INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In patients with acute lung injury (ALI), it has recently been suggested that positive end-expiratory pressure (PEEP)-induced alveolar recruitment can be associated with some degree of lung overdistension (1). Computed tomographic (CT) scan allows a precise determination of lung volumes by the frequency histogram analysis function (2). Up to now, lung zones with a density between 1,000 and 500 Hounsfield units (HU) have been considered as normally aerated, those between 500 and 100 HU as poorly aerated, and those between 100 and +100 HU as nonaerated (3). However, the density threshold separating lung overdistension from normal lung aeration has not yet been firmly established. Except for the paper by Dambrosio and coworkers (1), no attempt has been made to quantify lung overdistension although its assessment could be of critical importance when comparing different ventilatory strategies in patients with ALI. The first goal of this study was to assess density histogram distribution in healthy volunteers at functional residual capacity (FRC) and at total lung capacity (TLC) in order to determine the density threshold allowing one to separate normally aerated from overdistended lung. Once determined in healthy volunteers, this threshold was used in ventilated patients with ALI in order to separate PEEP-induced alveolar recruitment from PEEP-induced lung overdistension.

In patients undergoing thoracic CT scan the intravenous administration of contrast material is usually recommended for delineating nonaerated lung parenchyma from pleural effusion. One can hypothesize that contrast material increases the density of lung parenchyma and induces a shift to the right of the density histogram distribution resulting in an overestimation of nonaerated regions of the lung. In the present study the influence of contrast material on density histogram distribution was also assessed.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Healthy Volunteers and Patients with ALI

Six authors of the present study (S.V., J.R., Q.L., Ph.C., P.G., and J.J.R., age = 37 ± 10 yr, 4 males and 2 females) underwent a spiral thoracic CT scan. All volunteers were nonsmokers and had no history of previous lung disease.

A thoracic CT scan was also performed in six patients (age = 55 ± 18 yr, 4 males and 2 females) with ALI defined as a Pa_O2 below 300 mm Hg at fraction of inspired oxygen (FI_O2) 1 and zero end-expiratory pressure (ZEEP) and bilateral hyperdensities on the bedside chest X-ray (12). The protocol was considered as a part of routine clinical practice and no informed consent was obtained from patients' next of kin. Five patients were admitted to the surgical intensive care unit of La Pitié-Salpêtrière Hospital (Department of Anesthesiology) after major vascular (n = 2), neurologic (n = 1), and orthopedic (n = 2) surgery, whereas one patient was admitted for an acute medical illness. ALI was secondary to bronchopneumonia (n = 2), pulmonary contusion (n = 2), and aspiration (n = 2). The mean lung injury severity score was 2.4 ± 0.7 (13) and the mean quasi-static respiratory compliance, calculated as tidal volume divided by inspiratory plateau pressure minus auto-PEEP, 47 ± 17 ml/cm H₂O. All were deeply sedated with a continuous intravenous infusion of fentanyl 250 µg/h, flunitrazepam 1 mg/h and paralyzed by vecuronium 4 mg/h. They were ventilated using volume-controlled mechanical ventilation by means of a César ventilator (Taema, Antony, France). PEEP was individualized in each patient and set 2 cm H₂O above the lower inflection point of the pressure-volume (P-V) curve determined as follows. During volume-controlled mechanical ventilation, the inspiratory/expiratory (I/E) ratio was set at 80%, the respiratory frequency at 5 breaths/min, and the tidal volume (VT) at 1,500 ml. Using these ventilatory settings, a constant inspiratory flow of 9 L/min was administered to the patient during 9.6 s, generating a P-V curve on the screen of the ventilator. A cursor present on the screen was used to determine the lower inflection point of the P-V curve. In a preliminary study, we verified that this method of measuring compliance was equivalent to the method of reference, the gross syringe method. As a mean, a PEEP of 13 ± 3 cm H₂O was applied to the six patients with ALI.

Thoracic CT Scan Procedure

Lung scanning was performed from the apex to the diaphragm using a Tomoscan SR 7000 (Philips, Eindhoven, The Netherlands). The exposures were taken at 120 kV and 250 mA. All images were observed and photographed at a window width of 1,600 HU and a window level of 700 HU. Contiguous axial sections of 10-mm thickness were reconstructed from the volumetric data.

