Effects of superimposed tissue weight on regional compliance of injured lungs

Respiratory Physiology & Neurobiology 228 (2016) 16–24

Contents lists available at ScienceDirect

Respiratory Physiology & Neurobiology journal homepage: www.elsevier.com/locate/resphysiol

Effects of superimposed tissue weight on regional compliance of injured lungs夽 Mariangela Pellegrini a,b , Savino Derosa a , Angela Tannoia a , Christian Rylander c , Tommaso Fiore a , Anders Larsson b , Göran Hedenstierna d , Gaetano Perchiazzi a,b,∗ a

Department of Emergency and Organ Transplant, Bari University, Bari, Italy Hedenstierna Laboratoriet—Surgical Sciences, Uppsala University, Uppsala, Sweden c Department of Anaesthesia and Intensive Care Medicine, Sahlgrenska University Hospital, Göteborg, Sweden d Hedenstierna Laboratoriet—Medical Sciences, Uppsala University, Uppsala, Sweden b

a r t i c l e

i n f o

Article history: Received 8 May 2015 Received in revised form 6 March 2016 Accepted 6 March 2016 Available online 11 March 2016 Keywords: Lung compliance Computed tomography Acute lung injury Image registration Mechanical ventilation Gravity

a b s t r a c t Computed tomography (CT), together with image analysis technologies, enable the construction of regional volume (VREG ) and local transpulmonary pressure (PTP,REG ) maps of the lung. Purpose of this study is to assess the distribution of VREG vs PTP,REG along the gravitational axis in healthy (HL) and experimental acute lung injury conditions (eALI) at various positive end-expiratory pressures (PEEPs) and inflation volumes. Mechanically ventilated pigs underwent inspiratory hold maneuvers at increasing volumes simultaneously with lung CT scans. eALI was induced via the iv administration of oleic acid. We computed voxel-level VREG vs PTP,REG curves into eleven isogravitational planes by applying polynomial regressions. Via F-test, we determined that VREG vs PTP,REG curves derived from different anatomical planes (pvalues < 1.4E-3), exposed to different PEEPs (p-values < 1.5E-5) or subtending different lung status (pvalues < 3E-3) were statistically different (except for two cases of adjacent planes). Lung parenchyma exhibits different elastic behaviors based on its position and the density of superimposed tissue which can increase during lung injury. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Respiratory system compliance (CRS ) is an indicator of the severity of acute respiratory distress syndrome (ARDS) and is used for titrating ventilatory settings for ARDS treatment (Matamis et al., 1984). A growing body of evidence shows that intrapulmonary differences in the elastic proprieties of the lung are also potential sources of ventilator-induced lung injury (VILI) due to the increased

夽 The study was performed at the Hedenstierna laboratory, University Hospital, Uppsala, Sweden. ∗ Corresponding author at: Department of Emergency and Organ Transplant, Section of Anaesthesia and Intensive Care Medicine, University of Bari, c/o Centro di Rianimazione—Policlinico Hospital, Piazza Giulio Cesare, 11, 70124 Bari, Italy. E-mail addresses: [email protected] (M. Pellegrini), [email protected] (S. Derosa), [email protected] (A. Tannoia), [email protected] (C. Rylander), tommaso.fi[email protected] (T. Fiore), [email protected] (A. Larsson), [email protected] (G. Hedenstierna), [email protected] (G. Perchiazzi). http://dx.doi.org/10.1016/j.resp.2016.03.005 1569-9048/© 2016 Elsevier B.V. All rights reserved.

strain between neighboring groups of alveoli with different time constants (Steinberg et al., 2004). Several studies have demonstrated that the healthy lung is inherently inhomogeneous because of gravitational (Milic-Emili, 1986) and local forces (Ma et al., 2013). This results in vertical but opposing gradients of alveolar distension (Simon et al., 2005) and relative ventilation (i.e., the ratio between ventilated gas and the volume of gas already present in a lung region) (Milic-Emili et al., 1966 and Milic-Emili, 1986). More recent findings showed that along the gravitational axis, lung density, quantified via computed tomography (CT) and estimated superimposed pressure, increases from nondependent to dependent zones in both normal and ARDS lungs (Gattinoni et al., 1991; Pelosi et al., 1994). Our group has recently defined a method of assessing the topographic distribution of lung compliance (Perchiazzi et al., 2014) by using a series of CT images obtained at different inflation volumes. One relevant message of that study was that compliance, measured at the airway opening, is a lumped parameter that does not necessarily reflect the extent of regional differences of lung elasticity. Moreover, it provided a reliable method with which to compute

