| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Heart of England NHS Foundation Trust, Bordesley Green East, Birmingham, B9
5SS, United Kingdom.
2 Department of Radiology, Guy's and St. Thomas' NHS Foundation Trust, London,
United Kingdom.
3 Department of Medical Physics, Guy's and St. Thomas' NHS Foundation Trust,
London, United Kingdom.
4 Department of Clinical Perfusion, Guy's and St. Thomas' Foundation NHS Trust,
London, United Kingdom.
Received March 20, 2007;
accepted after revision May 19, 2007.
Funded in part by an educational grant from the Royal College of
Radiologists, 2004.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionAPPENDIX 1: Mathematic Model...References |
|---|
MATERIALS AND METHODS. A closed pulsatile cardiopulmonary bypass circuit was connected to tubes traversing a water bath to simulate the abdominal aorta and inferior vena cava. Three smaller interconnecting acrylic plastic tubes were connected as branches to the aortic tubing to simulate branch vessels. One of the three tubes, the control, had no holes in it, one had a 100-µm hole, and one had a 280-µm hole. The leakage rates were predetermined with a cardiac output of 2 and 4 L/min and with a mean arterial pressure (MAP) ranging from 30 to 100 mm Hg for each hole size. The following studies were performed for each of the predetermined leakage rates. For study 1, 16-MDCT was performed using bolus tracking after 35 mL of contrast medium had been injected into a simulated peripheral vein. For study 2, DSA was performed using a 4-French straight catheter placed 10 cm proximal to the holes (selective first aortic branch cannulation). For study 3, DSA was performed with a catheter placed in the small branch at the site of the hole (highly superselective). For study 4, 16-MDCT was performed with a catheter placed as in study 2, 10 cm proximal to the holes, for the detection of lower leakage rates. Cine loops of MDCT and DSA images were examined by two blinded observers to detect extravasation from the holes in the tubes (i.e., the branch arteries). Interobserver agreement was studied using Cohen's kappa statistic.
RESULTS. The threshold to detect bleeding was as follows for each study: For IV contrast-enhanced MDCT (study 1), it was 0.35 mL/min; DSA with a catheter 10 cm proximal to the holes (study 2), 0.96 mL/min; DSA with a catheter at the holes (study 3), 0.05 mL/s or lower; and intraarterial selective MDCT (study 4), 0.05 mL/s or lower. The ease of detection improved with increasing MAPs and larger volumes of leakage. Interobserver correlation was excellent.
CONCLUSION. In vitro, IV contrast-enhanced MDCT is more sensitive than first-order aortic branch-selective DSA in detecting active hemorrhage unless the catheter position is highly superselective and is close to the bleeding artery. These results suggest that MDCT can be used as the initial imaging technique in the diagnosis of active hemorrhage if the clinical condition of the patient allows.
Keywords: abdominal trauma • digital subtraction angiography • emergency radiology • hemodynamics • hemorrhage • MDCT • trauma
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAPPENDIX 1: Mathematic Model...References |
|---|
Over the past decade, several investigators have shown that active contrast extravasation is a specific sign of intraabdominal bleeding and predicts the need for active intervention. When this sign is seen, hemorrhage is assumed to be occurring at a brisk rate [4]. However, cases have been reported in which active bleeding was seen on CT but was not seen on angiography performed soon after CT [1, 5]. In vivo, bleeding can be intermittent, even from minute to minute [6], and may stop due to tamponade, spasm, or thrombosis. Therefore, a direct clinical comparison of CT and angiography is difficult even when the two tests are performed one immediately after the other.
|
|
|
We designed an in vitro experiment to compare the detection of active abdominal bleeding on CT and on angiography and to ascertain the threshold of detectable bleeding with each technique using a simple near-physiologic model.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAPPENDIX 1: Mathematic Model...References |
|---|
The phantom insert was placed in a water tank that was filled to 20 cm (Fig. 1B), and the inserts were placed with a clamp at a depth of 10 cm to simulate the abdomen. To prevent movement of water during CT due to table movement, the tank was filled with soaked, fully expanded foam sponges and the holes in the tubes were covered with sponge sleeves.
