Departments of 1Anaesthesiology and 2Radiology, Johannes Gutenberg-University Mainz, Langenbeckstr. 1, D-55131 Mainz, Germany*Corresponding author: Department of Anaesthesiology, Johannes Gutenberg-University, Mainz, Langenbeckstr. 1, D-55131 Mainz, Germany
Accepted for publication: April 12, 2001
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Br J Anaesth 2001; 87: 45968
Keywords: measurement techniques, dynamic computed tomography; lung, pulmonary time constants; lung, lavage; pig
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A recent CT acquisition technique, called dynamic multiscan CT, allows continuous image acquisition with a temporal resolution of 250 ms. This superior resolution allows changes in radiologic attenuation of the lung during mechanical ventilation to be followed. Neumann and colleagues investigated the temporal behaviour of poorly aerated lung regions and atelectases during inspiration and expiration in pigs with experimental lung damage. Using a monoexponential relationship, they described the kinetics of mean radiologic density of the lung (mean lung density, MLD) and of two density ranges describing poorly aerated lung and atelectasis.2 We considered that although aeration of the healthy lung may have one-compartmental behaviour, aeration of damaged lungs may be better described by considering more than one compartment. To obtain maximum sensitivity for ventilation-induced changes in aeration, we analysed those lung density ranges which had shown the greatest area change with inspiratory and expiratory manoeuvres in a previous study of similarly treated animals (see Fig. 5).3 To prove our hypothesis, we used an experiment to allow us to discriminate different : (1) the repeated lavage model, which has been described as a model for lung collapse and recruitment phenomena;2 and (2) rectangular airway pressure steps between ZEEP and 50 cm H2O, in order to allow both atelectasis formation and nearly complete recruitment.
|
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Study procedure
Baseline values of haemodynamics, blood gases, and acidbase status were obtained. Two scout views were taken during end-inspiratory and end-expiratory breathholds and a transverse, supradiaphragmatic slice was defined by a reference scan. Thus, a fixed table position was defined which allowed the visualization of lung parenchyma between the apex of the heart and the diaphragm, both during inspiration and expiration.
Airway pressure step manoeuvres were performed with healthy lungs and again after surfactant-depletion by repeated lavage of the lungs. Lung lavages were performed 9±1 times with 1 litre of isotonic Ringers solution.4 Inotropic support was provided by continuous infusion of 3±2 µg kg1 h1 epinephrine after lung lavages.
To generate a quasi-rectangular increase or decrease in airway pressure, the respirator was switched to its CPAP mode, and the CPAP level was increased swiftly in one step from end expiration (ZEEP=0 cm H2O) to 50 cm H2O or from CPAP of 50 cm H2O to ZEEP, respectively. The CT acquisition was started 5 s before the airway pressure step and continued for at least 30 s after the step change. Volume-controlled ventilation was resumed afterwards, until baseline haemodynamics and blood gases were restored.
CT acquisition parameters
In all dynamic multiscan CT acquisitions, the tube voltage was set to 120 kV and the tube current to 110 mA. A matrix of 512x512 and a slice thickness of 1.0 mm was used. Images were reconstructed using the high-resolution algorithm. Total rotation time of the x-ray tube was 750 ms allowing for an overlapping temporal increment, i.e. an effective temporal resolution, of 250 ms.
Quantitative analysis of CT scans
In each lung image, the total lung area was determined semi-automatically using a dedicated software (Pulmo-Software®, Siemens, Germany). An interactive correction was carried out if the total lung area resulting from the automatic segmentation did not represent the anatomical lung area. Such errors occurred predominantly in damaged lungs. The boundaries of the lung were defined by the ribs, the aorta, and the heart. All segmentations were performed by one investigator to avoid operator-related variation. The total lung area (density range from 1024 to +200 HU) was automatically divided into fractional areas of defined densities.3 The density range from 910 to 700 HU was used to describe aerated lung parenchyma in healthy porcine lungs. The range from 910 to 300 HU represented aerated tissue in the lavaged lungs model. For each density range, the corresponding relative lung area was evaluated planimetrically.