M. Pellegrini et al. / Respiratory Physiology & Neurobiology 228 (2016) 16–24

a distribution map of volume and pressure inside the lung at the voxel level. We have previously shown that regional strain is posturedependent in healthy lungs and that intrapulmonary differences between dependent and non-dependent planes are less pronounced in the prone than in the supine position (Perchiazzi et al., 2011). Considering that global lung compliance measured at the airway opening is used to both assess the degree of lung impairment and titrate respiratory support, we deemed it relevant to explore the effects of weight force on the distribution of the elastic properties in the lung. The purpose of the present study was to determine the relationship between regional lung compliance and the compression from superimposed tissue weight, by using two-dimensional maps of the distribution of pressure and volume variation in healthy lung conditions and an experimental model of human ARDS (eALI) at two levels of applied positive end-expiratory pressure (PEEP) and twelve inflation volumes. We hypothesize that the effect of gravity on regional lung compliance is moderated by the superimposed weight of the lung itself. 2. Methods After approval by the local animal ethics committee at Uppsala University, the present study was executed in agreement with the National Research Council guide regarding the “Principles of laboratory animal care” (NIH publication no. 86-23, revised 1985). 2.1. Animal preparation Five healthy pigs with a mean weight of 26.0 ± 2.8 kg after sedation underwent general anesthesia. Anesthesia induction was achieved via an intramuscular injection of atropine (0.04 mg/kg), tiletamine-zolazepam (5 mg/kg, Zoletil; Boeringer Ingelheim, Copenhagen, Denmark), and medetomine (5 ␮g/kg, Dormitor Vet; Orion Pharma, Sollentuna, Sweden). Before the injection of the muscle relaxant, endotracheal intubation was guaranteed via surgical tracheostomy by using a cuffed tube (6.0 Hi-Contour; Mallinckrodt Medical, Athlone, Ireland). An intravenous infusion of ketamine (20 mg/kg/h, Ketaminol; Vetpharma, Zurich, Switzerland), fentanyl (5 mg/kg/h, Pharmalink, Spånga, Sweden), and pancuronium (0.24 mg/kg/h, Pavulon; OrganonTeknika, Gothenburg, Sweden) in buffered glucose 2.5% (Rehydrex; Fresenius Kabi, Uppsala, Sweden) delivered at a rate of 7 ml/kg/h was used to maintain anesthesia. Mechanical ventilation was provided by a mobile ventilator (Servo-I, Maquet, Solna, Sweden) that delivered a baseline ventilation using a volume-controlled, constant flow modality with a tidal volume (VT ) of 9 ml/Kg and a respiratory rate (RR) of 20 bpm; the inspiratory-to-expiratory (I:E) ratio was 1:2, and there was a PEEP of 5 cm H2 O, with a fraction of inspired oxygen (FiO2 ) equal to 0.5. Oxyhemoglobin saturation (SpO2 ) was continuously measured via a transcutaneous sensor placed at the animal’s ear. An esophageal catheter (Oesophageal catheter, Erich Jaeger GmbH, Höchberg, Germany) was positioned in the distal third of the esophagus, using the Baydur technique (Baydur et al., 1982) to obtain continuous measurements of esophageal pressure (PESO ). A second balloon catheter was located lower within the gastrointestinal duct to continuously measure gastric pressure (PGA ). Pressure (PAW ) and flow (V’AW ) were continuously measured at the airway opening. Three pressure transducers (Digimaclic Pressure Transducers, Special Instruments GmbH, Nördlingen, Germany) were used to measure PAW , PESO , and PGA , while V’AW was acquired via a Fleisch pneumotacograph (Laminar Flow Element type PT, Spe-

17

cial Instruments GmbH, Nördlingen, Germany) positioned between the endotracheal tube and the ventilator and connected to a differential pressure transducer (Diff-Cap Pressure Transducer, Special Instruments GmbH, Nördlingen, Germany). All respiratory signals were acquired via an analog-to-digital converter card (DAQ-card AI-16XE50, National Instruments Corp., Austin, USA) controlled by the Biobench Software (ver.1.0, National Instruments Corp., Austin, USA) at a sampling frequency of 200 Hz. The inspiratory and expiratory airway volumes (VAW ) were obtained via flow integration. After instrumentation, the animals were mechanical ventilated for 60 min to stabilize the pigs’ hemodynamic and respiratory conditions. eALI was induced via an injection of oleic acid (OA) 0.1 ml/kg (Apoteksbolaget, Göteborg, Sweden) through a central venous catheter at repeated doses of 0.5 ml. The target index of lung injury was a SpO2 value less than or equal to 80%. During OA injection, adrenalin, in boluses of 0.01 mg, was used to avoid decreases in systemic arterial pressure. 2.2. Image acquisition and analysis protocol The procedure used to generate two-dimensional maps of pressure and volume distribution from the CT images has been recently published by our group (Perchiazzi et al., 2014) and will only be summarized here (Fig. 1). After a recruitment maneuver (RM) consisting of a constant airway pressure of 40 cm H2 O for 40 s, the inspiratory capacity (IC) of each animal was calculated. The entire IC was divided into twelve iso-volumetric steps, which were used to define the gas volumes to be delivered to the lungs before taking each CT scan (VT = (IC/12) × 1, (IC/12) × 2, (IC/12) × 3. . . up to IC). The animals underwent inspiratory hold maneuvers (IHMs) corresponding to the twelve inspiratory volumes, which were administered as monotonically increasing volumes. In order to restore a steady-state condition, each IHM was separated from the following ones via 2–3 min of tidal breathing. Whole-lung spiral CT scans (Somatom Sensation 16, Siemens, Erlangen, Germany) were performed for each IHM (120 KV, 80 mAs). At the end of the procedure, a CT acquisition at the zero end expiratory pressure (ZEEP) was performed. Each CT scan of the whole lung required less than 6 s of apnea. The described sequence, composed of RM, twelve IHM steps, and ZEEP, was performed in each animal at two levels of PEEP before and after the induction of eALI. Hence, the studied conditions were as follows: healthy lungs at a PEEP of 5 cm H2 O (HL5), healthy lungs at a PEEP of 10 cm H2 O (HL10), injured lungs at a PEEP of 5 cm H2 O (eALI5), and injured lungs at a PEEP of 10 cm H2 O (eALI10). For each studied condition and sequence of IHMs, five transverse planes were selected among the whole-lung CT scans, covering the lung from the para-diaphragmatic to the apical level. The five transverse planes had 25 mm of distance between them along the longitudinal axis. Hence, the distance between the most paradiaphragmatic CT plane and the most apical one was fixed at 100 mm. The slice thickness was 5 mm. Customized scripts for the Image Processing and Statistics Toolboxes of MatLab R2010 (MatLab, The MathWorks, Natick, USA) were purposely created by one of the authors (G.P.) in order to perform the image analysis. In each slice, the perimeter of the lung parenchyma was manually selected in order to avoid potential flaws that are typical of automatic segmentation algorithms, which may not distinguish atelectatic areas from chest-wall structures, due to their similar HU density values. The collected CT images of the lung underwent an image registration process, the details of which have been described previously (Flusser and Zitova, 2003; Perchiazzi et al., 2014). The process of image registration aims to enable comparisons between images