In the healthy volunteers, CT scan was performed at FRC and TLC both before and after injection of 80 ml of contrast material (Omnipaque; Nycomed, Paris, France). FRC was considered as the lung volume at the end of a passive expiration after a period of quiet breathing. TLC was obtained by asking each healthy volunteer to reach his or her maximal inspiratory lung volume and to perform an active expiratory effort against a 30-cm-high column of water in order to maintain a positive alveolar pressure of 30 cm H₂O. Apnea was maintained during CT scan acquisition (15 to 20 s). Great care was taken to avoid any gas leak during the acquisition at TLC.

In patients with ALI, CT scan images were acquired at FRC using ZEEP and a PEEP of 13 ± 3 cm H₂O, obtained after clamping the expiratory circuit at end-expiration. Airway pressure was continuously monitored during the CT scan acquisition in PEEP to ensure that the preset value was effectively applied. Contrast material (80 ml) was intravenously injected before the CT scan acquisition in order to differentiate pleural effusion from consolidated lung parenchyma. For transportation and during the CT scan procedure, the patients were accompanied by two intensivists. Mechanical ventilation was provided using an Osiris ventilator (Taema, Antony, France). A Propaq 104 EL monitor (Protocol System, North Chicago, IL) allowed the continuous monitoring of electrocardiogram (ECG), airway pressure, and hemoglobin saturation.

Lung volumes and density histograms were quantified by a method previously described and validated (2). Briefly, the radiologist manually traced the right and left lung outlines with the roller ball on each spiral CT section being unaware of the lung volume status. Lung areas and mean lung density values were determined by using the region of interest function. Frequency histograms of the densities in HU were subsequently generated for each region of interest by using the analysis function. The frequency distribution of CT numbers of the entire lung was computed for 50 compartments, from 1,000 HU to +100 HU, examining a 22-HU segment for each compartment. The lung volume of each compartment was calculated by multiplying the following: number of lung pixels times square pixel size times section thickness. Total lung volume was obtained by adding the lung volume of each compartment. Lung zones with a density below 500 HU were considered as normally aerated, those between 500 and 100 HU as poorly aerated, and those between 100 and +100 HU as nonaerated (5).

Hemodynamic and Respiratory Measurements

All patients had in place an arterial and a fiberoptic thermodilution Swan-Ganz catheter (CCO/Sv_O2/VIPTD catheter; Baxter Healthcare Corporation, Irvine, CA) for cardiovascular monitoring. Arterial pressure, ECG, and cardiac filling pressures were recorded at a high sample rate of 100 Hz on the MP 100 WS data acquisition and analysis system (Biopac System Inc., Goleta, CA) and Macintosh personal computer (Apple Computer Inc., Cupertino, CA) connected to the analog port of the hemodynamic monitor Merlin (Hewlett Packard, Palo Alto, CA). Cardiac output was measured using the thermodilution technique with simultaneous withdrawing of systemic and pulmonary arterial blood samples. Arterial oxygen tension (Pa_O2), mixed venous oxygen pressure (P_O2), and pH were measured using a conventional analyzer whereas hemoglobin concentration, arterial and mixed venous oxygen saturations were measured using an OSM3 hemoximeter (Radiometer Copenhagen, Neuilly-Plaisance, France). Standard formulas were used to calculate cardiac index, pulmonary shunt, oxygen delivery, and oxygen consumption. In each patient, expired CO₂ was measured with a nonaspirative 47210A infrared capnometer (Hewlett-Packard, Andover, MA) positioned between the endotracheal tube and the Y-piece of the ventilator. Expiratory CO₂ curves were recorded on the MP 100 WS data acquisition and analysis system. After simultaneous withdrawing of an arterial blood sample, the ratio of alveolar dead space (VDA) to VT was calculated according to the equation: VDA/VT = 1  PET_CO2/Pa_CO2, where PET_CO2 is end-tidal CO₂ measured at the plateau of expiratory CO₂ curve.

