Departments of 1 Anaesthesiology and 2 Radiology, Johannes Gutenberg-University, School of Medicine, Langenbeckstrasse 1, D-55131 Mainz, Germany. 3 Department of Anesthesia, University of Pennsylvania, 3400 Spruce Street, Philadelphia, PA 19104, USA 4 Present address: Veterans Administration Medical Center, Los Angeles, University of California, 11301 Wilshire Boulevard, Los Angeles, CA 90070, USA
Corresponding author. E-mail: klm@mail.uni-mainz.de This study contains parts of the doctoral thesis of Elena Ribel.
Accepted for publication: June 20, 2003
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Methods. Six anaesthetized pigs underwent pressure-constant ventilation (FIO2=1.0, inspiratory:expiratory ratio=1:1) before and after induction of lung damage by saline lavage. Mean airway pressure (P) was varied (8, 13, 18, 23, 28, 33, and 38 cm H2O) in random order. At each P
level, dCT acquisitions were performed over several respiratory cycles (Somatom Plus4, Siemens; supradiaphragmatic transverse slice; thickness=1 mm; temporal resolution=100 ms). During scanning at each P
, arterial and mixed venous blood were obtained for blood gas analysis and shunt calculation. In each CT image, fractional areas (FA) of defined density ranges representing ventilated lung and atelectasis were determined by planimetry using dedicated software. The FA data of individual 100 ms scans were averaged over several respiratory cycles, and expressed as mean FA in percentage of total lung area at each P
. For atelectatic lung parenchyma a quantitative relationship of the respective mean FA to shunt fraction was studied using regression analysis.
Results. Under steady-state conditions, mean FA of atelectasis correlated linearly with the calculated shunt fraction (healthy lungs, r=+0.76; lavaged lungs, r=+0.89). There is a non-linear relationship between mean FA of ventilated lung parenchyma and mean FA of atelectasis with PaO2.
Conclusions. We conclude that dCT allows assessment of the effects of ventilator adjustments and resultant P; changes upon lung aeration and oxygenation rapidly, and with good spatial and temporal resolution. This may benefit patients with acute lung injury, whose ventilatory pattern may be optimized as early as during their first diagnostic workup.
Br J Anaesth 2003; 91: 699708
Keywords: complications, acute respiratory distress syndrome; lung, lavage; lung, respirator therapy; measurement techniques, dynamic computed tomography; pig
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
We hypothesized that dCT was capable of imaging the lung during continuous respiration (i.e. without interrupting the respirator settings), and that density measurements in dCT image series correlated with clinical parameters of lung oxygenation (i.e. the pulmonary shunt-fraction).
In this study using pigs, we therefore compared animals with ventilated, but otherwise healthy lungs, as well as with surfactant-depleted lungs. We assessed dynamic CT-based pulmonary densitometry and calculated pulmonary shunt-fraction from blood gas analysis during controlled pressure-constant ventilation with varied mean airway pressure (P).
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
After instrumentation, the animals were transferred to the CT unit (Somatom Plus 4, Siemens, Erlangen, Germany), positioned supine, and connected to an ICU ventilator (Servo 900 C, Siemens, Erlangen, Germany), and switched to a pressure-constant ventilation (PCV) mode.
In the CT scanner, blood gas status was analysed both in a continuous fashion, using an intra-arterial sensor (Paratrend 7, Philips Inc., Best, The Netherlands) and intermittently by sampling and analysing arterial and mixed venous blood (ABL 500/OSM 3, Radiometer, Copenhagen, Denmark). Both inspiratory and expiratory oxygen and carbon dioxide concentrations were measured using a side-stream gas analyser (Capnomac Ultima, Datex-Ohmeda, Helsinki, Finland).
Imaging
Dynamic CT imaging (multiscan technique) was performed with the following parameters: tube voltage=120 kV, tube current=110 mA, matrix=512x512 mm, and slice thickness=1.0 mm, resulting in a voxel size of 0.34x0.34x1.0 mm. Images were reconstructed using a high-resolution algorithm. An effective temporal resolution of 100 ms was achieved using an overlapping temporal increment (sliding window), with a total x-ray tube rotation time of 750 ms.
