline (precontrast) radiodensity, indicating an increase in regional air content from dependent to nondependent regions. In parallel with the differences in baseline radiodensity, the height of the regional dye curves is lower in nondependent regions, indicating lower PBFgm.


Figure 1
View larger version (91K):
[in this window]
[in a new window]
  		Fig. 1. Example regional time-intensity curves [electron beam computed tomography (EBCT)]. Left, transverse computed tomography (CT) images of a dog in the supine position; right, regional time-intensity curves, Hounsfield units (HU). Top, region of interest (ROI) over a main pulmonary artery (PA) to obtain input reference indicator curve; bottom, curves from selected lung parenchymal ROIs.

 
Figure 2 shows a color-coded map of PBFgm over the lung fields from a supine animal. There is marked heterogeneity of PBFgm and a suggestion of an increase in PBFgm when comparing nondependent to dependent regions. Figure 3 shows a color-coded map of PBFgm in the prone position, indicating that there is a more uniform distribution. To demonstrate the effect of normalization by volume, relative number of alveoli, air content, and non-air tissue content of regions, data for PBFraw, PBFgm, PBFair, PBFalv, non-air fraction, air fraction, and relative number of alveoli were averaged in 3-mm intervals from dependent to nondependent regions of the lung as shown in Fig. 4. Note that there is little difference in the pattern of blood flow distribution using PBFraw, PBFair, PBFgm, and PBFalv in the prone position, because of the minimal gradient in tissue, air, and alveoli fraction. In contrast, there is a steeper gradient in PBFraw in the supine position, which is partly accentuated by normalization for air content, but compensated by normalization for relative number of alveoli. To determine the position along the vertical height of the lung at which maximal flow was observed (VMFP), we averaged data in 3-mm intervals from dependent to non-dependent regions and then picked the vertical position of the maximal mean four PBFs, relative to the lowest point in the image. The means of VMFP positions are shown in Table 2. In all cases, the peak occurred in a more dependent position supine compared with the prone body posture, but only the VMFP of the distribution of PBFgm for the whole lung was significantly different (P < 0.05). The mechanisms producing gradients above and below the peak could be different, being dominated by lung collapse or other issues related to low lung volumes in dependent regions and dominated by vascular distensibility or changing zonal conditions above the peak.


Figure 2
View larger version (108K):
[in this window]
[in a new window]
  		Fig. 2. Color coded map of fractional non-air pulmonary blood flow (PBFgm) in an animal in the supine position. Each panel shows one transverse slice of lung from apical (top left) to basal (bottom right) regions. Legend for the color scale, in milliliters per minute per gram non-air tissue is shown next to the bottom right panel.

 

Figure 3
View larger version (89K):
[in this window]
[in a new window]
  		Fig. 3. Color coded map of PBFgm in an animal in the prone position. Each panel shows one transverse slice of lung from apical (top left) to basal (bottom right) regions.

 

Figure 4
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 4. PBF raw unit volume data (PBFraw), PBF normalized to regional fraction air (PBFair), relative number of alveoli (PBFalv), PBFgm, mean fraction air content, mean fractional non-air content, and number of alveoli average in 3-mm intervals from one representative animal. Note the similar pattern of blood flow for PBFraw, PBFair, PBFalv, and PBFgm in the prone position, but a steeper relative decline in PBFraw and PBFair compared with PBFgm and PBFalv in the supine position. This is due to the steeper increase in fractional air content from dependent to nondependent regions in the supine position. Arrows indicate the positions of maximal flow along the height of the lung.

 

View this table:
[in this window]
[in a new window]
  		Table 2. Vertical maximal flow position of the distribution of PBFraw, PBFair, PBFalv, PBFgm

 
Figure 5 shows scatterplots of regional PBFgm from a representative animal in both prone and supine positions. Note again the marked heterogeneity of PBFgm in all graphs. In this example, there appears to be an overall gradient in PBFgm from dependent to non-dependent regions that is more marked in the supine position. Average vertical gradients determined by linear regression of scatterplots of PBFraw, PBFgm, PBFair, PBFalv for all animals are shown in Fig. 6. When regional blood flow is normalized by the four different parameters, there are significant differences in gradients between prone and supine positions for the whole lung, supine being more positive (steeper increase from non-dependent to dependent regions) than prone (P < 0.05; Refs. 20, 23). For left and right lungs, there were significant differences in gradients between supine and prone positions when normalizing to PBFraw, PBFair, and PBFalv. However, when normalized per unit non-air tissue weight (PBFgm), the gradients were significantly different for the right but not the left lung. In both supine and prone positions, the vertical gradients in PBFraw, PBFair, and PBFalv were not significantly different between the left and right lungs. However, the vertical gradient between left and right lungs when using PBFgm was significantly different in the supine position but not in the prone position.