Statistical Analysis

The results are expressed as mean ± SD in the text and tables and as mean ± SEM in the figures. The parameters derived from the CT scan analysis of healthy volunteers at FRC and at TLC, before and after injection, were compared by a two-way analysis of variance for 1 within factor (FRC and TLC) and 1 between factor (presence or absence of contrast material). A Student paired t test was used to compare patients' parameters. Level of significance was considered as 5%.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Density Histogram Distribution at FRC in Healthy Volunteers and in Patients with ALI

Table 1 summarizes the lung volumes of the six healthy volunteers computed from the CT scan analysis. At FRC, 92 ± 3% of the overall lung parenchyma was normally aerated, 7 ± 3% ws poorly aerated, and less than 1% was nonaerated. The distribution of density histograms was monophasic and non-Gaussian with a peak located at 791 ± 12 HU and a mean CT number value of 691 ± 54 HU. Figure 1 shows the individual density histograms of the six healthy volunteers. Virtually no lung parenchyma had a CT number below 900 HU. In two healthy volunteers, 1.1% and 0.7% of the overall lung volume corresponding to 50 and 24 ml of lung parenchyma were characterized by CT numbers ranging between 900 and 912 HU. In the four other volunteers, lung parenchyma had a density always above 900 HU.

View larger version (30K): [in this window] [in a new window]   Figure 1.   Individual frequency histogram distribution of the six healthy volunteers at FRC (open circles) and TLC (closed circles) with contrast material. Only the initial part of the density histogram (between 1,000 and 400 HU) is represented. The dashed line indicates the density threshold for lung overdistension.

Table 2 summarizes the lung volumes of the six patients with ALI computed from the CT scan analysis. At FRC, 35 ± 29% of the overall lung parenchyma was normally aerated, 33 ± 19% was poorly aerated, and 32 ± 18% was nonaerated. As shown in Figure 2, the distribution of the density histograms was biphasic in three patients with a first peak located at 773 ± 34 HU and a second peak located at 5 ± 13 HU. In the remaining three patients, the distribution of the density histogram was monophasic and non-Gaussian characterized by a progressive increase in the volume of the lung along the Hounsfield unit scale with a peak located at 3 ± 25 HU.

View larger version (32K): [in this window] [in a new window]   Figure 2.   Individual frequency histogram distribution of the six patients at FRC (open circles) and a PEEP value of 13 cm H2O (closed circles). The entire HU scale is represented. The dashed line indicates the density threshold for lung overdistension.

Density Histogram Distribution at TLC in Healthy Volunteers

In healthy volunteers, overall lung volume increased by 79 ± 35% at TLC and 30 cm H₂O of alveolar pressure whereas peak CT number decreased to 886 ± 26 HU. The large majority of the increase in lung volume (74 ± 30%) was characterized by CT numbers below  900 HU. Because at FRC more than 99% of the lung parenchyma was characterized by CT numbers above 900 HU, this value can be reasonably considered as the density threshold separating normally aerated lung parenchyma from lung overdistension.

Density Histograms at PEEP in Patients with ALI

In patients with ALI, overall lung volume increased by 47 ± 19% at PEEP 13 cm H₂O whereas mean CT number decreased to 538 ± 171 HU. PEEP induced a mean alveolar recruitment, defined as a reduction of the volume of lung areas characterized by densities ranging between 100 and +100 HU, of 320 ± 196 ml. PEEP-induced alveolar recruitment was observed in each individual. PEEP also increased the volume of normally aerated lung areas by 1,191 ± 486 ml. According to the previously described threshold of 900 HU, a mean overdistension of 238 ± 320 ml was associated with PEEP-induced alveolar recruitment. In fact, this phenomenon occurred only in three patients (Figure 2). In ZEEP conditions, each of these three patients demonstrated a biphasic distribution of density histograms suggesting that PEEP-induced lung overdistension mainly occurs when a large amount of normally aerated lung coexists with a large amount of nonaerated lung before PEEP implementation.

Influence of the Injection of Contrast Material

In healthy volunteers, contrast material had no significant effect on density histograms distribution (Figure 3), on normally aerated, poorly aerated, and nonaerated lung volumes (Table 1), and on mean CT number of the overall lung parenchyma (687 ± 55 HU at FRC and 809 ± 38 HU at TLC after contrast material versus 691 ± 54 HU and 824 ± 37 HU without contrast material).