The perfusing fluid in the circuit was a 3:1 mixture of 5% dextrose and Hartmann solution to match the specific gravity and viscosity of circulating blood. One milliliter of contrast medium (iohexol [Omnipaque 300, Schering Health Care]) was added, the amount arrived at by trial and error, to raise the CT attenuation of the fluid to 30 H, which is the approximate density of unopacified blood.
The parameters that were varied in the study were the perfusion rate Q (L/min) and the MAP (mm Hg). The leakage rate from the holes was measured using different combinations of these parameters. The oxygenator reservoir (Fig. 1A) was filled to 0.5 L. The total volume of circulating fluid in the system was 800 mL. Because low leakage rates (< 1 mL/min) were used in this study, the fluid loss from the holes was too little to be measured accurately from the drop in the fluid level in the reservoir. Therefore, the leakage rate from the holes was measured in air over a sufficient time to collect a sample on the order of 5 mL, and the leakage rate was calculated by dividing the leaked volume by the time over which the sample was acquired.
The possible errors in making the measurements in free air compared with those in fluid at depth were investigated using a column of water at 63 mm Hg through a Perspex tube with a larger-than-3-mm-diameter hole. The extravasation rate in free air was estimated by measuring the amount of fluid leaked from the hole over a 5-minute period, and the rate was calculated as milliliters per minute. The extravasation rate in fluid at a depth was calculated by measuring the drop of fluid level in a reservoir using the same parameters over a 5-minute period. The mean of three measurements was calculated. The conditions were consistent with those applicable to the Bernoulli equation—that is, steady nonrotational flow for an ideal fluid.
During all measurements, the surface of the water in the tank was kept level with the surface of the water in the pump reservoir to ensure that there was no difference in hydrostatic pressure. For the MDCT study, a contrast injector introduced contrast medium into a tube, simulating an injection into a peripheral vein, 30 cm before the reservoir (filled to 0.5 L) where the contrast material underwent mixing analogous to blood in the lungs before being pumped into the aorta.
The phantom was placed in a 16-MDCT unit (MX8000 IDT, Philips Medical Systems) and scanned using a volume scanning protocol for the abdominal aorta. Bolus-tracking software was used, and a region of interest (ROI) on the arterial side of the aortic section close to the branch insert was chosen. Standard injector settings were used to administer a contrast bolus of 30 mL at 3.5 mL/s. This reduced amount of contrast medium, approximately one fourth, was used to account for the reduced circulating volume in this setup compared with in vivo. The volume scanning over the pipe inserts was started when the CT attenuation value in the ROI rose above 140 H (arterial phase). For MAPs of 20, 30, 40, 50, and 60 mm Hg and a perfusion rate of 4 L/min, the same volume scanning was repeated after manually scanning at 70 seconds (portal phase) and at 2 minutes (delayed phase). The settings chosen were 120 kVp, the automatic setting for milliamperes (mAs), a 0.75-mm collimation, and a pitch of 1.5 on a 16-MDCT scanner.
The phantom was also examined in an angiography room using a biplane digital subtraction angiography (DSA) system (V3000, Philips Medical Systems) with a 22-cm field-of-view intensifier at 2 frames per second over the same range of perfusion rates (Q) and MAPs as those used for the CT experiments. A 4-French straight catheter with multiple side holes was inserted through a Tuohy-Borst adaptor into the (simulated) arterial tree and contrast medium was injected in two positions.
Sets of Images Obtained
IV contrast-enhanced MDCT—The phantom was scanned at flow
rates (Q) of 2 and 4 L/min and at MAPs of 30, 40, 50, 60, 70, 80, 90,
and 100 mm Hg in the arterial phase (25- to 30-second delay per bolus
tracking). Scans were also obtained at 70 seconds and at 2 minutes after
injection with a Q of 4 L/min and MAPs of 20, 30, 40, 50, and 60 mm
Hg. The leakage was so obviously visible that scanning using the higher MAPs
was not performed in the portal or delayed phase.
Intraarterial contrast-enhanced MDCT—Another CT series was obtained after injecting 20 mL of contrast medium at 4 mL/s through a catheter placed 10 cm proximal to the holes to simulate CT first-order branch-selective arteriography at 4 L/min and with MAPs of 10 and 20 mm Hg. The purpose of this series was to detect leaks lower than that detected on IV contrast-enhanced CT. Higher leakage rates were thought to be superfluous and were not studied.