18

M. Pellegrini et al. / Respiratory Physiology & Neurobiology 228 (2016) 16–24

Fig. 1. Iso-gravitational alignment of CT scans. Transverse CT images were acquired at five levels from base to apex. The alignment with the iso-gravitational plane for each of the tested conditions was checked and then categorized into eleven iso-gravitational stripes. Images represent the compliance maps of the lung, which were obtained via the technique already described in Perchiazzi et al. (2014) for the four studied conditions of the lung. Vessels and bronchi are less mobile than the surrounding parenchyma and this last can be subjected to locally relevant forces. In fact, on one side of the apparent direction of vessels/bronchi motion on the transversal plane, parenchyma is relatively more strained; opposite to this, parenchyma is relatively more compressed.

of different dimensions because the images were taken at twelve inflation volumes. Volume maps (VMAP ) were produced by transforming the HU densities into gas volumes voxel by voxel using Formula (1) (Gattinoni et al., 1986a,b): Vgas = Vvox ·

−HU 1000

(1)

where Vvox is the volume of one voxel, Vgas is the volume of the gas contained in the voxel, and HU is the density measured in Hounsfield units. Similarly, the volume of tissue occupying one voxel is yielded by Eq. (2):



Vtissue = Vvox · 1 −

−HU 1000

 (2)

where Vtissue is the volume of tissue contained in the voxel. Images from consecutive volumes of the inflation sequence were then subtracted between them, thus generating volume increment maps (VMAP ). In order to estimate lung compliance, it is necessary to compute both VMAP and a corresponding map of changes in transpulmonary pressure (PTP,MAP ) voxel by voxel. PTP,MAP was estimated by correcting the global transpulmonary pressure for the local factors that affect the lung (Perchiazzi et al., 2014). Considering the lung to be composed of a stack of fluid layers (Krueger et al., 1961), it is possible to calculate the hydrostatic pressure exerted by each layer on the one below by knowing the voxel-by-voxel lung density, as provided by the CT. Thus, the density is used to estimate the hydrostatic forces relative to a CT slice, which are derived from the HU values assigned by the CT scanner

M. Pellegrini et al. / Respiratory Physiology & Neurobiology 228 (2016) 16–24

to the voxels composing the same image, without any intervention by the authors. In this way, by measuring PESO and noting the position of the esophagus along the gravitational axis, it is possible to estimate the pleural pressure exerted on each isogravitational lung layer. For each iso-gravitational lung layer with height y, the transpulmonary pressure variation (PTP ,y ) is as follows: PTP,y = (PAW,plat − PAW,EE ) − (PESO,plat,y − PESO,EE,y )

(3)

where PAW,plat is airway pressure during IHM, PESO,plat is the esophageal pressure at the same time and at height y, and PAW,EE and PESO,EE,y are the corresponding airway and esophageal pressures, respectively, at the end of expiration. In the four lung conditions, lung compliance at inspiratory vital capacity and at airways opening (as CL = V/PTP ) was noted in order to monitor the progress of lung injury during its induction.

19

Table 1 Haemodynamic and gas exchange data. (mean ± standard deviation) Values sampled before and 30 min after the induction of eALI. In both conditions, values reported in Table were sampled at a PEEP level of 5 cm H2 O and at a FiO2 of 0.50. HR: Heart rate; MAP: mean arterial blood pressure; CVP: central venous pressure; MPAP: mean pulmonary artery pressure; PCWP: pulmonary capillary wedge pressure; CO: Cardiac Output; Hb: hemoglobin concentration; PaO2 : partial pressure of oxygen in systemic arterial blood; PaCO2 : partial pressure of carbon dioxide in systemic arterial blood. Healthy HR [bpm] MAP [mmHg] CVP [mmHg] MPAP [mmHg] PCWP [mmHg] CO [l/min] Hb [g/dl] PaO2 [mmHg] PaCO2 [mmHg]

97 93 9 24 11 3.2 8.7 247 39.7

± ± ± ± ± ± ± ± ±

16 17 3 6 2 0.5 0.6 81.5 6.0

eALI 109 88 11 38 13 3.7 9.5 105 52.6

± ± ± ± ± ± ± ± ±

14 17 2 6 3 0.7 0.6 47 7.5

2.3. Analysis of the gravitational course of elasticity The analyzed conditions were twelve inspiratory volumes at two levels of PEEP in healthy and eALI states for each animal. Because, in each condition, five transversal CT slices were selected for further analysis, a first step was to check their alignment with regard to the same iso-gravitational plane throughout the sequence of inspired volumes. Each CT scan, as described above, allowed the creation of a pair of maps: one VMAP and one PTP,MAP . The following analysis was performed to divide each of these matching pairs into eleven coordinated gravitational stripes. These divided the images at exactly the same height according to a common isogravitational plane (see Fig. 2, panel E). In this way, we can pool the local V and PTP data from the five levels and the five animals in each condition. The mean values of coordinated V and PTP were computed and plotted. Henceforth, we number the isogravitational planes of the lung from plane 1 (as the most nondependent) to plane 11. Moreover, V vs PTP relationships were fitted to a polynomial model (MatLab R2010, Curve Fitting Toolbox, MathWorks, Natick, USA) having the following form: y = ax3 + bx2 + cx + d

(4)

where y = PTP and x = V and coefficients a, b, c, and d do not have a univocal physical significance (see discussion). These regressions were tested for statistical significance by using an F-test. In this statistical test and in all the following ones, we defined ␣ = 0.05. Furthermore, the curves obtained at the two lung conditions and two PEEP levels were compared (by applying F-tests) in the following ways: 1) within the same PEEP level, comparing healthy vs eALI using curves derived from the same isogravitational plane; 2) within the same lung condition, comparing the PEEP of 5 cm H2 O vs the PEEP of 10 cm H2 O using curves derived from the same isogravitational plane; and 3) within the same lung condition and PEEP level, comparing curves derived from different isogravitational planes. In relation to the necessity of performing multiple comparisons, we applied the Bonferroni correction: the level of ␣ was reduced to ␣/n (where n is the number of simultaneous comparisons), resulting in ␣ = 0.05/10 = 0.005. We had as a null hypothesis that the corresponding isogravitational planes showed equal courses between healthy and injured conditions and between PEEP levels. If the null hypothesis was rejected, an additional factor or factors had an effect on the regional pressure vs volume curve. The hydrostatic superimposed pressure exerted on each of the eleven layers was plotted, along with