Radiological interpretation
To verify the quantitative analysis by density ranges, every image series was evaluated by a board-certified radiologist (H.-U.K.) for the following criteria: general lung aeration, presence of atelectasis, and a ventral-to-dorsal gradient. An image-by-image evaluation was done to describe the temporal evolution of these criteria.
Identification of lung compartments, their time constants () and fractional sizes (A)
The time behaviour of a specific density range during inflation or deflation was analysed under the a priori assumption of two-compartmental behaviour. The change in aeration (air content) of one homogeneous compartment in response to a rectangular airway pressure increase or decrease can be described by a mono-exponential wash-in or wash-out function, respectively:5
(V, expired volume; , time constant).
In this study, was derived from the temporal dynamics of the area of a defined density range in subsequent CT images. In the following,
A denotes the lung area, which becomes aerated by the airway pressure increase from ZEEP, or collapses during the airway pressure decrease to ZEEP. The lung area at ZEEP was defined as the baseline for the aeration manoeuvre (and as the endpoint of collapse, respectively), and was not taken into account for this analysis. To separate the two time constants
, the increase or decrease of aerated lung area in response to the airway pressure change was fitted by a least-squares fitting procedure using a bi-exponential relationship, i.e.,
In this relationship, A1 represents the fraction of
A, which follows a time constant
1.
A2 represents the fraction of
A with a temporal behaviour characterized by
2.
First, the temporal behaviour of the area of the predefined density range was plotted. The bi-exponential equation (above) was then introduced as user-defined fitting function. Data points were fitted with the software package Origin® (Origin 4.0, Microcal Software Inc., USA). As the fitting procedure requires a selection of starting values for the parameters, we chose the following parameters, which gave reproducible results in simulation curves: A1=80%,
A2=20%,
1=1 s, and
2=10 s. These settings were applied to all data sets, i.e. before and after lung damage, and gave a satisfactory goodness of fit (chi-squared=2.06 (0.75); mean (SD) for all fits performed). An example of a fit result is given in Figure 1.
|
Haemodynamics and gas exchange measurements
Before lung damage and after stabilization after repeated lung lavages, we recorded aortic and central venous pressures, and cardiac output. For assessment of gas exchange, arterial and mixed venous blood gases were measured.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
|
Deflation
A deflation process with a single fast compartment was present in all animals, with a time constant of 1.2 s (range 1.11.3 s). In only one of five animals a second, longer of 4.5 s was evident, which comprised 7% of the total ventilated lung area. In four of five healthy animals, the deflation process fulfilled the prospectively defined criterion of a fast one-compartmental behaviour.
Lavaged lungs
Inflation
Here, a short time constant of 0.5 s (range 0.50.6 s) was detected in 86% (range 7887%) of the lung area inflated by the manoeuvre. The second, longer of 9.1 s (range 816.8 s) was found in 14% (range 1322%) of the inflation area (see Fig. 4A).
|
Individual data for all animals are listed in Table 2.
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
In healthy lungs, the inflation and deflation processes are characterized best by a one-compartmental behaviour, whereas after lavage the co-existence of two compartments with clearly different was observed. A short inspiratory
was found for 86% of the total lung area ventilated or recruited by the manoeuvre. The remaining 14% of finally aerated lung area inflated quite slowly with a long
of about 10 s. In contrast, during expiration, 94% of the damaged lung collapsed very quickly with a
of less than 1 s, compared with slower deflation with a
of more than 1 s in the healthy lungs.
Radiological interpretation of the dynamic series of CT scans gave a more descriptive assessment of the temporal dynamics of lung aeration with ventilation and recruitment phenomena in ARDS.
During inflation, areas that were already filled expanded first, followed by a slow recruitment of atelectasis. Thus, the short in inspiration appeared to represent ventilation of already aerated alveoli, whereas the longer
reflects recruitment of atelectatic areas.
During deflation, rapid collapse of large portions of lung parenchyma, with a rapid increase of atelectatic and hypoventilated dense lung areas, causes the short expiratory . The long expiratory
in ARDS represents expiration from lung areas, which remain air-filled even at ZEEP.