Selective DSA—Cine loop images were obtained with a 4-French catheter placed 10 cm proximal to the holes into a Y connector to simulate selective aortic first-order branch cannulation. Twenty milliliters of contrast material was injected through an automatic injector at 8 mL/s. As with MDCT, these injections were performed at flow rates (Q) of 2 and 4 L/min and at MAPs of 30, 40, 50, 60, 70, 80, 90, and 100 mm Hg.
Highly selective DSA—Cine loop images with the catheter at the level of the holes to simulate highly selective cannulation were obtained using the same variables as described earlier. Ten milliliters of contrast medium was hand injected.
Image Analysis
Two of the authors who were blinded to the leakage rates analyzed the DSA
and CT images independently on cine loops. All workstation functions,
including maximum intensity projections and multiplanar reconstructions, were
available for review. The reviewers recorded a yes–no response according
to whether they detected extravasation (Tables
1 and
2). Scores were assigned to the
images in a random sequence, and the data were analyzed several weeks after
the experimental study. A researcher not involved in scoring the images kept
the key and also frequently changed the order of the tubing inserts in the
phantom to prevent an identifiable pattern of leakage. Cohen's kappa statistic
was used to test interobserver agreement. A kappa value of 0.8 or greater was
considered excellent correlation.
|
|
To validate the flow dynamics and assess the applicability of this phantom to what happens in vivo, a mathematic model that predicts a relationship among the variables that influence leakage rates from a bleeding artery was devised. The expected bleeding rates obtained using the mathematic model were plotted against the observed bleeding rates on graphs; these finding are discussed further and illustrated in Appendix 1.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAPPENDIX 1: Mathematic Model...References |
|---|
Extravasation rates measured in air with the 100- and 280-µm holes at differing cardiac outputs and MAPs and their detection with MDCT and DSA are shown in Tables 1 and 2, respectively. The leakage rates increase with MAP and hole diameter. The leakage rates at 10 and 20 mm Hg were too small to measure experimentally. They have been extrapolated for the 100-µm hole at 4 L/min from a plot of leakage rates versus mean hole pressures and were estimated to be 0.05 and 0.20 mL/min for MAPs of 10 and 20 mm Hg, respectively.
The minimum leakage rates detectable in the four image studies were as follows: For IV contrast-enhanced MDCT, the minimum leakage rate detected by both observers was 0.35 mL/min (Table 1). All other higher leakage rates were detected with increasing subjective ease. For MDCT after arterial injection of contrast medium (CT mesenteric arteriography) through a catheter into a simulated first aortic branch, the minimum leakage rate was 0.05 mL/min—that is, the lowest threshold was detected. The threshold of detection on DSA with the catheter in a Y connector 10 cm away from the holes, simulating first aortic branch cannulation like the superior mesenteric artery, was 0.87 mL/min for one observer and 0.96 mL/min for the other (Tables 1 and 2). Like with MDCT, the higher leakage rates were detected by both observers with increasing ease. Leaks during DSA with the catheter at the hole (highly superselective position) were detected at all cardiac outputs and all MAPs including 10 and 20 mm Hg—that is, the lowest threshold of 0.05 mL/min was detected.
There was excellent interobserver correlation for the detection of leaks on
both CT and DSA (
= 1 for all CT studies and 0.82 for DSA). The volume
of extravasated contrast medium increased, and the ease of detection improved
on the 2-minute delayed phase of imaging in the five sequences performed (Fig.
2A,
2B).
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAPPENDIX 1: Mathematic Model...References |
|---|
High-speed narrow-collimation MDCT allows large volume coverage, produces images with decreased motion and respiratory artifacts, and can be accurately timed to acquire data in the arterial or venous phase. The excellent z-axis resolution, near isotropic voxel resolution, and ability to localize bleeding accurately should improve the diagnostic value of CT [1, 14]. However, CT is hampered by the fact that contrast medium is delivered IV and becomes diluted by unopacified blood at the point of leakage. A detailed experimental study in swine using controlled extravasation of contrast-enhanced blood showed that helical CT can depict active colonic extravasation below 0.4 mL/min [9]. In that study, a mathematic model was used with computer simulation based on iodine mass and volume balances, and the authors used dilution as a function of time to arrive at the figures. They suggested that with a smaller slice thickness, higher mean aortic enhancement, and decreased conspicuity threshold, the sensitivity of CT could be as low as 0.2 mL/min or lower, matching the high sensitivity of technetium-99 m RBC studies.