the pressure-volume curve of the single layers into which the parenchyma was divided. The variation in superimposed pressure between the highest and the lowest delivered volumes was computed and checked for statistical significance in each layer for the four tested conditions (Student T-test). The mean weight of the superimposed tissue over the various lung inflation volumes was computed in the eleven isogravitational planes. By applying the Student T-Test for the paired samples, we investigated whether the induction of eALI and the modification of PEEP determined a statistically significant change in superimposed weight. Moreover, using the polynomial model described above, we calculated the regression of the superimposed tissue weight vs plane position, in healthy and eALI conditions and computed their statistical significance by applying the F-test. In order to assess the homogeneity of the volume distribution, we computed the differences in gas volume between the mostand least-inflated regions of the lung throughout all the delivered volumes and compared them via the Student T-test. From each regression curve, we computed the second derivative in an attempt to calculate its inflection points. This was necessary to assess the position and concavity of the curvatures generated by the third-degree polynomial, as seen in Eq. 2. We studied the inflection points positioned in the range of the fitted data and computed the mean and standard deviation separately for the four tested lung conditions (HL5, eALI5, HL10, and eALI10). Moreover, by using the Student T-test, we assessed whether there was any statistically relevant difference in inflection point positioning between the two PEEP levels and between healthy and injured lung. We computed the average density of lung parenchyma corresponding to nonaerated tissue voxels in all the studied conditions and verified whether there was any statistically significant difference between healthy and eALI conditions, by applying the Student T-test. 3. Results All animals survived the experiment. The main hemodynamic data are shown in Table 1. After the induction of eALI, CL decreased from 33.7 ± 7.3 to 20.5 ± 4.7 [ml/cm H2 O] (measured at a PEEP of 5 [cm H2 O] and half vital capacity). Considering all tested conditions (twelve lung volumes, two PEEP levels, two lung conditions (HL and eALI), and five anatomical planes), 1200 coordinated V and PTP couples were obtained (Fig. 1). We obtained four pencils of regional V vs PTP curves (Fig. 2); each curve depicted the behavior of one (out of eleven) iso-gravitational plane and was composed of twelve points. All 44 curves were described by polynomial equations whose goodness-

20

M. Pellegrini et al. / Respiratory Physiology & Neurobiology 228 (2016) 16–24

Fig. 2. Regional volume vs pressure curve of the lung. Panels A–D refer to the four studied conditions: on the y-axis, the median volume of gas in each of the eleven planes of the studied animals is reported. Please note that the gas volume reported in the graph, is relative to CT slices having a thickness of 5 mm. The overall volume/pressure curve at airways opening is presented in the online supplementary material. On the x-axis, the median of the regional transpulmonary pressure is plotted. In Panel E, the legend for the plots above is provided, which uses a coded color scale to maintain the correspondence between the anatomical section and the line on the plot.

of-fit parameters made them statistically significant regressions upon F-test (see Fig. 3 and Online Supplementary Table 1 and Figs. 3–6). A visual inspection of the curves indicated that they shared a uniform morphology, which was expressed by a single curvature in all but four curves. The curve from the most dependent areas of the lung received a higher gas volume than the less dependent curves for similar PTP values. The four curves exhibiting a different morphology originated from the most dependent areas of the eALI lungs: the three bottom levels at a PEEP of 5 cm H2 O (Fig. 2, panel B) and the most dependent level at a PEEP of 10 cm H2 O (Fig. 2, panel D). These four curves presented an evident double curvature, showing initially low compliance values and then assuming a pattern similar to the others. The mean differences in gas volume between the most and the least inflated lung planes were as follows (data expressed as mean ± standard deviation, as [ml]): 0.33 ± 0.12 for HL5, 0.63 ± 0.24 for eALI5, 0.18 ± 0.09 for HL10 and 0.44 ± 0.18 for eALI10. The induction of eALI increased the weight of the superimposed tissue in all the eleven isogravitational planes (see Fig. 4), and this increase was statistically significant (p=0.0034). As expected, the induction of eALI increased inhomogeneity, as indicated by the increase in the difference in regional lung inflation, both at a PEEP of 5 (p = 0.0011) and at a PEEP of 10 cm H2 O (p = 0.0003). Moreover, increasing PEEP from 5 to 10 cm H2 O reduced the regional inhomogeneity both during healthy conditions (p = 0.0045) and during eALI (p = 0.037). When comparing curves derived from the same isogravitational level and PEEP (see Fig. 2) and differing in terms of lung status (healthy vs injured), all the curves were fitted by statistically different equations, except when studying plane 1 and plane 3 at a

PEEP of 5 cm H2 O. Similarly, at the same lung status, when PEEP varied from 5 to 10 cm H2 O, the same isogravitational plane subtended statistically different regression curves (see Supplementary Online Table 2, part A). Examining the course of the regression curves derived from the same lung status and applied PEEP but from different planes, out of 220 comparisons, the F-test detected statistically significant differences in all cases, except for these two contiguous plane pairs: plane 1 vs 2 (during HL5, HL10, and eALI5) and plane 5 vs 6 at HL10 (see Supplementary Online Table 2, Part B to E). Out of 44 extracted inflection points, 15 were outside the range of data (interval between the lowest and the highest sampled value) for which the regression was computed. Thus, they were discarded for the following calculations. The mean inflection points (given as co-ordinates using the following notation: mean volume ± standard deviation [ml], mean pressure ± standard deviation [cm H2 O]) were 0.07 ± 0.04 [ml], 15.6 ± 6.1 [cm H2 O] for healthy lungs and 0.19 ± 0.14 [ml], 24.4 ± 8.8 [cm H2 O] for eALI lungs. See Table 2 for a detailed report. Inflection points were placed in statistically different positions when comparing healthy vs eALI lungs (p = 0.0034), as well as in subgroups at the two PEEP levels: HL5 vs eALI5 (p = 0.0029) and HL10 vs eALI10 (p = 0.04). The difference between the positions of the inflection points for the two PEEP levels (p = 0.052) and in the eALI5 vs eALI10 subgroup were not significant (p = 0.083), although in healthy conditions [HL5 vs HL10 (p = 0.001)], a difference was found (see Supplementary Online Table 3). The hydrostatic superimposed pressure increased down the lung, as can be seen in Fig. 4. In fact the polynomial regression of the superimposed tissue weight vs plane position, was