In a similar animal model of ARDS, Wegenius and colleagues described a quantal behaviour of atelectatic lung detectable by CT. This behaviour is because surfactant-depleted alveoli are either fluid-filled and thus collapsed, or air-filled and open, but almost no intermediate states are present.6 This theory is supported by our observation that only after lung damage, i.e. after development of widespread expiratory alveolar collapse, two-compartmental behaviour appears. During inflation, the short reflects inflation of open alveoli (i.e. a ventilation process), and the longer
represents the slower opening of collapsed alveoli at their respective opening pressures, for example the recruitment process. During deflation, fast expiration of air is accompanied by the rapid formation of large atelectatic areas, which explains the large compartment with a short
of <1 s.
As we aimed to develop a diagnostic method to characterize compartments of lung aeration in ARDS, we chose a lavage-induced form of lung damage to reproduce the ventilation and recruitment processes after surfactant-depletion according to Wegenius and colleagues.6 The lavage model simulates only one aspect of ARDS, i.e. loss of surfactant.7 In our study, the appearance of the CT images was consistent with the well-described radiological criteria for healthy lungs during general anaesthesia, and for lungs in an early stage of ARDS, respectively.8 9 Images also closely resembled those published by Neumann and co-workers, who compared saline-lavage in pigs with ARDS induced by oleic acid and endotoxin administration, using dynamic CT.2
Various definitions of lung aeration, defined by densitometry of CT images, co-exist in the radiology literature. Neumann and co-workers used the increase or decrease of MLD to quantify the changing air content in serially acquired CT images.10 MLD reduces the entire distribution function of density values, for example the attenuation histogram, in the lung to a single-number descriptor. Therefore, it will reflect not only the change of aeration in a lung region of interest, but also varying contributions of intravascular and extravascular water content. Temporal dynamics of MLD are, therefore, determined not by the processes of aeration (or collapse) alone, but by the net temporal behaviour of all density domains or compartments of a region together. A decrease in MLD may thus represent effects as different as recruitment, ventilation, overdistension, or even barotrauma. Such processes are, however, clearly discernible if, within the spectrum of attenuations, a density range is selected which is known to specifically reflect the effects of interest. We wished to detect rapid changes in parenchymal aeration, for example recruitment and ventilation, with high sensitivity, and not to compare absolute gas content of healthy and diseased lungs. Therefore, we studied aeration changes by following the temporal behaviour of specific density windows, which we had tuned to our animal model and imaging parameters in a previous study.3 In the literature, a density range of 910 to 500 HU is used to detect aerated lung areas. In this previous study, airway pressure was increased and decreased in a stepwise manner (5 cm H2O every 5 s) during dynamic multiscan CT. In healthy and lavaged lungs, the fractional area change of attenuation ranges of increasing bandwidth (constant lower threshold of 910 HU, higher threshold increasing from 800 HU to 200 HU in steps of 100 HU) was determined planimetrically and compared. We found a density window from 910 HU to 700 HU for healthy lungs, and from 910 HU to 300 HU for lavage ARDS to be most sensitive to airway pressure-dependent changes in lung aeration.3 These response curves are shown in Figure 5. They are explained by the fact that after lavage a greater parenchymal water content causes ground-glass opacities which nevertheless respond strongly to aeration. Another way to assess aeration of lung tissue, which also avoids the shortcomings of regional analysis of MLD, is to calculate the total amount of air on a pixel-by-pixel basis as described by Puybasset and colleagues.11 This approach translates, in a strictly proportional fashion, x-ray attenuation (in Hounsfield Units) into fractional gas content, and calculates absolute FRC from this and from the volume of voxels. To obtain volumetric data that are meaningful to the clinician, this technique requires static thin-section CT acquisition of the entire lung volume. At present, however, fast dynamic assessment of multiple slices or even the entire lung volume is not yet technically possible. Once this is achieved, our technique will allow calculation of the compartmental distribution of the tidal volume and the pertinent time constants, similar to that of Puybasset and colleagues. However, as our study was restricted to a fast dynamic but two-dimensional analysis of only one large transverse lung slice, attenuation was not transformed further into gas content in that slice, particularly because relative compartmental size and time constants are independent of such a transformation.