Image analysis
In each CT image, the lung parenchyma was detected and differentiated (i.e. segmented) from non-pulmonary tissues automatically using dedicated software. This segmented total cross-sectional lung area was divided into fractional areas of predefined density ranges, which differentiate atelectatic from ventilated lung parenchyma.5 A density range of 300 to +200 Hounsfield units (HU) was used to define atelectasis, whereas a density range of 910 to 500 HU was used to define ventilated lung parenchyma. The area sustained by each density range was expressed as a fraction of the total cross-sectional lung area (fractional area (FA) in %) in order to allow a within-subject comparison of data. Figure 1 illustrates cyclic changes of ventilated and atelectatic lung area during PCV. For each measurement, the FA data of all individual 100 ms scans were averaged over several respiratory cycles, and expressed as mean FA in % of total lung area at the P chosen.
|
A surfactant-depletion model of ARDS was induced by repetitive lung lavage with warmed isotonic Ringers solution until a PaO2/FIO2 ratio of 100 was achieved.6 If ionotropic support was required to maintain stable haemodynamics following lung lavage, a continuous infusion of 3 (SD 2) µg kg1 h1 epinephrine was administered.
The above procedure was then repeated in the damaged lungs by changing P in random order to the seven steps already described.
Statistical analysis
Values are given as median, 25th and 75th percentiles. Comparisons between groups (healthy vs lavaged animals) were made using Wilcoxon signed rank test if a minimum of n=6 data points were available in each group. Differences were considered to be significant when P<0.05. Correlation coefficients between density-specific FA and QS/QT (shunt) were calculated using linear regression analysis. Mean bias and limits of agreement between mean FA of atelectasis and conventionally calculated shunt were determined according to Bland and Altman.7
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Haemodynamic and arterial blood gas data
Haemodynamic and blood gas results obtained before and after lavage at each P are shown in Tables 1 and 2.
|
|
|
Mean fractional area of atelectasis and the calculated shunt fraction each showed a linear correlation in healthy and in lavaged lungs. In healthy and lavaged lungs, a correlation coefficient of r=+0.76 and r=+0.89, respectively were calculated. The CT-based atelectasis fraction overestimated the conventionally calculated shunt fraction systematically by 3.4% (bias) within limits of accuracy of ±7.5% (Fig. 3A).
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
In this study, the mean atelectatic lung area showed a linear correlation with shunt-fraction. The mean ventilated lung area (and consequently atelectasis) correlated non-linearly with PaO2. This non-linearity, mainly caused by the buffering effect of haemoglobin, is apparent in the iso-shunt lines (at high shunt-fractions) as described by Nunn and colleagues,8 and was recently investigated in more detail by Petros and colleagues.9 Whereas in these studies FIO2 was varied, in our study the amount of atelectasis was changed to obtain different PaO2 at a constant FIO2 of 1.0. In contrast to the data sets underlying Nunn and colleagues iso-shunt diagrams, however, our experimental setup made it difficult to produce reliable data points in the low PaO2 range (i.e. with the largest amount of atelectasis), for the following reason: in injured lungs and at low airway pressures, this leads to unstable haemodynamics, which explains that in most animals no steady state was achieved under these conditions.
In this study, dCT images were analysed on an image-by-image basis in order to average density measurements over several respiratory cycles. As we aimed to correlate image-based density measurements with conventional parameters of oxygenation, we produced a range of PaO2 values and shunt fractions as large as possible by studying healthy and surfactant-depleted lungs. The lavage ARDS model is an established animal model, specifically known to induce large amounts of atelectasis.6 As expected, the amount of atelectasis at low P clearly differed before and after saline lavage. However, even in the lavaged lungs, nearly complete recruitment of atelectases occurred at high P
. Neumann and colleagues1 compared the response of lung aeration with airway pressure steps in different animal models of ARDS and implied a one-compartmental behaviour of lung aeration. In previous work, we identified two widely separate respiratory time constants in lavage-ARDS, and therefore proposed at least a two-compartment model to discriminate the more rapid aeration process of already open alveoli from slower recruitment of atelectasis in the lavage ARDS model.2
As all measurements were performed during FIO2=1.0, the calculated venous admixture represents a true shunt fraction.10 If a mixture of oxygen and nitrogen is used, the resultant FIO2 has to be taken into account when correlating venous admixture to atelectasis measured throughout an image series.
Densitometry in CT imaging
A variety of CT methods to quantify ventilation and atelectasis have already been described in the literature and have been discussed previously.2 In the current study, we correlated fractional lung areas of different density ranges with PaO2 and shunt. Several studies have demonstrated that specific density windows can be assigned to defined physiological conditions of lung parenchyma, like ventilated lung parenchyma, atelectasis, or overdistension.1113 Other parameters such as mean lung density1 or total amount of air calculated pixel by pixel as described by Puybasset and colleagues14 reduce a continuous spectrum of density values or histogram to a single-number descriptor. Such data reduction needs to be controlled for artefacts by an image-by-image radiological interpretation to exclude influences such as barotraumas or pneumothoraces.
