Introduction

In acute lung injury (ALI) and in the acute respiratory distress syndrome (ARDS), severe impairment of oxygenation is caused by an inhomogenous ventilation-perfusion \({\left( {{\mathop {\text{V}}\limits^ \cdot }_{{\text{A}}} /{\mathop {\text{Q}}\limits^ \cdot }} \right)}\) distribution and an increase in shunt fraction resulting from atelectasis formation. The amount of aerated lung is markedly reduced due to alveolar collapse and flooding [1, 2], which is found predominately in the dependent parts of the lungs [1, 3]. Mechanical ventilation has been shown to further aggravate the \({{\mathop {\text{V}}\limits^ \cdot }_{{\text{A}}} /{\mathop {\text{Q}}\limits^ \cdot }}\) mismatch [4]. One of the commonly used treatment concepts in such situations is to open up atelectasis, aiming at the recruitment and maintenance of ventilated lung volume to improve oxygenation [5].

Although significant controversy exists over the optimal maneuver to achieve alveolar recruitment, no gold standard exists for how to measure the success of such a maneuver. Computed tomography (CT) has been used to detect recruitment [6] with the distributions of the CT values as measured in Hounsfield units (HU) assigned to lung tissue ventilation [7]. In these early studies, only single slices had been analyzed, and the results were extrapolated to the whole lung. This had been demonstrated to be unreliable due to the uneven distribution of aerated and nonaerated lung from apical to basal [8]. Therefore, images of the whole lung by multiple exposures have been evaluated for this purpose [9–11]. Recruitment was assessed by the increase of intrathoracic gas and by histograms showing CT-value distributions. However, long exposure times with a single slice CT scanner and gross calculation of differently aerated lung volumes have prevented subtle analysis in the quantitative distribution of ventilation volumes for transition into clinical practice [9]. The aim of this study was to investigate whether the effects of a standard recruitment maneuver on the volumes of differently aerated lung parenchyma in acute lung injury could be determined more easily and faster by use of a multi-slice CT scanner in static conditions of acute lung injury.

Materials and methods

After governmental approval, eight anesthetized and paralyzed female pigs (31.3±1.9 kg) were orotracheally intubated and ventilated in constant flow mode with an inspiratory oxygen concentration (FiO2) of 100%, a tidal volume of 8 ml/kg and a positive end-expiratory pressure (PEEP) of 10 cmH2O throughout the study. Handling of animals conferred to the guidelines laid out in the Guide for the Care and Use of Laboratory Animals [12]. Anesthesia was maintained with continuous infusions of propofol (Fresenius Kabi, Bad Homburg, Germany), fentanyl (Janssen-Cilag, Neuss, Germany) and pancuronium (Schwabe-Curamed, Karlsruhe, Germany) and a continuous infusion of Ringer's solution maintained at 5 ml/kg h−1.

Gas flow and airway pressure were measured between the tracheal tube and the Y-piece of the ventilator circuit by a membrane differential pressure transducer, connected to the ventilator (Galileo Gold, Hamilton Medical, Rhäzüns, CH, accuracy ±5%). Tidal volume (VT) was derived from the integrated flow signal. Expiratory volumes were corrected as proposed by Jonson et al. [13].

Experimental protocol

Acute lung injury was induced through repeated lavages with saline warmed to 38°C and allowed to stabilize until the PaO2 was constantly below 13.3 kPa for 60 min [14]. Afterwards, the animals were placed supine on the CT table and MSCT was performed during end-expiratory and end-inspiratory breath-holds of approximately 5.5 s [1, 15]. Between scans, tidal ventilation was resumed for 30 s. Then the actual recruitment maneuver was performed by inflating the lungs three times with a pressure of 45 cmH2O for 40 s [7, 16–18] with ten normal tidal breaths in between inflations. Fifteen minutes after the recruitment maneuver, a second set of MSCT scans was obtained. After the experiment, the animals were killed with a barbiturate overdose.