Neumann and colleagues evaluated the temporal dynamics of lung inflation and collapse with a fast CT technique, assuming a priori a one-compartmental behaviour of the lung. They found that both recruitment and collapse of lungs damaged with saline lavage, oleic acid and endotoxin-induced ARDS occurred mainly during the first 4 s. In lavage-ARDS, a longer (
2.0 s) was observed in expiration than in inspiration (
0.7 s).2 Our results support these findings. These authors fitted their data to follow a mono-exponential relationship, and the contribution of our slow compartment may have been underestimated by their choice of an upper CPAP level of
40 instead of the 50 mbar we used. They concluded that in their experimental settings, expiratory times <0.6 s would be necessary to avoid cyclic alveolar collapse during mechanical ventilation, which is also supported by the findings of our study for the lavage-ARDS model.
In a related study of porcine oleic-acid-induced ARDS, Neumann and co-workers used several PEEP levels ranging from 10 to 25 cm H2O and an inspiratory peak pressure of 15 cm H2O above PEEP to determine the minimum PEEP level as well as the critical inspiratory and expiratory time to avoid lung collapse.11 They found that a PEEP level >20 cm H2O or an expiration time 0.6 s were required to largely prevent expiratory lung collapse. As the short deflation
1 in our ARDS series ranged between 0.4 and 1.0 s, we predict that reducing expiration time to
0.6 s would allow deflation of only about 60% of the fast compartment, causing FRC to increase, and would completely prevent collapse of the slow compartment (
2 >7 s). However, whereas titration of PEEP or expiration times takes time and requires repeated assessments by blood gas analysis or CT, the technique to derive compartmental time constants of lung aeration from one rectangular CPAP manoeuvre may reduce the time to optimize the ventilation pattern in an individual ARDS patient.
Recently, electrical impedance tomography (EIT) has been introduced as another image-based technique to quantify regional lung aeration. Kunst and colleagues showed, that the lower and upper inflection point can be determined with EIT in porcine lavage ARDS lungs.12 In contrast with dynamic CT, EIT can be used as a bedside technique without any radiation exposure. CT, on the other hand, allows much better temporal and spatial resolution. Further studies are necessary to compare these methods and their potential roles for clinical decision-making.
A current limitation of our technique is that the resolution of the CT data is only achieved in one cross-sectional slice of the lung. We selected a slice between apex of the heart and diaphragm, which allows a fair approximation of total ventilated lung area.2 13 To reduce consecutive errors in planimetry and densitometry, the cross-sectional slice for image acquisition was determined by a scout view and a reference scan before each experiment. This slice definition was important to ensure that in every CT image only the lung parenchyma between the apex of the heart and the diaphragm was measured. Also, as the scanner table is immobile during the dynamic multiscan acquisition, this slice cannot be followed during inspiration or expiration although the lung area of interest moves slightly over time with the cranio-caudal respiratory motion of the lung. This may explain the occasional, counterintuitive observation in our series as well as in others,2 that ventral lung regions actually appear to become smaller again during late inspiration (compare Fig. 3B and C), or that density in ventral regions decreases during expiration. Such effects could be caused by respiratory motion of the (preferentially ventilated) subcardial accessory lobe of the pig lung. Deviation from a two-compartmental behaviour, i.e. existence of a third compartment, is another possibility yet to be studied. In the near future, multislice CT scanners will allow simultaneous acquisition of several slices. These image series, acquired at different cranio-caudal levels, may also allow dynamic lung volume changes to be calculated during ventilation.
The temporal resolution of the CT scanner was limited to 250 ms in this study. This resolution allows 6 data points to be obtained during 1.5 s; for example, a of 0.5 s is the minimum to be determined by least-squares fitting. Scanning at even better temporal resolution would be necessary if
<0.5 s are expected. Fast dynamic CT acquisitions with an effective temporal resolution up to 100 ms are already available, and further improvement may come soon. An even better temporal resolution of about 50 ms is already obtainable by electron beam scanners (EBCT),14 but this technology is not widely available for clinical use, and causes increased radiation exposure compared with spiral CT scanners.15
Even dynamic CT uses a lot of radiation. Although dosimetry was not performed in this study, preliminary data on humans show a dose of 47 mGy for a 10 s dynamic CT measurement, using the same scanner and similar acquisition parameters,16 compared with a dose of approximately 3040 mGy when undergoing a spiral CT of the thorax.