Figure 5
View larger version (20K):
[in this window]
[in a new window]
  		Fig. 5. Distributions of PBFgm in one animal in the prone (right) and supine (left) positions. On each side, data for both lungs are shown at top, with data from individual lungs shown in bottom panels. Note the apparent steeper decline in PBFgm in the supine posture in this animal.

 

Figure 6
View larger version (22K):
[in this window]
[in a new window]
  		Fig. 6. Vertical gradient of PBFraw, PBFair, PBFalv, PBFgm in both supine and prone positions. Top, vertical gradient in whole lung; bottom, vertical gradient in the left and right lung. Note that vertical gradient was measured from the nondependent lung surface. NS, not significant.

 
Average vertical gradients in PBFs between supine and prone positions were obtained by processing the tissue data between the VMFP and non-dependent lung surface, shown in Fig. 7. In the whole lung, only PBFair was significantly different between supine and prone positions. In both supine and prone positions, vertical gradients of the four PBFs were not significantly different between left and right lung. For the right lung, only PBFgm was not significantly different between supine and prone positions. For the left lung, only PBFair was significantly different between supine and prone position.


Figure 7
View larger version (27K):
[in this window]
[in a new window]
  		Fig. 7. Vertical gradient of PBFraw, PBFair, PBFalv, PBFgm above the vertical maximal flow position in both supine and prone positions. Vertical gradient was obtained by processing the parenchymal region between nondependent lung surface and the vertical maximal flow position.

 
We assessed overall heterogeneity of PBFraw, PBFair, PBFalv, and PBFgm by calculating the total coefficient of variation of the four PBFs (Fig. 8). The overall coefficient of variation quantifies the total heterogeneity of blood flow, which amounted to 42, 52, 36, and 32% in the supine position and 33, 41, 29, and 28% in the prone position for PBFraw, PBFair, PBFalv, and PBFgm, respectively. The overall heterogeneity of PBFraw, PBFair, and PBFalv was statistically different between the two body postures for the whole lung data set, but not PBFgm (P = 0.05). For left and right lungs, the distributions of PBFraw, PBFair, and PBFalv were significantly more heterogeneous in the supine position than the prone position (P < 0.05), but the heterogeneity of PBFgm was significantly greater in the supine position only for the right lung alone. In the prone position, there was no significant difference between left and right lungs using the four normalizations of PBFs. In the supine position, the distributions of PBFraw, PBFair, and PBFalv were significantly more heterogeneous in the right lung than the left lung but not PBFgm. Mean r2 values were obtained from linear regression of PBFair, PBFraw, PBFalv, and PBFgm against vertical position (see Table 5). There were significantly different r2 values for only PBFgm between supine and prone positions (P < 0.05). In one animal, we performed repeat blood flow measurements to assess repeatability (Fig. 9). The correlation was good (r2 = 0.89 in prone, 0.74 in supine position), with a slope near 1.0, indicating reasonable reproducibility of PBFgm.


Figure 8
View larger version (24K):
[in this window]
[in a new window]
  		Fig. 8. Coefficient of variation (CV) of PBFraw, PBFair, PBFalv, PBFgm in both supine and prone positions. Note that CVs of PBFraw, PBFair, and PBFalv showed the statistically same pattern, but not PBFgm.

 

View this table:
[in this window]
[in a new window]
  		Table 5. r2 from linear regression of PBFraw, PBFair, PBFalv, and PBFgm against vertical position

 

Figure 9
View larger version (11K):
[in this window]
[in a new window]
  		Fig. 9. Repeatability of PBFgm. Data from one animal are shown. Repeat contrast injections were performed to assess the repeatability of the regional blood flow measurements. Left and right, data from the supine and prone positions, respectively.

 
Because CT has advanced considerably from the time the dogs in this paper were studied, we show in Fig. 10 data obtained from a more recent MDCT scanner. Regional time-intensity curves in ROIs from a pulmonary artery (upper) and parenchyma (lower) are shown from images derived using a Siemens Sensation 16 MDCT scanner. Note that attenuation (HU) from time-intensive curves increases as ROIs move from non-dependent to dependent lung areas, indicating vertical gradients of pulmonary blood flow. These slices are 1.2 mm thick and this ROI represents considerably finer sampling of regional perfusion than was possible using the 8-mm-thick EBCT image data. These data were acquired under an animal study protocol approved by the University of Iowa Animal Care and Use Committee.