CT measurements

MSCT of the whole lung was performed with a 16-slice CT scanner (SOMATOM Sensation 16 Siemens, Forchheim, Germany). Scan parameters were chosen as follows: 16×1.5-mm collimation, 120-kV tube voltage with an effective tube current-time product of 100 mAseff., a gantry rotation time of 500 ms and 36-mm table feed per rotation. For image reconstruction, an effective slice thickness of 2 mm and a reconstruction increment of 1 mm were used. An optimized sharp kernel (B50f) and a field-of-view of 280×280 cm2 with a 512×12 matrix were applied. All images were assessed at a window width of 1,200 HU and a window center of −600 HU. For analysis of MSCT images, a commercially available software was used (Pulmo Option, Syngo, Siemens). First the lung parenchyma was automatically delineated at the inner pleura and then corrected by hand by an experienced reader (DH). CT values within a range of −1,000 to +100 HU were assigned to lung parenchyma, representing the range from air only to totally de-aerated lung tissue. The fractions of density distributions were automatically calculated by the CT software in steps of 100 HU. From this data, the area covered by each HU interval per slice was calculated. The values of the different lung volumes were calculated by multiplying the resulting areas with the slice thickness. Lung parenchyma was characterized as hyperinflated (VHYP) from −1,000 to −900 HU, as normally aerated (VNORM) from −900 to −500 HU, as poorly aerated (VPOOR) from −500 to −100 HU and as non-aereated (VNON) from −100 to +100 HU [15]. The gas content was calculated as:

\(V_{{GAS}} = \frac{{V \times HU_{{MEAN}} }}{{ - 1000}}\) and the tissue mass was calculated as: VTISS=VTOT−VGAS. The lung volumes consisted of VGAS+VTISS, e.g., a mean HU of −500 representing 50% gas and 50% tissue [15]. All volumes were calculated for end-inspiratory and end-expiratory scans separately. The tidal distribution was calculated by subtracting end-expiratory from end-inspiratory values.

For each slice from apical to basal, the total cross-sectional area and the mean area taken by the particular increments in HU were calculated. To compare the results with those obtained by just single slice analysis, the same procedure was applied to the single slice cranial to the diaphragma with the greatest cross-sectional area in the expiratory exposure before the recruitment maneuver. The slices at the same position were then analyzed for the other exposures also. Instead of volumes, data are given as fractions of the total cross-sectional area.

All data are reported as mean±standard deviation (SD). “Before” and “after” recruitment maneuver values were compared using Wilcoxon's signed ranks test. Recruitment was defined as the increase in aerated and the decrease in non-aerated volumes after the recruitment maneuver. Statistical significance was accepted at P<0.05 (SPSS 11.0, Chicago, IL).

Results

Comparison of breath-hold volumes before and after the recruitment maneuver

During end-inspiratory and end-expiratory breath hold, there was a significant increase in the poorly aerated lung volumes (VPOOR) and a decrease in nonaerated volumes (VNON) (Fig. 1). This was associated with an increase in the PaO2 from 9.5±2.7 to 12.6±3.8 kPa (P=0.017). Total end-inspiratory lung volume increased from 1,589±139 ml before to 1,713±150 ml after the recruitment maneuver (P=0.01), while VGAS increased from 838±84 ml before to 939±128 ml after the recruitment maneuver (P=0.012). The distribution of this gas went preferentially to poorly aerated and normally aerated lung, which was more pronounced in the regions of −700 to −500 HU (Fig. 2). No significant changes in the better ventilated regions of VNORM from −900 to −700 HU were observed. VTISS did not change (819±56 ml before vs 838±64 ml after, P=0.263).

Fig. 1
figure 1

One exemplary slice near the diaphragma of a pig with saline lavage-induced acute lung injury. Upper panel (a) before the recruitment maneuver and lower panel (b) after the recruitment maneuver. I: inspiratory breath hold; E: expiratory breath hold

Full size image
Fig. 2
figure 2

Volumes for each 100-HU compartment from −1,000 to +100 HU. Upper panel: values at inspiratory hold; lower panel: values at expiratory hold. denotes P<0.05 for comparison between “before” and “after” the recruitment maneuver. Boxplot: parameter median, interquartile range and standard deviation