M. Pellegrini et al. / Respiratory Physiology & Neurobiology 228 (2016) 16–24

21

Fig. 3. Relationship between superimposed pressure and regional pressure vs volume curves. Separate transpulmonary pressure (left y-axis, expressed as [cm H2 O]) vs volume (x-axis, expressed as [ml]) plots of the three representative anatomical planes (depicted as blue circles); for each plane, the corresponding superimposed pressure has been plotted (red triangles, referring to the right y-axis, expressed as [cm H2 O]). Plane 11 is the most gravitationally dependent. Plane 1 is the most nondependent. Observe the different regional volume ranges presented by each anatomical plane during lung inflation, as well as the corresponding pressure reached; please note that the superimposed pressure is already included in the reported calculation of transpulmonary pressure. eALI: experimental Acute Lung Injury. (The graphs relative to all the eleven anatomical planes are reported in the supplementary material).

8

[g]

Superimposed ssue

*

( )

6

4

Healthy ARDS

2

Table 2 Co-ordinates of inflection points. (mean ± standard deviation). Data deriving from polynomial curve regression, in the two conditions (healthy and after induction of eALI) and in the four combination of the two PEEP levels of 5 and 10 cm H2 O and the two lung conditions. Inflection points

Pressure [cm H2 O]

Healthy eALI Healthy, PEEP = 5 eALI, PEEP = 5 Healthy, PEEP = 10 eALI, PEEP = 10

15.6 24.4 14.1 26.5 17.2 23.3

± ± ± ± ± ±

6.1 8.8 7.8 11.4 3.6 7.8

Volume [ml] 0.073 0.191 0.104 0.290 0.042 0.134

± ± ± ± ± ±

0.047 0.145 0.050 0.133 0.008 0.126

Isogravitaonal planes

0

1

2

3

4

5

6

7

8

9

10

11

Fig. 4. Superimposed tissue. Mean superimposed tissue, measured in the eleven isogravitational planes (plane 1: nondependent; plane 11: most dependent) under healthy condition and eALI. Bars represent the standard error. The asterisk (*) marks a statistically significant increase of weight of the superimposed tissue in eALI as compared to healthy conditions, computed by applying the F-test to the polynomial regression curves (p = 0.0034).

statistically significant both in healthy (R = 0.99 and p < 1 × 10−6 ) and eALI conditions (R = 0.99 and p < 1 × 10−6 ). Application of Student T-test for paired samples showed that curves of superimposed tissue weight vs plane position were statistically different when comparing healthy vs eALI conditions (p = 0.003) or PEEP 5 vs PEEP 10 (p < 1 × 10−4 ). When superimposed pressure was studied in relation to the delivered inspiratory volume, it had a slightly descending course

(from low to high delivered inspiratory volume) in all tested conditions, which was more evident in the dependent planes (Fig. 3). However, in each single plane, in all the four study conditions, there was no statistically significant change of superimposed pressure between being exposed to the highest or to the lowest delivered volume.The highest variation in superimposed pressure was found in the most dependent plane, which had a mean of 3.0 ± 0.7 [cm H2 O] over the four experimental phases. The subpopulation of nonareated lung voxel presented an average volumetric mass (expressed as [gr/cm3 ]) in the healthy lung of 1.023 ± 0.187 and 1.013 ± 0.136 in the eALI. This difference was statistically significant (p = 3.686 × 10−8 ). However ALI lungs presented a superior quantity of nonaerated voxels equal to 12925 ± 1861 while in healthy lungs it was 10174 ± 584, also this difference was statistically different (p = 1.253 × 10−8 ).

22

M. Pellegrini et al. / Respiratory Physiology & Neurobiology 228 (2016) 16–24

4. Discussion The measurement of pressure and volume variation at the regional level was successfully achieved by applying a method that combined image analysis and respiratory mechanics measurements. The main conclusion of the reported experiments is that lung parenchyma exhibits different elastic behaviors according to its position with respect to the gravitational vector and according to the superimposed pressure, which depends on the density of the lung tissue and is consequently modified by the presence of lung injury. The course of the pressure vs volume curves has been characterized by statistically different polynomial regressions, which allowed the comparison of different lung conditions and the analysis of curve patterns. The novelty of the data reported in this paper is that we simultaneously estimated both pressure and volume variation at the regional level. Classical and recent papers have focused mainly on ventilation distribution, which is studied by using radionuclides (Rahn et al., 1956), computed tomography with and without Xenon as an inhaled contrast agent (Simon, 2005), positron emission tomography (Harris and Schuster, 2007), magnetic resonance (Hopkins et al., 2007), single-photon-emission computed tomography (Petersson et al., 2007), and microsphere distribution (Robertson and Hlastala, 2007). An effort to calculate the pressure acting on lung parenchyma was performed by Pelosi et al. (1994), who estimated the superimposed pressure by multiplying the lung density by the height of the iso-gravitational lung planes into which the lung CT slices had been divided. They focused on regional lung inflation; superimposed pressure was used to supply a theoretical basis to explain the gravitational gradients of ventilation. Instead, our method used superimposed pressure, airway opening pressure, and esophageal pressure (see Eq. (3)) in order to estimate transpulmonary pressure. Then, the assignment of PTP to space allowed the identification of the location of esophageal pressure measurement by tomography. As previously presented and validated (Perchiazzi et al., 2014), this allows the drawing of coordinated maps of PTP and V at the spatial resolution of a single voxel. 4.1. Regional volume vs pressure relation All the regional V vs PTP relations exhibited a curvilinear behavior. In Fig. 2, they are represented as starting from end-expiratory lung volume in order to make their pattern more evident. In fact, the curves describing dependent planes have higher volume-to-pressure ratios than the nondependent planes. The order of their course is altered in the three most dependent planes in eALI5 and the most dependent plane in eALI10. In these four cases, the apparent double curvature is determined by an initially low compliance that after an inflection point, changes to a higher value. Another notable difference is that (see Fig. 3) the nondependent planes of the lung have a higher initial end-expiratory volume, but they receive only a small fraction of the inhaled volume, which is distributed more greatly to the most dependent parts of the lung. Our results confirm classical observations about static gas distribution (Milic-Emili, 1986). In order to characterize and compare the different V vs PTP curves, we successfully applied polynomial regressions, the assumptions and limitations of which will be discussed further below. When the curves are examined in the same lung condition and applied PEEP, they have statistically different courses (with the exception of the contiguous planes 1 vs 2 and 5 vs 6 out of 220 cross-comparisons). This demonstrates that differently positioned lung planes (according to the gravitational vector) present different PTP vs V curves. The same occurs when comparing the curves derived from the same lung plane exposed to different