Figure 10
View larger version (100K):
[in this window]
[in a new window]
  		Fig. 10. Example regional time-intensity curves (MDCT). Sample locations of the feeding pulmonary (Pul) artery (input, top left) and parenchymal (output, bottom left) ROIs on transverse plane in supine position and gamma variate-fitted input (top right) and output (bottom right) curves corresponding to the locations.

 

   DISCUSSION
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
GRANTS
 DISCLOSURE
 ACKNOWLEDGMENTS
 REFERENCES
 
We use high speed X-ray CT to assess regional pulmonary blood flow using 4DCT (multiple slices imaged over time) in intact animals, and this methodology can be translated to current MDCT scanners. Although others (19) have studied what they have called "pulmonary perfusion via CT" to assess the presence of pulmonary emboli, these studies are not actually assessing perfusion but rather perfused blood volume. To assess perfusion quantitatively as we have done here, one must follow first-pass kinetics by scanning a section of lung over time as a sharp bolus of contrast traverses the lung parenchyma. The studies referred to above use a spiral mode of imaging during a prolonged infusion of contrast to identify differences in regional enhancement during a pseudo steady-state contrast enhancement. This is not perfusion. In addition, others (28) have followed a bolus over time as we have but they did not normalize the raw perfusion values and thus it is difficult to compare these measures against other data in the literature or between individuals or between body postures. In part, because of the difference in the selection of normalization factors for assessing regional differences in pulmonary perfusion or difficulty in assessing a value for the chosen normalization factor, considerable controversy has arisen regarding interpretation of results (43). In this study, we take advantage of CT to assess regional pulmonary perfusion and report data in small lung regions using various normalization factors: PBFraw, PBFair, PBFalv, and PBFgm to explore the effects of these normalization factors on reported gradients and estimates of flow heterogeneity. CT allows for the in vivo assessment of these various normalization parameters accurately and simultaneously. The important findings of this study follow. 1) PBFraw, PBFair, and PBFalv data were consistent with a gravitational influence over PBF distribution. However, PBFgm did not exhibit a strong posture dependence. 2) There was a significant difference in vertical gradient in only PBFgm between prone and supine positions in the right lung and between the left and the right lungs in the supine position, suggesting that the regional blood flow to dependent regions of the left lung may be more affected by the weight of the heart than the right lung.

Methodological issues.   The theoretical and experimental bases of dye dilution measurements of blood flow are well developed (4, 9, 15), and recently indicator dilution theory has been verified and used in conjunction with dynamic X-ray CT scanning to quantify regional perfusion patterns (8, 9, 39, 49) in such organs as the heart (50, 51), kidney (6, 26, 27), brain (3), and lungs (28, 37, 49, 51). The very limited number of pulmonary perfusion studies using dynamic X-ray CT scanning has, however, been restricted to measuring pulmonary blood flow in large regions of the lung (i.e., anterior, middle, and posterior regions). We have developed techniques to sample dynamic CT image data sets with a much finer resolution (samples as small as 2 mm by 2 mm in this current study) and to simultaneously quantify several regional pulmonary functional parameters with high spatial and temporal resolution.

Wolfkiel and Rich (48) tested the validity of the blood flow algorithm in dogs by comparing time-density curves of the pulmonary outflow tract, lung, and aortic outflow tracts, which were calculated by determining the fraction of the area under an aortic root time-density curve at the time of peak tissue concentration. They reported that the estimated washouts of indicator were minimum (2.4 ± 5.0% in the supine position and 3.9 ± 6.8% in the prone position), demonstrating that tracer washout is valid for pulmonary perfusion measurements made with a sharp, central intravenous bolus injection of contrast material, the protocol we used in the studies reported here. Two studies have shown a good correlation between microsphere distribution in lungs (51) or heart (50) vs. CT-based assessment of first-pass kinetics of a bolus-injected iodinated contrast agent using the same mathematics as proposed here. In Wu et al. (51), the authors showed an excellent correlation between CT-based and microspheres-based methods for the measurement of pulmonary blood flow, albeit the study used fairly large regions of interest. In Wong et al. (50), the authors showed a very strong correlation between microspheres and the CT-based measures but pointed out the need for careful matching of normalization factors in such studies. When the heart was removed for assessment of injected microspheres, myocardial blood is drained and thus in vivo "myocardial volume" is different from ex vivo "myocardial volume." At higher flow rates, these differences became significant and thus had to be accounted for. In the case of pulmonary blood flow assessment, in vivo vs. ex vivo correlation problems are compounded by the fact that the lung collapses on removal from the chest cavity and traditional microsphere sampling requires prior air drying of the lung with associated distortions. In addition to validation of the iodinated contrast agent in heart and lung, Bently et al. (6) has shown that the method is robust for the assessment of renal perfusion as well. In this study, the authors compared CT-based perfusion against implanted perivascular electromagnetic flow probes.