In the study reported here, we performed an in vitro study, for the reasons outlined earlier, using a phantom that permitted variation of MAPs to simulate physiologic dilution of administered IV contrast medium in a pulsatile heart–lung circuit. Although we used a completely different experimental arrangement than Kuhle and Sheiman [9] for our study, it produced sensitivity thresholds similar to those reported by Kuhle and Sheiman.
During angiography, a catheter position of 10 cm away from the holes was chosen to mimic what happens in practice, such as during an injection into the superior mesenteric artery to detect bleeding from the vasa recta of the bowel. This yielded a threshold sensitivity of 0.96 mL/min for DSA and 0.35 mL/min for IV contrast-enhanced CT. Therefore, in practice, IV contrast-enhanced MDCT appears more sensitive than selective first-branch DSA. When the catheter position was at the holes, which is usually impractical in clinical situations, DSA detected bleeding at 0.05 mL/min or less and was independent of MAP. Similarly, performing CT with the catheter in the first aortic branch (simulating CT mesenteric arteriography) increased the sensitivity of CT to 0.05 mL/min or less. Therefore, MDCT was more sensitive than DSA in detecting active bleeding.
The results of our study also suggest that IV contrast-enhanced CT would be ineffective in detecting bleeding if the MAP falls below 30 mm Hg. However, patients with this MAP level are usually too unstable to undergo CT. Subjective detection of the leak was easier in the delayed phase (2 minutes) of imaging than in the arterial phase, although this finding has limitations, which we discuss later in this article.
Limitations and Pitfalls
Detection of bleeding on CT depends on the differential optical density of
extravasated contrast medium and the substance into which extravasation
occurs. For example, it is easier to detect bleeding into fat than within
liver parenchyma. CT detection of bleeding also depends on whether the
extravasated contrast medium washes away or is circumscribed and collects into
a localized space [11]. In our
phantom, extravasation occurred in a nonlocalizing wet sponge, which is
probably equivalent to extravasation in abdominal fat. The figures for
CT-detected leakage rates in solid organs would therefore be higher than those
reported here.
Because of the pulsatile flow in the heart–lung machine, air was being aspirated into the circuit, particularly at high flow rates, when establishing the leakage rates at different MAPs before the main experiment. This effect was more pronounced for the large hole size and higher flow rates. Nevertheless, because both MDCT and DSA were performed using an identical arrangement, the observed superior sensitivity of CT compared with DSA is valid, although the numeric value of the threshold may vary.
There was no capillary bed or glomerular filtration in our circuit because introduction of a hemodialysis filter would have made the phantom more complicated. This may partly account for the easier detection of leaks at 2 minutes (Fig. 2A, 2B) rather than 30 seconds. This is particularly true because iohexol (Omnipaque) would have redistributed to the extracellular space in vivo. However, it should be borne in mind that the vascular half-life of these agents is on the order of 20 minutes and that only 18% of the administered contrast load is normally excreted by the kidneys in 30 minutes [15]. Therefore, at 2 minutes a large proportion of the contrast material would remain in the vascular tree in vivo.
Bowel peristalsis has not been accounted for in this study; however, in a clinical situation, it is usually more of a problem during angiography than during CT.
In addition to active contrast extravasation, an experienced angiographer often relies on other signs, such as vessel cutoff, mural irregularity, spasm, diffuse vasoconstriction, arteriovenous fistula, and false aneurysm, to localize the site of bleeding [1]. Although these vascular abnormalities, particularly the latter two, are visible on MDCT, visualization by the reporting radiologist may be more difficult on MDCT than on DSA.
Practical Applications
Detection of active contrast extravasation in abdominal trauma was reported
in 1989 [16]. Its incidence on
MDCT is thought to be approximately 13%
[17]. This sign is used to
guide emergency treatment in the spleen
[18,
19], liver
[20], bowel
[21], pelvis
[22], and other organs. The
superior sensitivity of CT could make this technique an excellent screening
tool in the detection of bleeding and could potentially prevent unnecessary
invasive angiography or laparotomy. Given the small amount of bleeding that
can be detected on MDCT, small active contrast extravasation may be treated
conservatively, particularly if contained within a solid organ and not
communicating with the peritoneal space
[1,
23].