In healthy and lavage-ARDS lungs in this study, PaO2 increased with higher mean airway pressures. At a mean airway pressure of 38 cm H2O, a slight decrease in PaO2 was noted in healthy lungs. This may have been attributable to a decrease in pulmonary blood flow when regional intra-alveolar pressure in overinflated areas exceeded capillary perfusion pressure.
At present, the technique of dCT focuses on the quantification of ventilation, but provides no information on pulmonary perfusion. Thus, the correlation between dynamic lung density distribution and measures of oxygenation described in this study presumes intact pulmonary perfusion. Further development of this technique and/or the additional, independent analysis of regional pulmonary perfusion could address regional ventilationperfusion equilibrium.
A challenge of the method presented is the large number of CT images to be evaluated for quantitative analysis. This problem was addressed by developing a dedicated software tool, which automatically segments the lung in each single CT image. To validate this automated lung segmentation tool, we compared 120 software-based CT evaluations (mean density values of healthy and lavaged lungs) with interactive lung segmentation by a radiologist. There was excellent correlation between both methods with a coefficient r=0.99.5
There are some shortcomings of the dCT technique: presently, it allows imaging only in one predefined transverse slice. We selected a supradiaphragmatic slice, which allows scanning of the maximum transverse lung area in this animal model. However, Lu and colleagues15 found an inhomogeneous distribution of PEEP-induced reopening of collapsed alveolar regions, leading to misrepresentation of ventilated lung volume by imaging of a single plane only. Multi-slice scanners, which allow image acquisition with high temporal resolution simultaneously in multiple separate slices are already commercially available. However, these slices are still in close proximity to each other so that no true three-dimensional reconstruction of the lung is possible. With ongoing development of scanner technology, real-time imaging of the ventilation processes throughout the entire lung may be available in the near future.
In native CT imaging, no distinction between blood, tissue and interstitial water can be made, which potentially leads to an error through blood redistribution during the respiratory cycle in positive pressure ventilation.16 However, as intrapulmonary blood volume is 10% of the total circulatory blood volume (in humans, dogs and rabbits), this influence will not significantly affect quantitative analyses in native dCT imaging.17 The 34% overestimation (bias) of shunt by dCT-based atelectasis measurement may be explained by the inability of our setup to separate blood from atelectatic parenchyma.
Finally, dCT evaluations must quantify ventilation indirectly in contrast to other radiological and nuclear medicine techniques that use radioactive or hyperpolarized gases as direct imaging agents. On the other hand, dCT has the ability to visualize lung morphology in high detail, allowing additional quantification of non-ventilated lung areas. Frequent clinical problems in ARDS which could also be evaluated by dCT include pleural effusion, pneumothorax, bullae, and interstitial fibrosis. In addition, ventilator-associated lung injury can also be detected by dynamic CT evaluation (e.g. barotrauma, volutrauma).
CT exposes the patient to radiation, a fact that is also relevant in dCT imaging because of the large number of images necessary for quantitative evaluation. Heussel and colleagues18 measured a radiation exposure of 47 mGy in a dynamic CT measurement of 10 s duration, using the same scanner and acquisition parameters similar to those used in this study. This dose is approximately equivalent to the total dose necessary for a spiral CT of the entire chest. Further investigations will be necessary to optimize imaging procedures, and to reduce radiation exposure during dynamic acquisition.