Our normalization to determine PBFgm divided PBF by the regional non-air content that is easily obtained from precontrast baseline images. The non-air volume includes true tissue volume, blood volume, and interstitial or alveolar fluid volumes. Ideally we would normalize by true tissue volume, although this quantity is difficult to obtain with CT scanning. It is theoretically possible to obtain blood volume from the ratio of the area under the regional curve to area under the arterial input curve. With this quantity, one could subtract blood volume from non-air volume to obtain tissue volume. However, in practice, this frequently results in a negative or otherwise unreasonable tissue volume because of the inherent noise variability in both signals. For this study, we therefore chose the simpler normalization by non-air content. Instead, normalization by relative number of alveoli could be considered as normalization by unit lung tissue.

We controlled for temporal density changes due to motion artifacts by acquiring image data during breath hold at 0 cmH2O airway pressure (FRC) and gated scanning to the animal's ECG. Under these conditions, alveolar pressure should also be 0 cmH2O so that most, if not all, of the lung should remain in zone 3 flow conditions during the time the functional data were acquired. Furthermore, we used selection criteria to eliminate samples containing major blood vessels and/or major airways so that the data should accurately represent functional lung parenchymal regions.

Gravitational influences on PBF distribution.   Murphy et al. (35) used EBCT assessments of local pulmonary blood flow in humans and showed increased flow to dependent regions in supine subjects. This study used a peripheral vein injection of bolus contrast material, which produces a wider input bolus, making it more likely that the basic assumption of minimal venous washout at the time of peak in regional dye curve is not valid. Wolfkiel and Rich (48) compared dynamic X-ray CT regional pulmonary enhancement in supine and prone dogs. He analyzed relatively large areas of lung and reported a significant difference between anterior, middle, and posterior regions of lung in the supine position but only a slight decline in flow to posterior (nondependent) regions in prone position. Similarly, we noted a reduction of the vertical gradient in PBFraw from the supine to prone posture (Fig. 6). However, regional differences in PBFraw could in part reflect regional differences in amount of blood vessel-containing tissue between regions, or regional tissue content. To correct for this effect, we expressed flows per gram of non-air content, PBFgm (Tables 3 and 4 and Fig. 6). This analysis was performed using linear regression, which can be markedly affected if the underlying trend in PBFgm is not linear. In fact, most plots of PBFraw, PBFair, PBFalv, and PBFgm vs. vertical distance up the lung, especially in the prone position, showed an inverted U configuration, with a peak near the center of the lung and declines in flow above and below the peak. To further analyze this effect, we determined the position of the peak in the four different types of normalization of blood flow using flows average in 3-mm intervals up the lung. This position was located more toward the dependent lung in the supine compared with prone position in both right and left lungs (Table 2). Factors determining the position of this peak were not determined in this study. The reduction in flow in extreme dependent lung has been termed "zone 4" to emphasize that it is not entirely consistent with the hydrostatic theory of pulmonary flow distribution originally put forth by West and colleagues (10, 25, 45). Factors that might cause the reduction in flow in the extreme dependent regions include 1) an increase in vascular resistance at low lung volumes; 2) an increase in interstitial fluid pressure that compresses extra-alveolar vessels; and 3) the weight of the heart compressing lung. The role of each of these factors and their interaction with normalization factors in determining distribution on the small scale we report in this work needs to be investigated in future study.


View this table:
[in this window]
[in a new window]
  		Table 3. Gradients in PBFgm in nondependent lung

 

View this table:
[in this window]
[in a new window]
  		Table 4. Gradients in PBFgm in dependent lung