Clinical consequences
Different respiratory time constants and their individual contributions to lung ventilation and alveolar recruitment can be found by dynamic CT acquisition using clinical spiral CT scanners. Such analyses may help to rapidly and individually optimize ventilator settings in ARDS patients, avoid cyclic alveolar collapse and reopening, maximize ventilated lung, and reduce respirator-induced lung injury. Dynamic CT techniques open a new diagnostic field, providing a regional analysis of lung function with correlation to the underlying pathology. Further studies are necessary to establish the clinical value of this method.
![]() |
Acknowledgements |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 Neumann P, Berglund JE, Fernandez Mondejar E, Magnusson A, Hedenstierna G. Dynamics of lung collapse and recruitment during prolonged breathing in porcine lung injury. J Appl Physiol 1998; 85: 153343
3 Markstaller K, Kauczor H-U, Eberle B, et al. Dynamic multiscan CT during artificial respiration: comparison of different density ranges in healthy lungs and in a lavage ARDS model. Fortschr Roentgenstr 1999; 170: 57580
4 Lachmann B, Robertson B, Vogel J. In vivo lung lavage as an experimental model of respiratory distress syndrome. Acta Anaesthesiol Scand 1980; 24: 2316[ISI][Medline]
5 Nunn JF. In: Heinemann B, ed. Nunns Applied Respiratory Physiology. Oxford, 1993; 58693
6 Wegenius GA, Sjöstrand UH, Svensson BA, Lichtwrack-Aschoff M, Markström AM, Nordgren AK. Radiological assessment of surfactant depletion injury. Appl Cardiopulm Pathophysiol 1996; 6: 11929
7 Rosenthal C, Caronia C, Quinn C, Lugo N, Sagy M. A comparison among animal models of acute lung injury. Crit Care Med 1998; 26: 9126[ISI][Medline]
8 Reber A, Engberg G, Sporre B, et al. Volumetric analysis of aeration in the lungs during general anaesthesia. Br J Anaesth 1996; 76: 7606
9 Wegenius G. Radiology in the diagnosis of Adult Respiratory Distress Syndrome with special reference to pulmonary oedema its distribution and detection. Acta Radiol Suppl Stockh 1988; 372: 5372[Medline]
10 Neumann P, Berglund JE, Mondejar EF, Magnusson A, Hedenstierna G. Effect of different pressure levels on the dynamics of lung collapse and recruitment in oleic-acid-induced lung injury. Am J Respir Crit Care Med 1998; 158: 163643
11 Puybasset L, Cluzel P, Gusman P, Grenier P, Preteux P, Rouby J-J, CT Scan ARDS Study Group. Regional distribution of gas and tissue in acute respiratory distress syndrome. I. Consequences for lung morphology. Int Care Med 2000; 26: 857869[ISI][Medline]
12 Kunst PW, Böhm SH, Vazquez-de-Anda G, et al. Regional pressure volume curves by electrical impedance tomography in a model of acute lung injury. Crit Care Med 2000; 28: 17883[ISI][Medline]
13 Gattinoni L, Pelosi P, Vitale G, Pesenti A, DAndrea L, Macheroni D. Body position changes redistribute lung computed tomographic density patterns in patients with acute respiratory failure. Anesthesiology 1991; 74: 1523[ISI][Medline]
14 Rogalla P, Enzweiler C, Schmidt E, Taupitz M, Bender A, Hamm B. Thoracic diagnostics with electron beam tomography. Radiologe 1998; 38: 102935[ISI][Medline]
15 Becker C, Schatzl M, Feist H, et al. Assessment of the effective dose for routine protocols in conventional CT, electron beam CT and coronary angiography. Fortschr Roentgenstr 1999; 170: 99104
16 Heussel CP, Hafner B, Thelen M, Kauczor H-U. Computed Tomography of the trachea: paired inspiratory/expiratory spiral-CT and cine-CT during respiration. Eur Radiol 2000; 10: 273