Alternative imaging techniques
Currently there is no other image-based technique in clinical use for estimating regional lung aeration in patients with ARDS. Electron beam CT (EBCT) reaches a very high temporal resolution (<50 ms for a volumetric assessment of the total lung). However, EBCT uses relatively high radiation doses and suffers from an inferior spatial resolution compared with spiral CT scanning. Because of technical improvement of spiral CT scanners, EBCT is effectively obsolete.19
Conventional ventilation scintigraphy yields only low spatial resolution,20 whereas ventilation SPECT (using 133Xe, 127Xe or 81mKr) overcomes this limitation. However, to administer the radioisotope gas, both methods require breathing circuit disconnection, which would lead to alveolar collapse in the patient on continuous high positive pressure ventilation. Positron emission tomography (PET) of the lungs using dissolved injected 13N2 as a tracer is a promising new radioisotope-based approach to follow regional distribution of both ventilation and perfusion.21 It is still limited by the availability of PET scanners and has not been applied in clinical ARDS. Novel magnetic resonance imaging (MRI) techniques visualizing inhaled hyperpolarized noble gas (3He or 129Xe) with high temporal resolution (130 ms per image and less),22 allow direct imaging of the ventilation process during a respiratory cycle, and a spatial resolution that is currently unmatched by any nuclear medicine technique. Although these methods have great potential for applications in thoracic radiology and pulmonary physiology, they are not likely to have a significant impact on patients with acute lung injury. In ARDS, the mismatch of ventilation and perfusion is mainly attributable to atelectasis,23 either permanently or cyclically with respiration. Thus, atelectatic lung parenchyma is to be included in quantitative image-based analysis. All techniques using inhaled gas imaging to assess ventilation will miss severely hypoventilated or atelectatic areas. Electric impedance tomography (EIT) represents another relatively new technique that allows quantification and spatial resolution of lung aeration at the bedside without radiation exposure. Kunst and colleagues24 used EIT to determine the lower inflection point and the upper deflection point in an ARDS model. Compared with dCT, EIT still offers considerably less spatial resolution and no direct visualization of atelectasis. Further studies will be necessary to define the potential and clinical significance of both methods.
Clinical implications
In clinical practice, lung aeration is assessed in ventilated ARDS patients by auscultation, chest radiography and arterial blood gas analysis. This approach uses oxygenation failure to quantify the size of the alveolar shunt compartment in the three-compartment model of the lung. Hence, the PaO2/FIO2 ratio indirectly indicates the size of the ideal alveolar compartment (i.e. the open and perfused parts of the lung).
Another, more advanced physiological lung model predicts that, particularly in the diseased lung, an abnormally broad spectrum of ventilation-perfusion (VA/Q) ratios are present. This technique falls short of clinical applicability (e.g. in the assessment of the effect of changes in ventilator settings in the ARDS patient). The multiple inert gas elimination technique (MIGET) uses the different solubilities of six inert gases to calculate fifty different VA/Q ratios during steady state.25 Baumgardner and colleagues26 modified conventional MIGET by incorporating micropore mass inlet spectroscopy (MMIMS-MIGET), which requires much smaller blood samples and allows laboratory analysis at a much more rapid turnover. While the clinical applicability of MMIMS-MIGET holds promise, all MIGET-derived techniques calculate virtual VA/Q compartments over the whole lung, without spatial or temporal allocation to topographical lung regions.
Image based measurements like dCT and EIT add two interesting factors to conventional lung function tests: high temporal and high spatial resolution. The former offers a new understanding of ventilatory processes during continuous respiration (e.g. the importance of ventilatory frequency on regional lung inflation and deflation processes).1 2 If averaged over time, as in this study, dynamic measurements will reflect a more realistic assessment of ventilatory processes than obtained from static measurements, for example the static pressurevolume curve.
Regional lung function tests will be particularly important in inhomogeneous ventilated lungs (e.g. in ARDS), and may be more sensitive than global measurements such as PaO2 or shunt.
Perspectives
CT-based resolution of cyclic lung density variations during ventilation offers an exact regional quantification of lung aeration in healthy and pathologic pulmonary conditions and offers an additional method to assess oxygenation and shunt fraction without the need to sample mixed venous blood. Furthermore, changes in response to altered ventilator settings (e.g. P, ventilatory frequency, and I:E ratio) may be studied. The application of this technology in the clinical arena may allow quantification of the participation of specific lung regions in gas exchange and the definition of the anatomical substrate of gas exchange abnormalities early and rapidly during the diagnostic workup of ARDS patients.