View larger version (14K): [in this window] [in a new window]   Figure 3.   Frequency histogram distribution along the HU scale (between 1,000 and +100 HU) of the six healthy volunteers at FRC without (open squares) and with (open circles) contrast material and at TLC without (closed squares) and with (closed circles) contrast material.

Cardiorespiratory Effects of PEEP

Table 3 summarizes the hemodynamic and respiratory effects of PEEP in patients with ALI. There was a significant increase in Pa_O2 and a significant decrease in pulmonary shunt, in agreement with the reduction of nonaerated lung volumes, related to PEEP-induced alveolar recruitment. No significant change was observed in the other parameters.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main result of this study is that, in healthy volunteers, the density threshold separating normally aerated lung parenchyma from lung overdistension is around 900 HU. Using this density threshold in six patients with ALI, we could demonstrate that PEEP-induced alveolar recruitment was accompanied by a certain degree of lung overdistension in three of them.

CT Scan Assessment of Lung Overdistension

A thoracic CT scan allows an accurate measurement of pulmonary volume and density of the lung parenchyma. By combining both measurements, the volume of normally aerated, poorly aerated, nonaerated, and overdistended lung regions can be determined giving the possibility of assessing the effects of a given ventilatory mode on pulmonary morphology.

In fact, two different approaches can be used for measuring lung volume of these different compartments. The first one that was developed by Gattinoni and coworkers (3) is based on the assumption that the mean CT numberor the mean densityof a given lung volume correlates with the respective proportion of gas and tissue within this lung. The CT number characterizing each individual pixel is expressed in HU and defined as the attenuation coefficient of the X-ray by the material being studied minus the attenuation coefficient of water divided by the attenuation coefficient of water. By reference, the CT number of water is 0 HU. The CT number is scaled by a factor of 1,000 and the CT number of air is 1,000 HU. A volume of lung with a mean CT number of 500 HU is considered as being composed of 50% gas and 50% tissue. Using this analysis, it is possible to compute the volume of gas and tissue and the gas/tissue ratio. One limitation of this approach is that it divides the lung into two compartments (gas and tissue) and does not take into account the fact that aerated parts of the lung can be normally aerated, poorly aerated, or overdistended.

Another approach is to measure lung volumes by analysis of the distribution of density histograms (2). This technique is based on the calculation of the volume of one pixel, called a voxel. It can be calculated that, for a field of 35 times 35 cm², a section thickness of 1 cm and a zoom of 1, a voxel has a volume of 0.0047 ml. An appropriate correction factor must be applied if a different zoom is used for acquisition of the images. The overall lung volume is calculated as the total number of pixels times the volume of one voxel. If the lung is divided into different compartments defined by different ranges of CT numbers, the number of voxels included within the boundaries of each compartment gives its volume. According to Gattinoni and coworkers (5), three different lung compartments are classically considered: normally aerated lung characterized by CT numbers ranging between 1,000 and 500 HU, poorly aerated lung characterized by CT numbers ranging between 500 and 100 HU, and nonaerated lung characterized by CT numbers above 100 HU. These definitions were derived from the distribution of CT numbers obtained from three 9-mm-thick CT sections in eight healthy subjects (5). The present study performed on the whole lung confirms that the density threshold selected for differentiating normally aerated from poorly aerated lung parenchyma is valid: at TLC with an alveolar pressure of 30 cm H₂O, less than 2% of the healthy volunteers' lung parenchyma was characterized by CT numbers above 500 HU.