 
Vascular influences on gravity-independent pulmonary perfusion distribution have been reported by Beck and Rehder (5). They suggested that vascular conductance plays an important role in pulmonary perfusion distribution by showing that the dorsal-basal regions of the canine lung receive the highest blood flow in the supine and prone lung when the gravitational influences were effectively subtracted in isolated lungs. It may be that in the supine canine lung gravitational and pulmonary vascular influences are acting synergistically to drive blood flow to the dorsal lung regions, giving rise to the ventral-dorsal perfusion gradient we and others have observed. In the prone posture, however, gravity would be acting to drive blood flow to the ventral lung regions while vascular geometry, such as conductance, would be driving blood flow to the dorsal lung regions, so that the two influences would balance each other, resulting in a more homogenous perfusion distribution (46). The present data are consistent with this model, considering both the differences in vertical gradients (Fig. 6) and the changes in position of the maximal blood flow (Table 2) between prone and supine postures. However, there may be other influences at work in these intact lungs, as demonstrated by the data in Fig. 7. Consistent with the hydrostatic model of PBF distribution, this analysis shows that there is no prone-supine difference in the gradient in only PBFgm in regions above the VMFP. Below the VMFP, there is a significant difference in gradient in PBFgm between prone and supine position in the right lung only and between the left and the right lung in the supine position only. We suggest the regional flow to dependent regions of the left lung is affected by the weight of the heart more than in the right lung, explaining the difference. This is similar to previous findings of Hoffman and Ritman (23) who suggested that the intrathoracic position of the heart may be an important factor in the distribution of regional lung air content. The authors noted that, in the prone posture, the sternum supports the heart, whereas in the supine posture, the heart, in part, is supported by the lungs. Albert and Hubmayr (1) also reported the compressive force resulting from heart weight toward the lung by measuring the relative volume of lung located directly under the heart in the supine and prone postures. The authors found that in the supine position, the greater portion of the left lung was located under the heart than the right lung, but no lung was located under the heart while subjects were in the prone position. With the continued refinement of methods for determining regional lung function, such as what we report here, studies to sort out mechanisms for regional differences in blood flow, ventilation, and regional air content are needed.

Effect of normalization on vertical gradient of blood flow distribution.   The previous studies using different imaging modes have reported pulmonary blood distribution with different normalizations (2, 28, 34, 36, 48). In our study, we found that vertical gradient of blood flow distribution was affected by vertical distribution of normalization components such as unit volume, unit alveolar gas, unit relative number of alveoli (per unit lung tissue), unit non-air (tissue+blood) contents in both body postures, especially in the supine posture shown in Fig. 6. Treppo et al. (40), using PET, compared blood flow (PBFraw) normalized by unit thorax volume with blood flow (PBFair) normalized per unit gas content. In PET, normalization by air content is advantageous due to the compensation for the partial volume effects in voxels containing large vessels. The authors showed that the vertical gradient of air content in the supine position had additive effects on the vertical gradient in PBFair (15.1 ± 0.49%/cm) compared with PBFraw (9.55 ± 0.73%/cm), which is consistent with our results. Pulmonary blood flow normalized by the relative number of alveoli was used in previous studies (2, 29) because alveoli are uniformly expanded at TLC and alveolar number is proportional to regional volume (32). Our study showed that the vertical gradient of PBFalv was more uniform than the distribution of PBFraw. Anthonisen and Milic-Emili (2) using Xe scintigraphy, compared perfusion distributions at FRC normalized by per unit volume with per alveolus and reported that vertical gradients of blood flow per alveolus was smaller than vertical gradients of blood flow per unit volume due to the existence of zonal change from two to three in vertical distribution of blood flow per alveolus. Our findings demonstrated that vertical distributions of PBF based on four different normalization factors did not cause significant difference of vertical gradients of blood flow distribution between the left and the right lung in either the prone or supine body postures, with the exception of PBFgm assessed in the supine position.

Effect of normalization on heterogeneity of blood flow distribution.   Different normalizations led to the different degrees of heterogeneity (coefficients of variation) of blood flow distributions. Our results, and other studies using distributions of injected microspheres in the lung, suggest a strong nongravitational component to heterogeneity of PBF (13, 14, 41). The coefficients of variation (CV) of PBFraw, PBFair, PBFalv, and PBFgm in the gravitational direction we report (32–52% supine; 28–40% prone; Fig. 8) indicate considerable flow heterogeneity within isogravitational planes. Glenny et al. (13), using a microsphere technique, found a heterogeneity of PBF normalized by dry tissue (CV = 0.44 supine, CV = 0.39 prone) that was greater than our PBFgm (CV = 0.32 supine, CV = 0.28 prone). With the use of the fractal relationship presented by Glenny and colleagues (11, 14), perfusion heterogeneity should have been 1.5 times greater at the volume of the ROIs used in this study. Other contributing factors to lower heterogeneity seen in this study could be due to 1) the fact that we used selection criteria to assess regions that were relatively free of partial volume small blood vessels or airways; 2) our inclusion of blood along with tissue in the normalization used to calculate PBFgm; 3) our lung regions were limited to the 7.6 cm of z-axis coverage of the scanner; and 4) our data represent blood flow at FRC whereas microspheres represent perfusion occurring during active respiration. It is well recognized that at higher lung volumes, perfusion is decreased to the non-dependent lung regions, increasing overall heterogeneity. Treppo et al. (40) using PET reported the value of the spatial heterogeneity of blood flow normalized by volume and air content ([PBFraw] CV = 0.41 supine, CV = 0.25 prone; [PBFair] CV = 0.47 supine, CV = 0.18 prone). Their results in the supine position were close to our results (supine: CV = 0.43 PBFraw, CV = 0.51 PBFair), but in the prone position our results (prone: CV = 0.33 PBFraw, CV = 0.41 PBFair) were actually greater than their findings. The CV of flow has been shown to depend on the sample size, and our flow measurements were from 0.070 cm3 volume samples compared with 1.9 cm3 [Glenny and Robertson (13)] and 0.5 cm3 [Treppo et al. (40)] samples (41). We would expect, therefore, that the CV would be higher in our studies due to the finer sampling possible with our method. This difference in degree of heterogeneity could be related to species differences but remains unexplained. It may require a study to directly compare the two methods of determining heterogeneity to settle the issue. r2 values shown in Table 5 indicate the degree to which gravity (vertical gradient) contributes to the total heterogeneity in distribution of blood flow with four different types of normalizations. Gravitational contributions (r2 value; Table 5) to PBF under the various normalization conditions had the same trend as shown for total heterogeneity of PBF distribution based upon CV (Fig. 8). With the use of either r2 or CV, only PBFgm showed a significant difference between the supine and the prone positions.