![]() |
Acknowledgement |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 Markstaller K, Eberle B, Kauczor HU, et al. Temporal dynamics of lung aeration determined by dynamic CT in a porcine model of ARDS. Br J Anaesth 2001; 87: 45968
3 Tokics L, Hedenstierna G, Strandberg A, Brismar B, Lundquist H. Lung collapse and gas exchange during general anesthesia: effects of spontaneous breathing, muscle paralysis, and positive end-expiratory pressure. Anesthesiology 1987; 66: 15767[ISI][Medline]
4 Tokics L, Hedenstierna G, Svensson L, et al. V/Q distribution and correlation to atelectasis in anesthetized paralyzed humans. J Appl Physiol 1996; 81: 182233
5 Markstaller K, Arnold M, Dobrich M, et al. A software tool for automatic image-based ventilation analysis using dynamic chest CT-scanning in healthy and in ARDS lungs. Fortschr Roentgenstr 2001; 173: 8305
6 Lachmann B, Robertson B, Vogel J. In vivo lung lavage as an experimental model of the respiratory distress syndrome. Acta Anaesthesiol Scand 1980; 24: 2316[ISI][Medline]
7 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; i: 30710
8 Lawler PG, Nunn JF. A reassessment of the validity of the iso-shunt graph. Br J Anaesth 1984; 56: 132535[Abstract]
9 Petros AJ, Dore CJ, Nunn JF. Modification of the iso-shunt lines for low inspired oxygen concentrations. Br J Anaesth 1994; 72: 51522[Abstract]
10 Hlastala MP, Colley PS, Cheney FW. Pulmonary shunt: a comparison between oxygen and inert gas infusion methods. J Appl Physiol 1975; 39: 104851
11 Wu MT, Chang JM, Chiang AA, et al. Use of quantitative CT to predict postoperative lung function in patients with lung cancer. Radiology 1994; 191: 25762[Abstract]
12 Markstaller K, Kauczor HU, Eberle B, et al. Multi-rotation CT during continuous ventilation: comparison of different density areas in healthy lungs and in the ARDS lavage model. Fortschr Roentgenstr 1999; 170: 57580
13 Kinsella M, Muller NL, Abboud RT, Morrison NJ, DyBuncio A. Quantitation of emphysema by computed tomography using a density mask program and correlation with pulmonary function tests. Chest 1990; 97: 31521[Abstract]
14 Puybasset L, Cluzel P, Gusman P, et al. Regional distribution of gas and tissue in acute respiratory distress syndrome. I. Consequences for lung morphology. CT Scan ARDS Study Group. Intensive Care Med 2000; 26: 85769[CrossRef][ISI][Medline]
15 Lu Q, Malbouisson LM, Mourgeon E, et al. Assessment of PEEP-induced reopening of collapsed lung regions in acute lung injury: are one or three CT sections representative of the entire lung? Intensive Care Med 2001; 27: 150410[CrossRef][ISI][Medline]
16 Drummond GB. Computed tomography and pulmonary measurements. Br J Anaesth 1998; 80: 66571[CrossRef][ISI][Medline]
17 Milnor WR, Jose AD, McGaff CJ. Pulmonary vascular volume, resistance and compliance in man. Circulation 1960; 22: 1307[ISI][Medline]
18 Heussel CP, Hafner B, Lill J, et al. Paired inspiratory/expiratory spiral CT and continuous respiration cine CT in the diagnosis of tracheal instability. Eur Radiol 2001; 11: 9829[CrossRef][ISI][Medline]
19 Tajik JK, Chon D, Won C, Tran BQ, Hoffman EA. Subsecond multisection CT of regional pulmonary ventilation. Acta Radiol 2002; 9: 13046[CrossRef]
20 Morrell NW, Nijran KS, Jones BE, Biggs T, Seed WA. The underestimation of segmental defect size in radionuclide lung scanning. J Nucl Med 1993; 34: 3704[Abstract]
21 Venegas JG. Noninvasive measurement of local VA, Q, and VA/Q distributions by PET. In: Hlastala MP, Robertson HAT, eds. Complexity in Structure and Function of the Lung. Lung Biology in Health and Disease, Vol. 121. New York: Marcel Dekker Inc., 1998
22 Schreiber WG, Weiler N, Kauczor HU, et al. Ultrafast MRI of lung ventilation using hyperpolarized helium-3. Fortschr Roentgenstr 2000; 172: 12933
23 Hedenstierna G, Neumann P. Gas exchange in acute respiratory failure. Minerva Anestesiologica 1999; 65: 3837[Medline]
24 Kunst PW, Vazquez de Anda G, Bohm SH, et al. Monitoring of recruitment and derecruitment by electrical impedance tomography in a model of acute lung injury. Crit Care Med 2000; 28: 38915[ISI][Medline]
25 Wagner PD, Saltzman HA, West JB. Measurement of continuous distributions of ventilation-perfusion ratios: theory. J Appl Physiol 1974; 36: 58899
26 Baumgardner JE, Choi IC, Vonk-Noordegraaf A, et al. Sequential V(A)/Q distributions in the normal rabbit by micropore membrane inlet mass spectrometry. J Appl Physiol 2000; 89: 1699708