Two important issues in acute gastrointestinal bleeding are the site of hemorrhage and whether the bleeding is ongoing or has stopped. In 75% of patients with massive gastrointestinal bleeds, bleeding ceases with conservative treatment [24] and negative angiography is common. Localization of gastrointestinal bleeding is important because surgery without preoperative localization results in more extensive bowel resection and carries higher intra- and postoperative mortality [25]. Therefore, appropriately performed CT could be helpful. Colonoscopy can be difficult to perform during acute bleeding and can fail to reveal the site of hemorrhage in up to 40% of cases [26]. Nuclear medicine studies, using technetium sulfa colloid or technetium-labeled erythrocytes, are sensitive techniques for showing lower gastrointestinal bleeding at rates as low as 0.1 mL/min. However, localization of the bleeding is inaccurate because of colonic peristalsis and the movement of radiotracer.
The results of several clinical studies and case series [27, 28] have also suggested the usefulness of CT in the detection of acute gastrointestinal bleeding that was undetectable by standard means. CT angiography using a dual-detector CT unit detected 90% of cases of colonic angiodysplasia in a group of elderly patients, obviating invasive diagnostic angiography [29]. In a more recent study [30] of 26 patients, a sensitivity and a specificity of MDCT of 91% and 99%, respectively, were reported assuming DSA as the gold standard. However, the two false-positive cases in that series could well have been due to bleeding below the threshold of a standard DSA study. In addition to its high sensitivity, CT is more likely to depict abnormalities outside the bowel and may reveal findings even in the absence of active hemorrhage.
For detection of occult gastrointestinal hemorrhage, CT mesenteric arteriography should be considered in which CT is performed after contrast injection through a catheter placed in the aorta or superior mesenteric artery. In our phantom, CT mesenteric arteriography detected bleeding rates of approximately 0.05 mL/min. As early as 1984, Stanley et al. [31] suggested the usefulness of postangiography CT in detecting occult gastrointestinal hemorrhage missed on conventional angiography. Ettorre et al. [32] were able to establish the site of hemorrhage in 72% of patients with obscure gastrointestinal bleeding by performing CT after a mid aortic injection of contrast medium, a technique that was shown to be more sensitive than conventional angiography.
In view of the findings of this study, we now routinely perform IV contrast-enhanced MDCT for the detection of acute lower gastrointestinal bleeding. In our anecdotal experience, we have found active bleeding in patients who underwent superselective DSA based on MDCT visualization of active bleeding even after the selective first-order aortic run was negative. We obtain unenhanced, arterial, and delayed series of images and use water as oral contrast medium before triaging to surgery or angiography.
Conclusion
In vitro, in this setup, IV contrast-enhanced MDCT has a lower threshold
(0.35 mL/min) than first-order selective DSA (0.96 mL/min) in the detection of
active bleeding. CT is a quick, inexpensive, and reliable diagnostic
investigation that can localize the site of bleeding and guide appropriate
treatment and may prevent a false-negative DSA.
APPENDIX 1: Mathematic Model of the Extravasated Flow Rate
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAPPENDIX 1: Mathematic Model...References |
|---|
The Bernoulli equation is given by the following:
 |
is the fluid density, 
is the flow
velocity, y is the vertical height in the direction opposite
gravitational force, g is the acceleration due to gravity, and 
is a constant along a given streamline. During the experiment, the holes in
the tubes were kept level with the pump; hence, y is a constant
throughout and can be eliminated from the equation. In the system under consideration, the effect of viscous forces is known to be relagtively small. Therefore, the drop in pressure from the arterial side of the phantom insert to the venous side of the phantom insert is small in comparison with the back pressure exerted in the circuit due to the valve near the pump input. To estimate the small pressure drop across the acrylic plastic (Perspex, Lucite International) inserts, Poiseuille's law was applied across the two branches—that is, the main arterial branch and the Perspex insert branches. Because the pressure difference and, hence, energy losses due to viscosity are small across these branches, one may apply this equation in combination with the Bernoulli equation, even though the latter strictly requires an energy-conservative system.