Full size image

Comparison of inspiratory-expiratory differences, resembling tidal ventilation

The distribution of tidal ventilation to differently aerated regions is shown in Fig. 3. No significant differences were found for tidal volumes measured by spirometry (230±16 ml) or determined by MSCT (209±52 ml, P=0.10). Tidal volume was not significantly increased after the recruitment maneuver, which can be expected in volume-controlled ventilation, but inspiratory plateau pressure decreased from 31±3 cmH2O to 27±3 cmH2O (P=0.03). For other ventilatory parameters, see Table 1. A significant portion of the tidal breath is distributed to nonaerated and poorly aerated parenchyma, representing expiratory collapse and inspiratory re-opening of the alveoli. The inspiratory-expiratory volume change in all non-aereated regions (62±18 ml) was not significantly reduced after the recruitment maneuver (43±26 ml, P=0.114). On the other hand, the distribution of ventilation to normally aerated regions appeared to be increased, but did not reach statistical significance (P=0.102).

Fig. 3
figure 3

Distributions of differently aerated lung volumes [in ml] during a tidal breath before (left side) and after (right side) a recruitment maneuver, obtained by inspiratory-expiratory differences. Positive values represent an increase, negative values a decrease of the respective volume during inspiration. The shown data represent the absolute values for hyperinflated, normally, poorly and non-aerated lung volumes. The range from −400 to −300 HU is not shown due to very small values (−1±7 ml before and 2±17 ml after recruitment). Data presented as mean±SD

Full size image
Table 1 Ventilation parameters
Full size table

Comparison of single slice with whole lung MSCT analysis

Great differences in the cross-sectional area and the distributions of HU can be observed from apical to basal (Fig. 4). In the analysis of a single slice just cranial to the diaphragma, the recruitment maneuver resulted in a decrease in non-aerated lung and an increase in poorly and normally aerated lung (Fig. 5). The percent change correlation was highly significant (r=0.96, P=0.001) compared to those observed with whole lung analysis, but significantly underestimated the increase in normally and poorly aerated lung (Table 2).

Fig. 4
figure 4

Cross-sectional area of each slice from apical to basal and the parts of it taken by nonaerated, poorly aerated, normally aerated and hyperinflated lung. a Before recruitment, expiration; b inspiration; c after recruitment, expiration; d inspiration

Full size image
Fig. 5
figure 5

Distribution of differently aerated lung (a) before recruitment and (b) after recruitment

Full size image
Table 2 Percent changes in lung volumes (whole lung analysis) or cross-sectional area (single slice) of differently aerated lung after a recruitment maneuver
Full size table

Discussion

In a porcine model of saline lavage-induced acute lung injury, we calculated the volumes of differently aerated lung parenchyma by attenuations of HU. Thereby, it is possible to observe the differential distribution of ventilation induced by a lung recruitment maneuver (Fig. 2). Mainly, the intrathoracic gas volume increased, and there was a shift from non-aerated to poorly and normally aerated lung volumes. The broad availability and improved performance of MSCT scanners have increased the acceptance of lung CT scanning in the treatment of ARDS [19] and have led to routine use in specialized centers.

To investigate whether changes in the amount of aerated and non-aerated lung volumes could be assessed quantitatively, we chose a model of acute lung injury that is known to be easily recruitable. Furthermore, the preferential distribution of atelectasis to the dependent lung could also be demonstrated in patients with ARDS by use of CT scanning [15]. Even though the general limitations of transferring experimental animal results into human pathology have to be applied, the pig lavage model has been used for several investigations [11, 20, 21]. To investigate the effects on the whole lung, we employed static conditions, even though dynamic studies of tidal ventilation have shown that a dynamic component of tidal recruitment might not be adequately reflected during breath-hold exposures of inspiration and expiration [21, 22]. However, to our knowledge, no software is available to analyze dynamic studies of the whole lung.