PEEPs or when examining the same lung plane before and after the induction of eALI. The induction of eALI shifts the V vs PTP curves to the right (i.e., the PTP coordinate of the inflection points increases) at both PEEP levels, demonstrating a global reduction in compliance. Instead, the change in PEEP shifts, in an orderly way, only the V vs PTP curves of healthy lungs. This can be explained by the fact that when the lung become inhomogeneous for the induction of eALI, the effect of PEEP modifies the mechanical behavior of the various planes according to their local conditions (i.e., if the damage has involved that particular plane or not). The overall effect is that during eALI, the univocal change in the position of the inflection points becomes statistically undetectable. However, the association of the regional lung PTP vs V curves with their position according to the gravitational axis does not necessarily assign exclusive causative mechanisms to gravitational forces. Nevertheless, lung elasticity can be affected not only by the weight of superimposed parenchyma per se. In fact, observing Supplementary Fig. 2, it can be noticed that there is a significant pressure gradient between the gravity dependent and non-dependent lung, particularly in the eALI conditions. In presence of increased permeability, the complex equilibrium between fluid filtration and reabsorption plays an important role and contributes to the mechanical load of the elastic structures (Staub, 1978). In fact, in our experimental model, the average density of the nonareated areas in eALI lungs was lower than in healthy conditions, resembling the density of fluids produced by protein exudation. Although showing a lower volumetric mass, voxels pertaining to nonareated tissue population are more represented in eALI than in healthy conditions. These findings support the idea that not only the weight forces, but also the distribution and the characteristics of fluids in the lung can play an important role. To translate the morphology of the PTP vs V curve into physical equivalents and interpret them, it is necessary to remark on the complexity exhibited by the behavior of normal lung tissue. In fact, while pulmonary alveolar structure per se can be considered uniform on the basis of histological examination and the distribution of elastin and collagen, apparent differences are caused by geometric constraints and by other loadings (different distension, gravity, and surface tension) (Fung, 1988). This is due to the fact that mechanical properties are not the simple sums of microscopic components (fibers, cells, and fluids) (Suki et al., 2005) but also relate to the way in which these components are interconnected: this generates behaviors that in the theory of models, are defined as emergent phenomena (Suki and Bates, 2011). This is the basis for the well-known (Milic-Emili, 1986) nonlinear behavior of PTP vs V curves: over the years (Murphy and Engel, 1978), various types of regression equation have been proposed to define their courses. Superimposed pressure is not the unique force that modulates the degree of inflation of alveoli exposed to the same airway pressure but located in different lung regions. The method used to estimate it stems from the idea that the lungs act as a fluid-like material in which each layer exerts a pressure on the one below that is proportional to its density (Krueger et al., 1961). This method does not take into consideration non-orthogonal forces across the layers or the fact that in principle, some areas may work as attractors rather than as simply hanging onto the one below. However, this is a widely used approach (Cressoni et al., 2014), and the results of this approximation are more accurate regarding the most dependent alveoli. Superimposed tissue density explains the Starling resistor (Conrad, 1969) effect on dependent small airways and the burden exerted on dependent alveoli well (Derosa et al., 2013). The composite effect of superimposed pressure and lung inflation may be responsible for the double curvature seen in the curves derived from the most dependent parts of the lung. This behavior