Using Poiseuille's law, where Q0 is total flow,
Q1 is the flow in the main arterial (aortic) branch of
internal radius R1 and length L1; and
Q2 is the flow in each of the three arterial branches in
the Perspex insert, each of which has an internal radius
R2 and length L2. The viscosity of the
solution is 
. The flow in each of the segments is calculated as follows:
 |
 |
The ratio of flows in the small artery tubes in the Perspex insert compared
with the ratio of flows in the main arterial branch is as follows:
 |
The ratio of flow rates depends on the resistance resulting from the
internal radius of the tube and the length of the pipe, with narrower and
longer pipes giving rise to more resistance and less flow. From flow
continuity, we have that the sum of the flow rates in all the arterial
branches is equal to the total flow rate:
 |
P is
given as follows (Fig. 1A):
 |
From inspection of the circuit shown in Figure
1A,
1B,
1C, the pressure at the input
to the pump will be P2, and the pressure
P1 will be slightly greater than this value under this
experimental arrangement. Furthermore, the pressure along arterial branches a,
b, or c will drop linearly along the length of these vessels under this model
from P1 down to P2 at the distal end
(Fig. 1A). In this case,
because the holes can be treated as being approximately halfway along the
arterial pipes, the pressure at these points will be approximately
[(P2+ P) / 2].
To estimate the extravasation rate at the hole, we applied the Bernoulli
equation. We assumed that the flow rates were small in relation to the flow in
the vessel and that extravasated liquid was forced by the excess pressure in
the fluid from the stationary boundary layer on the inside surface of the
tube. The extravasated fluid velocity, v, can be expressed in terms
of the excess pressure, P, inside the vessel as follows:
 |
 |
The instantaneous pressure P(t) at the hole site will be
equal to [(P1 – P) / 2]. The
extravasation rate therefore is proportional to the square of the radius of
the hole and varies as the square root of the fluid pressure in the vessel.
Under the experimental conditions used, the pressure deficit at the hole due
to viscous effects—that is, P / 2—was in some
cases up to 63% of the pressure P1 in the case of the
lowest leakage rates encountered.
One further correction must be made for the pulsatile nature of the flow.
As an approximation, the fluid is assumed to be incompressible, the influence
of elasticity in the vessel walls is neglected, and the flow is assumed to be
nonturbulent. The pump is simulated by a pressure, which varies in a 50%
on–50% off step cycle between 0 and Pmax. The mean
pressure is equal to Pmax / 2 by averaging over many pump
cycles. However, by replacing P1 by the instantaneous
pressure P(t) and integrating over one pump cycle, the mean
leakage rate over a pump cycle is found to be as follows:
 |
. The leakage rate increases approximately as the square root of the MAP and the radius of the hole squared.
This mathematic model gives a very approximate functional form of the extravasation rate being proportional to the square root of the MAP—that is, the peripheral resistance. For a given MAP and hole diameter, the model predicts that the leakage rate will fall slightly with increasing output because of the greater pressure drop at the hole due to the effect of viscosity. Using this mathematic model, we plotted the predicted extravasation rates for each MAP and for cardiac outputs of 2 and 4 L/min. We compared the predicted rates with the actual observed extravasation rates for the same parameters (Figs. 3 and 4). During the leakage measurements in air, we noted that the pulsatile nature of flow resulted in aspiration of air back into the leakage hole during part of the pump cycle and that visually this effect was more pronounced at 4 L/min than at 2 L/min. The effect was also more pronounced with the larger hole. For the smaller hole size, the model matches the measured leakage rates for Q = 2 L/min but the measured leakage rates are lower than the predicted rates for Q = 4 L/min (Fig. 2A, 2B), presumably because of this effect. The curves show that the measured leakage rates have approximately the same functional form as predicted by the model (Figs. 3 and 4).
|
|
Acknowledgments
We gratefully acknowledge the educational grant provided by the Royal
College of Radiologists in 2004 to undertake this study.
We also thank Schering Health Care, West Sussex, UK, for providing some of the contrast agents used for this study and Angshu Bhowmik for calculating the kappa statistic on the data.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAPPENDIX 1: Mathematic Model...References |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
); Q = 4 L/min, 0.1-mm hole (
). Q = 2 L/min,
measured leak, 0.1-mm hole (
); Q = 4 L/min, measured leak, 0.1-mm hole
(
).