In summary, we demonstrated the utility of measuring regional pulmonary blood flow in intact lungs of dogs using bolus injection of radiopaque contrast agent. Also, we showed that the characteristics of four different normalization factors such as unit lung volume, regional air content, non-air content, and relative number of alveoli can have an impact on the vertical gradient and heterogeneity of distribution of pulmonary blood flow in both body postures, and these normalization factors along with the ability to accurately estimate their values may play a role in much of the controversy over regional pulmonary blood flow heterogeneity. Pulmonary perfusion studies, especially using imaging modes, should consider appropriate normalization factors when analyzing the result for cross comparison among studies. In the present investigation, PBFraw, PBFair, and PBFalv showed posture dependence in both the left and the right lung. In contrast, we found a significant difference in perfusion (PBFgm) gradients from dependent to non-dependent regions between prone and supine body orientations in the right lung, but our data from left lung show no consistent postural variation (Tables 3 and 4). We suggest the heart may play a role in explaining these differences. Our data also show marked heterogeneity in regional perfusion.


   GRANTS
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
DISCLOSURE
 ACKNOWLEDGMENTS
 REFERENCES
 
This study was supported in part by National Heart, Lung, and Blood Institute Bioengineering Research Partnership Grant HL-064368.


   DISCLOSURE
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 DISCLOSURE
ACKNOWLEDGMENTS
 REFERENCES
 
E. A. Hoffman is a share holder in VIDA Diagnostics, which is the commercial software used in this study.


   ACKNOWLEDGMENTS
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 DISCLOSURE
 ACKNOWLEDGMENTS
REFERENCES
 
Current address for R. Larsen: Loma Linda Medical Center, Room 4433, 11234 Anderson St., Loma Linda, CA 92354.


   FOOTNOTES
 

Address for reprint requests and other correspondence: E. A. Hoffman, Dept. of Radiology, Univ. of Iowa College of Medicine, 200 Hawkins Dr., Iowa City, IA 52242 (e-mail: eric-hoffman{at}uiowa.edu)