The method that was used in the present study for determining the density threshold characterizing lung overdistension was based on two assumptions: (1) in healthy volunteers who are nonsmokers there are no overdistended lung regions at FRC, (2) lung overdistension is present at TLC when a pressure of 30 cm H₂O is applied to the alveolar space (14). As shown in Figure 2, the density threshold characterizing lung overdistension is likely around 900 HU: at FRC, more than 99% of healthy volunteers' lung parenchyma is characterized by CT numbers greater than 900 HU whereas at TLC, 30% of the same lung parenchyma is characterized by CT numbers ranging between 900 and 1,000 HU. In addition this threshold is in accordance with previous studies performed in patients with emphysema showing that the lung volume characterized by CT numbers below 900 HU correlates well with pulmonary function tests and histologic findings (15). It should be noted that this study does not allow the exact determination of the lung density characterizing an overdistended lung parenchyma, but rather gives a threshold from which normal aeration can be distinguished from overdistension. The ideal method would have been to anesthetize and intubate healthy volunteers and, after inflating their lungs to a pressure above 30 cm H₂O, to perform CT sections and measure the distribution of density histograms. For obvious ethical reasons, this methodology was impossible to implement. However, on several aspects, the method used in the present study is approaching this ideal maneuver. By performing a forced inspiration followed by a static expiratory effort against a pressure of 30 cm H₂O, it can be expected that some parts of the lung are overdistended. At TLC, it seems reasonable to assume that lung areas with a density already characteristic of some parts of the lung at FRC are not overdistended. Therefore, there is likely no overdistension at lung densities above 900 HU, a lung density virtually absent at FRC. It is possible that this way of determining the threshold for overdistension might result in some overestimation of the amount of overdistended lung. Given that many of the curves shown in Figure 1 at TLC are very steep in the region of 900 HU, a small change in the demarcation line can have a big effect on a percentage of the lung that is overdistended. However, because this threshold might be used in patients with ALI at risk of mechanical ventilation-induced lung barotrauma, it seems preferable to overestimate rather than underestimate mechanical ventilation-induced lung overdistension.

Effects of PEEP in Patients with ALI

Recently, Dambrosio and colleagues assessed mechanical ventilation-induced alveolar recruitment and lung overdistension in patients with adult respiratory distress syndrome (ARDS) using different ventilatory strategies (1). Analyzing lung morphology on three 1-mm-thick CT sections, these investigators defined the zone of overdistension as located between 1,000 and 800 HU. Using these definitions, they observed that 8% of the lung parenchyma of their patients was overdistended at FRC and ZEEP. Using PEEP, they observed at FRC that 26% of nonaerated lung regions were recruited whereas a 90% increase in lung overdistension was concomitantly observed. These results are in accordance with the results of the present study: PEEP-induced alveolar recruitment is often accompanied by lung overdistension. However, the magnitude of the phenomenon was likely overestimated by Dambrosio and colleagues, the CT numbers characterizing lung overdistension ranging between 900 and 1,000 HU rather than between 800 and 1,000 HU. It should also be noted that this threshold may slightly vary from one center to another. The use of CT numbers as absolute values is questionable (20) because they can be influenced by the type of CT scanners, the kilovoltages, and the reconstruction algorithm. Ideally, the density threshold for lung overdistension should be reassessed each time a new CT scanner is introduced.

Alveolar recruitment can be defined as a reduction in nonaerated or in poorly aerated lung volumes. In the present study poorly aerated lung volume did not change following PEEP whereas nonaerated lung volume significantly decreased in all patients, attesting to PEEP-induced alveolar recruitment. Simultaneously, arterial oxygenation significantly increased with a concomitant decrease in pulmonary shunt. Because cardiac index and the other hemodynamic parameters did not vary, these beneficial effects can be entirely attributed to PEEP-induced alveolar recruitment. Apparently, PEEP-induced lung overdistension was not associated with any deleterious hemodynamic effects.

Effects of Contrast Material on the Frequency Distribution of CT Numbers

Another result of the study is the lack of influence of the injection of contrast material on the density histogram distribution and on the subsequent calculations of the different lung volumes. This result is of importance because contrast material is often used when performing CT scan in patients with ARDS in order to differentiate lung parenchyma from pleural effusion. Considering the radius of a normal alveoli is 75 µm and the volume of a voxel as 0.0047 ml, it can be calculated that a voxel contains between 2,500 and 3,000 alveoli. The CT number characterizing a given voxel averages the radiological densities of gas, lung parenchyma, extravascular lung water, and blood of a large number of alveoli. One can hypothesize that at the alveolar space level, the volume of blood is negligible compared with the volume of gas, therefore explaining why contrast material has little influence on the frequency distribution of CT numbers.

In conclusion, this study shows that overdistended lung parenchyma is characterized by CT numbers ranging between 900 HU and 1,000 HU. The threshold of 900 HU allows a reliable determination of PEEP-induced alveolar overdistension in patients with ALI. The injection of contrast material does not affect this threshold nor the frequency distribution of CT numbers.