The understanding of the pathology of the ARDS has improved considerably since the introduction of CT [1]. The characterization of the disease severity was mainly done by looking at different gray scaling densities interpreted as infiltrates, consolidation or atelectasis during the acute phase [15] and after the resolution of ARDS [23]. The measurement of lung and intrathoracic gas volumes by CT has been validated previously [24, 25], but has not been used for as a routine procedure due to complicated and time-consuming exposure and analysis techniques. More recently, describing changes within the lung induced by certain interventions by the distribution of frequency distributions was attempted [7] with regard to the distribution of ventilation. Also, the lavage model in pigs has been used to describe changes seen in CT [22, 26–28], but mostly single slice analysis was used. The use of MSCT has enabled looking at the lung as a whole [29], acknowledging that interventions might induce changes in lung aeration at the apex differently from the basal lung regions [9]. In a study by Malbouisson et al. [8], a single slice CT scanner was used requiring an exposure time of >15 s. This, however, can interfere with ventilation and induce additional atelectasis formation in ARDS patients with highly unstable lungs [30]. In addition, recruitment was defined as the increase of intrathoracic gas, which could have also resulted from overdistended alveoli within the area marked as poorly or non-aerated lung. In contrast to previous investigations, we calculated actual volume shifts within regions of differently aerated lung parenchyma. As a consequence, we could demonstrate that the main mechanism of a standard recruitment maneuver is not an increase in normally aerated lung, but mainly an increase in poorly aerated lung, representing unstable units with low ventilation/perfusion distribution. Since we observed a significant improvement in oxygenation, the dependency of gas exchange on the amount of normally aerated lung volume has to be questioned in severe hypoxemic lung injury. Possibly poorly aerated regions with low ventilation-perfusion distribution and shunt perfusion are the main determinant of the relatively small changes in oxygenation, depending on regional perfusion alterations [31]. Since the interpretation of these results depends on the definition of the differently aerated lung compartments, the assignment of different density ranges to normally, poorly and non-aerated lung might lead to other results [22]. However, we have demonstrated that the main increases in aerated lung were in the range of −500 to −200 HU, whether this is called still normally or already poorly aerated. On the other hand, hardly any changes could be observed in the clearly well aerated ranges from −900 to −700, representing normally aerated lung.

Most importantly, though the amount of intrathoracic gas (and thereby the functional residual capacity) increased, the tissue amount [11] (also representing lung edema) and the distribution of the tidal breath to the differently aerated regions was unchanged. One of the known mechanisms of potentially harmful ventilation is the repeated collapse and re-opening of alveoli during tidal breath [32], characterized as tidal ventilation occurring in non-aerated and poorly aerated lung regions. On the other hand, the hyperinflated lung volume has been stable. This observation is important since a recruitment maneuver had been suspected to lead to hyperinflation and overdistension of certain areas of the lung. This study provides data that demonstrate that this is not true in a lavage model of ARDS, at least in volume controlled ventilation. With the pressure-controlled mode, the changes after the recruitment maneuver would have been different, with a significant increase in tidal volumes. Instead, we have observed a decrease in inspiratory distending pressure, which will further reduce the potentially injurious mechanisms of artificial ventilation. The question whether volume or pressure controlled ventilation is best for optimal recruitment is still unsolved [14]; therefore, both modes should be regarded as equal from this point of view. However, we have shown the importance of analyzing the whole lung, since the amount of recruited lung has been underestimated from the analysis of a single slice near the diaphragm. In this area the atelectasis commonly is found to be at its greatest, and the lung is additionally compressed by a superimposed vertical gradient and increased abdominal pressure [16]. To recruit these areas, extremely high pressures might be necessary, leading to overstretching of the non-atelectatic lung, which, however, was not observed in our study with an inspiratory plateau pressure around 30 cmH2O. Dynamic investigations have been used to characterize the effects of ventilation modalities on the distribution ventilation within the lung [28, 33], and in the future these could aid in the adjustment of ventilator settings.

In conclusion, we have shown that quantitative analysis of the whole lung by use of a MSCT scan is able to provide important insights into the consequences of a therapeutic intervention, e.g., a standard recruitment maneuver, as compared to a single slice analysis. Regions especially prone to atelectasis, e.g., the dependent lung, may be investigated in further studies with this technique. Further developments towards a reduction of the radiation dose [34] and improved analysis software could promote MSCT as a routine investigation in acute lung injury and ARDS.