The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


   REFERENCES
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 DISCLOSURE
 ACKNOWLEDGMENTS
 REFERENCES
 

  1. Albert R and Hubmayr R. The prone position eliminates compression of the lungs by the heart. Am J Respir Crit Care Med 161: 1660–1665, 2000.[Abstract/Free Full Text]
  2. Anthonisen N and Milic-Emili J. Distribution of pulmonary perfusion in erect man. J Appl Physiol 21: 760–766, 1966.[Free Full Text]
  3. Axel L. Cerebral blood flow determination by rapid-sequence computed tomography: a theoretical analysis. Radiology 137: 679–686, 1980.[Abstract/Free Full Text]
  4. Bassingthwaighte JB, Raymond GR, and Chan JIS. Nuclear Cardiology: State of the Art and Future Directions. St. Louis, MO: Mosby Year Book, 1993.
  5. Beck KC and Rehder K. Differences in regional vascular conductances in isolated dog lungs. J Appl Physiol 61: 530–538, 1986.[Abstract/Free Full Text]
  6. Bentley MD, Lerman LO, Hoffman EA, Fiksen-Olsen MJ, Ritman EL, and Romero JC. Measurement of renal perfusion and blood flow with fast computed tomography. Circ Res 74: 945–951, 1994.[Abstract/Free Full Text]
  7. Brundin LH, Rhodes CG, Valind SO, Jones T, and Hughes JMB. Interrelationships between regional blood flow, blood volume, and ventilation in supine humans. J Appl Physiol 76: 1205–1210, 1994.[Abstract/Free Full Text]
  8. Clough AV, Haworth ST, Roerig DL, Linehan JH, and Dawson CA. Microfocal angiography of the pulmonary vasculature. In: Medical Imaging 1998: Physiology and Function From Multidimensional Images. San Diego: SPIE, 1998.
  9. Clough AV, Linehan JH, and Dawson CA. Regional perfusion parameters from pulmonary microfocal angiograms. Am J Physiol Heart Circ Physiol 272: H1537–H1548, 1997.[Abstract/Free Full Text]
  10. Glazier JB, Hughes JM, Maloney JE, and West JB. Role of interstitial pressure in the distribution of pulmonary blood flow. J Physiol 190: 23P–24P, 1967.[Medline]
  11. Glenny R, Bernard S, and Robertson H. Pulmonary blood flow remains fractal down to the level of gas exchange. J Appl Physiol 89: 742–748, 2000.[Abstract/Free Full Text]
  12. Glenny RW, Bernard S, and Brinkley M. Validation of fluorescent-labeled microspheres for measurement of regional organ perfusion. J Appl Physiol 74: 2585–2597, 1993.[Abstract/Free Full Text]
  13. Glenny RW, Lamm WJE, Albert RK, and Robertson HT. Gravity is a minor determinant of pulmonary blood flow distribution. J Appl Physiol 71: 620–629, 1991.[Abstract/Free Full Text]
  14. Glenny RW and Robertson HT. Fractal properties of pulmonary blood flow: characterization of spatial heterogeneity. J Appl Physiol 69: 532–545, 1990.[Abstract/Free Full Text]
  15. Gould RG. Perfusion quantitation by ultrafast computed tomography. Invest Radiol 27: S18–S21, 1992.
  16. Greenleaf JF, Ritman JFL, Sass DJ, and Wood EH. Spatial distribution of pulmonary blood flow in dogs in left decubitus position. Am J Physiol 227: 230–244, 1974.[Free Full Text]
  17. Hakim TS, Dean GW, and Lisbona R. Effect of body posture on spatial distribution of pulmonary blood flow. J Appl Physiol 64: 1160–1170, 1988.[Abstract/Free Full Text]
  18. Hakim TS, Lisbona R, and Dean GW. Effect of cardiac output on gravity-dependent and nondependent inequality in pulmonary blood flow. J Appl Physiol 66: 1570–1578, 1989.[Abstract/Free Full Text]
  19. Herzog P, Wildberger JE, Niethammer M, Schaller S, and Schoepf UJ. CT perfusion imaging of the lung in pulmonary embolism. Acad Radiol 10: 1132–1146, 2003.[CrossRef][ISI][Medline]
  20. Hoffman EA. Effect of body orientation on regional lung expansion: a computed tomographic approach. J Appl Physiol 59: 468–480, 1985.[Abstract/Free Full Text]
  21. Hoffman EA, Acharya RS, and Wollins JA. Computer aided analysis of regional lung air content using three-dimensional computed tomographic images and multinomial models. Int J Math Model 7: 1099–1116, 1986.[CrossRef]
  22. Hoffman EA, Reinhardt JM, Sonka M, Simon BA, Guo J, Saba O, Chon D, Samrah S, Shikata H, Tschirren J, Palagyi K, Beck KC, McLennan G, Wang T, Schultz G, Hebestreit H, Hahn D, and Jakob PM. Characterization of the interstitial lung disease via density-based and texture-based analysis of computed tomography images of lung structure and function. Acad Radiol 10: 1104–1118, 2003.[CrossRef][ISI][Medline]