M. Pellegrini et al. / Respiratory Physiology & Neurobiology 228 (2016) 16–24

was hypothesized by Gattinoni et al. (1993), who first proposed the use of external PEEP to counterbalance it. 4.2. Technical issues and limitations All the extensive literature involving the use of CT methods (and radio-opacity in general) to infer gas volume variation in the lungs stems from the assumption of tissue mass conservation. Also, here, as in Fuld et al. (2008), we assumed that volume changes seen via CT are due to gas volume changes; Fuld demonstrated that changes in lung gas volume normalized by the initial gas volume depend directly on CT density with no need to further correct for tissue. However, we are aware that the conservation of tissue mass may not be perfectly achieved, because during ventilation, there can be a shift of blood between the lung and the surrounding tissues. For this reason, we measured the change in superimposed pressure plane-by-plane, investigating its variations in relation to delivered gas volume (Fig. 3, triangles). Only in the most dependent planes was a slight increase towards lower values at high delivered gas volumes observed, which never reached statistical significance. Papers examining the regional distribution of lung elasticity or transpulmonary pressure are lacking. For this reason, we decided to use a third-degree polynomial equation to fit the data from the eleven isogravitational stripes. We documented a tight regression. However, we cannot rule out that other functions may have had a good fit for the same pool of data. We needed the simplest equation that could have represented two physiological behaviors: nonlinear lung elastic properties and nonlinear weight force of tissue; third-degree polynomials are the simplest equations that can have two curvatures. This study has not been designed to compare different models (Murphy and Engel, 1978): curve fitting was used to ascertain whether there were different behaviors on the part of the different lung sections. It is not possible to assign a direct physical significance to the single coefficients of the equation: the iterative procedure of fitting tends to minimize the global estimation error, simultaneously leveling the computed variables. In order to compare images taken at different volumes, we have used a method known as image registration, which has already been used in other studies (Guerrero et al., 2006; Kaczka et al., 2011). As in a previous paper, we applied two-dimensional image registration and small lung volume increments in order to minimize potential errors in computing V maps derived from differences in the direction and magnitude of the motion of lung structures at different inflation volumes (Perchiazzi et al., 2014). In fact, atelectatic areas can work as “attractors” and may modify the expected direction of the motion of lung parenchyma. For this reason, image registration methods that use a single, pre-defined algorithm applied to an entire image are prone to registration inaccuracy (Zitovà and Flusser, 2003; Goshtasby, 1988). In the experiments reported here, we used the “piecewise linear” registration algorithm. This is based on the assumption that passing between consecutive inflation volumes, multiple linear elastic movements (also in the opposing direction) are possible. From a practical point of view, the technique consists of identifying corresponding structures in consecutive images by using marking points. These markers are then processed as vertices of elastic triangles (Delaunay triangulation) (Goshtasby, 1986). Although registration methods are based on mathematical theorems, for the purposes of physiological research, the possibility of confirming that anatomical structures within an image voxel are identical at two different inflation pressures is lacking. We deemed five levels to be qualitatively representative of the entire lung according to an interpolation principle already applied by Rylander et al. (2005) and later verified by Reske et al. (2010). In summary, this paper reports the estimation of both lung pressure and volume variation at the regional level. We demonstrated

23

that differently positioned lung planes (according to the gravitational vector) present different PTP vs V curves. Variation of PEEP or induction of eALI modify the PTP vs V curve of the single regions, however the application of a simple PEEP step change do not have univocal effects during eALI. Conflict of interest The authors declare that they have no competing interests. Acknowledgements We wish to express our gratitude to Agneta Roneus, Karin Fagerbrink, and Eva-Maria Hedin, laboratory technicians at Hedenstierna Laboratory, Uppsala University, for their assistance during the experiments; we also thank Monica Segelsjö of the Radiology Department at Uppsala University Hospital for her skillful CT management. This study was supported by grants from the Swedish Medical Research Council (5315); the Swedish Heart-Lung Fund; the School of Anesthesiology and Intensive Care Medicine, Bari University, Italy; and the Center of Innovative Technologies for Signal Detection and Processing (TIRES), Bari University, Italy. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.resp.2016.03.005. References Baydur, A., Behrakis, P.K., Zin, W.A., Jaeger, M., Milic-Emili, J., 1982. A simple method for assessing the validity of the esophageal baloon technique. Am. Rev. Respir. Dis. 126, 788–791. Conrad, W.A., 1969. Pressure-flow relationships in collapsible tubes. IEEE Trans. Biomed. Eng. BME-16, 284–295, http://dx.doi.org/10.1109/TBME.1969. 4502660. Cressoni, M., Chiumello, D., Carlesso, E., Chiurazzi, C., Amini, M., Brioni, M., Cadringher, P., Quintel, M., Gattinoni, L., 2014. Compressive forces and computed tomography-derived positive end-expiratory pressure in acute respiratory distress syndrome. Anesthesiology 121, 572–581. Derosa, S., Borges, J.B., Segelsjö, M., Tannoia, A., Pellegrini, M., Larsson, A., Perchiazzi, G., Hedenstierna, G., 2013. Reabsorption atelectasis in a porcine model of ARDS: Regional and temporal effects of airway closure, oxygen, and distending pressure. J. Appl. Physiol. 115, 1464–1473, http://dx.doi.org/10. 1152/japplphysiol.00763.2013. Flusser, J., Zitova, B., 2003. Image registration methods: a survey. Image Vis. Comput. 21, 977–1000, http://dx.doi.org/10.1016/S0262-8856(03)00137-9. Fuld, M.K., Easley, R.B., Saba, O.I., Chon, D., Reinhardt, J.M., Hoffman, E.A., Simon, B.A., 2008. CT-measured regional specific volume change reflects regional ventilation in supine sheep. J. Appl. Physiol. 104, 1177–1184. Fung, Y.C., 1988. A model of the lung structure and its validation. J. Appl. Physiol. 64, 2132–2141. Gattinoni, L., Mascheroni, D., Torresin, A., Marcolin, R., Fumagalli, R., Vesconi, S., Rossi, G.P., Rossi, F., Baglioni, S., Bassi, F., Nastri, G., Pesenti, A., 1986a. Morphological response to positive end expiratory pressure in acute respiratory failure. Compute. Tomogr. Study Intensive Care Med. 12, 137–142. Gattinoni, L., Pesenti, A., Torresin, A., Baglion, S., Rivolta, M., Rossi, F., Scarani, F., Marcolin, R., Cappelletti, G., 1986b. Adult respiratory distress syndrome profiles by computed tomography. J. Thorac. Imaging 1, 25–30. Gattinoni, L., Pelosi, P., Vitale, G., Pesenti, A., DAndrea, L., Mascheroni, D., 1991. Body position changes redistribute lung computed-tomographic density in patients with acute respiratory failure. Anestesiology 74, 15–23. Gattinoni, L., DAndrea, L., Pelosi, P., Vitale, G., Pesenti, A., Fumagalli, R., 1993. Regional effects and mechanism of positive end-expiratory pressure in early adult respiratory distress syndrome. JAMA 269, 2122–2127. Goshtasby, A., 1986. Piecewise linear mapping functions for image registration. Pattern Recog. 19, 459–466. Goshtasby, A., 1988. Image registration by local approximation methods. Image Vision Comput. 6, 255–261. Guerrero, T., Castillo, R., Sanders, K., Price, R., Komaki, R., Cody, D., 2006. Novel method to calculate pulmonary compliance images in rodents from computed tomography acquired at constant pressures. Phys. Med. Biol. 51, 1101–1112. Harris, R.S., Schuster, D.P., 2007. Visualizing lung function with positron emission tomography. J. Appl. Physiol. 102, 448–458, http://dx.doi.org/10.1152/ japplphysiol.00763.2006.