  23. Hoffman EA and Ritman EL. Effect of body orientation on regional lung expansion in dog and sloth. J Appl Physiol 59: 481–491, 1985.[Abstract/Free Full Text]
  24. Hoffman EA, Sinak LJ, and Ritman EL. Effect of body position on regional lung expansion: a computer tomographic approach. Physiologist 26: A-69, 1983.
  25. Hughes JMB, Glazier JB, Maloney JE, and West JB. Effect of lung volume on the distribution of pulmonary blood flow in man. Respir Physiol 4: 58–72, 1968.[CrossRef][ISI][Medline]
  26. Iwasaki T, Ritman EL, Fiksen-Olsen ML, Romero JC, and Knox FG. Renal cortical perfusion—preliminary experience with the dynamic spatial reconstructor (DSR). Ann Biomed Eng 13: 259–271, 1985.[ISI][Medline]
  27. Jaschke WR, Cogan MG, Sievers R, Gould R, and Lipton MJ. Cine-CT measurement of cortical renal blood flow. J Comput Assist Tomogr 11: 779–784, 1987.[ISI][Medline]
  28. Jones AT, Hansell DM, and Evans TW. Pulmonary perfusion in supine and prone positions: an electron-beam computed tomography study. J Appl Physiol 90: 1342–1348, 2001.[Abstract/Free Full Text]
  29. Kaneko K, Milic-Emili J, Dolovich M, Dawson A, and Bates D. Regional distribution of ventilation and perfusion as a function of body position. J Appl Physiol 21: 767–777, 1966.[Free Full Text]
  30. Lisbona R, Dean GW, and Hakim TS. Observations with SPECT on the normal regional distribution of pulmonary blood flow in gravity independent planes. J Nucl Med 28: 1758–1762, 1987.[Abstract/Free Full Text]
  31. Matsuoka S, Uchiyama K, Shima H, Terakoshi H, Nojiri Y, Oishi S, and Ogata H. Detectability of pulmonary perfusion defect and influence of breath holding on contrast-enhanced thick-slice 2D and on 3D MR pulmonary perfusion images. J Magn Reson Imaging 14: 580–585, 2001.[CrossRef][ISI][Medline]
  32. Milic-Emili J, Henderson J, Dolovich JAM, Trop MB, and Kaneko D. Regional distribution of inspired gas in the lung. J Appl Physiol 21: 749–759, 1966.[Free Full Text]
  33. Mintun MA, Ter-Pogossian M, Green MA, Lich LL, and Schuster DP. Quantitative measurement of regional pulmonary blood flow with positron emission tomography. J Appl Physiol 60: 317–326, 1986.[Abstract/Free Full Text]
  34. Mure M, Domino KB, Lindahl SGE, Hlastala MP, Altemeier WA, and Glenny RW. Regional ventilation-perfusion distribution is more uniform in the prone position. J Appl Physiol 88: 1076–1083, 2000.[Abstract/Free Full Text]
  35. Murphy D, Nicewicz J, Zabbatino S, and Altin R. Local pulmonary blood flow by ultrafast computed tomography: a pilot study. Chest 89: 451S, 1986.
  36. Musch G, Layfield JDH, Harris RS, Fischman MJ, and Venegas JG. Topographical distribution of pulmonary perfusion and ventilation, assessed by PET in supine and prone humans. J Appl Physiol 93: 1841–1851, 2002.[Abstract/Free Full Text]
  37. Robertson H, Krueger M, Saba O, Chon D, Shikata H, Hlastala M, Beck K, and Hoffman E. Comparison of regional pulmonary blood flow measured by CT indicator dilution (CTID) and intravenously injected fluorescent microspheres (FMS). Am J Respir Crit Care Med 169: A407, 2004.
  38. Sodani G, Sergiacomi G, Orlando A, Albani E, Romagnoli A, Squillaci E, Masala S, and Simonetti G. Perfusion MRI of the lung: preliminary results in twenty healthy volunteers. Radiol Med (Torino) 103: 45–54, 2002.
  39. Tajik JK, Tran BQ, and Hoffman EA. CT-based assessment of regional pulmonary blood flow parameters: an update. In: Medical Imaging 1999: Physiology and Function From Multidimensional Images. San Diego, CA: SPIE, 1999.
  40. Treppo S, Mijailovich SM, and Venegas JG. Contribution of pulmonary perfusion and ventilation to heterogeneity in V/Q measured by PET. J Appl Physiol 82: 1163–1176, 1997.[Abstract/Free Full Text]
  41. Walther SM, Domino KB, Glenny RW, and Hlastala MP. Pulmonary blood flow distribution in sheep: effects of anesthesia mechanical ventilation and change in posture. Anesthesiology 87: 335–342, 1997.[CrossRef][ISI][Medline]
  42. Wang T, Schultz G, Hebestreit H, Hebestreit A, Hahn D, and Jakob PM. Quantitative perfusion mapping of the human lung using 1H spin labeling. J Magn Reson Imaging 18: 260–265, 2003.[CrossRef][ISI][Medline]
  43. West JB, Glenny RW, Hlastala MP, and Robertson HT. Importance of gravity in determining the distribution of pulmonary blood flow. J Appl Physiol 93: 1888–1889, 2002.[Free Full Text]
  44. West JB. Distribution of pulmonary blood flow and ventilation measured with radioactive gases. Scand J Resp Dis Suppl 62: 9–13, 1966.