24

M. Pellegrini et al. / Respiratory Physiology & Neurobiology 228 (2016) 16–24

Hopkins, S.R., Levin, D.L., Emami, K., Kadlecek, S., Yu, J., Ishii, M., Rizi, R.R., 2007. Advances in magnetic resonance imaging of lung physiology. J. Appl. Physiol. 102, 1244–1254, http://dx.doi.org/10.1152/japplphysiol.00738.2006. Kaczka, D.W., Cao, K., Christensen, G.E., Bates, J.H.T., Simon, B.A., 2011. Analysis of regional mechanics in canine lung injury using forced oscillations and 3D image registration. Ann. Biomed. Eng. 39, 1112–1124, http://dx.doi.org/10. 1007/s10439-010-0214-0. Krueger, J.J., Bain, T., Patterson, J.L., 1961. Elevation gradient of intrathoracic pressure. J. Appl. Physiol. 16, 465–468. Ma, B., Sanderson, M., Bates, J.H.T., 2013. Airway-parenchymal interdependence in the lung slice. Respir. Physiol. Neurobiol. 185, 211–216, http://dx.doi.org/10. 1016/j.resp.2012.10.015. Matamis, D., Lemaire, F., Harf, A., Brun-Bruisson, C., Ansquer, J.C., Atlan, G., 1984. Total respiratory pressure-volume curves in the adult respiratory distress syndrome. Chest 86, 58–66. Milic-Emili, J., 1986. Static distribution of lung volumes. In: Macklem, P.T., Mead, J. (Eds.), Handbook of Physiology. Mechanics of Breathing (Section 3), vol. 3. American Physiological Society, Bethesda, pp. 561–574. Milic-Emili, J., Henderson, J.A., Dolovich, M.B., Trop, D., Kaneko, K., 1966. Regional distribution of inspired gas in the lung. J. Appl. Physiol. 21, 749–759. Murphy, B.G., Engel, L.A., 1978. Models of the pressure-volume relationship of the human lungs. Respir. Physiol. 32, 183–194. Pelosi, P., D’Andrea, L., Vitale, G., Pesenti, a, Gattinoni, L., 1994. Vertical gradient of regional lung inflation in adult respiratory distress syndrome. Am. J. Respir. Crit. Care Med. 149, 8–13, http://dx.doi.org/10.1164/ajrccm.149.1.8111603. Perchiazzi, G., Rylander, C., Vena, A., Derosa, S., Polieri, D., Fiore, T., Giuliani, R., Hedenstierna, G., 2011. Lung regional stress and strain as a function of posture and ventilatory mode. J. Appl. Physiol. 110, 1374–1383, http://dx.doi.org/10. 1152/japplphysiol.00439.2010. Perchiazzi, G., Rylander, C., Derosa, S., Pellegrini, M., Pitagora, L., Polieri, D., Vena, A., Tannoia, A., Fiore, T., Hedenstierna, G., 2014. Regional distribution of lung compliance by image analysis of computed tomograms. Respir. Physiol. Neurobiol. 201, 60–70, http://dx.doi.org/10.1016/j.resp.2014.07.001.

Petersson, J., Sánchez-Crespo, A., Larsson, S.A., Mure, M., 2007. Physiological imaging of the lung: single-photon-emission computed tomography (SPECT). J. Appl. Physiol. 102, 468–476, http://dx.doi.org/10.1152/japplphysiol.00732. 2006. Rahn, H., Sadoul, P., Farhi, L.E., Shapiro, J., 1956. Distribution of ventilation and perfusion in the lobes of the dog’s lung in the supine and erect position. J. Appl. Physiol. 8, 417–426. Reske, A.W., Reske, A.P., Gast, H.A., Seiwerts, M., Beda, A., Gottschaldt, U., Josten, C., Schreiter, D., Hlller, N., Wrigge, H., Amato, M.B., 2010. Extrapolation from ten sections can make CT-based quantification of lung aeration more precticable. Intensive Care Med. 36, 1836–1844. Robertson, H.T., Hlastala, M.P., 2007. Microsphere maps of regional blood flow and regional ventilation. J. Appl. Physiol. 102, 1265–1272, http://dx.doi.org/10. 1152/japplphysiol.00756.2006. Rylander, C., Tylén, U., Rossi-Norrlund, R., Herrmann, P., Quintel, M., Bake, B., 2005. Uneven distribution of ventilation in acute respiratory distress syndrome. Crit. Care 9, R165–R171, http://dx.doi.org/10.1186/cc3058. Simon, B.A., 2005. Regional ventilation and lung mechanics using X-ray CT. Acad. Radiol. 12, 1414–1422. Simon, B.A., Christensen, G.E., Low, D.A., Reinhardt, J.M., 2005. Computed tomography studies of lung mechanics. Proc. Am. Thorac. Soc. 2, 517–521. Staub, N.C.S., 1978. Pulmonary edema: physiological approach to management. Chest 74, 559–564. Steinberg, J.M., Schiller, H.J., Halter, J.M., Gatto, L.A., Lee, H., Pavone, L.A., Nieman, G.F., 2004. Alveolar instability causes early ventilator-induced lung injury independent of neutrophils. Am. J. Respir. Crit. Care Med., http://dx.doi.org/10. 1164/rccm.200304544OC. Suki, B., Bates, J., 2011. Lung tissue mechanics as an emergent phenomenon. J. Appl. Physiol. 110, 1111–1118. Suki, B., Ito, S., Stamenovic, D., Lutchen, K.R., Ingenito, E., 2005. Biomechanics of the lung parenchyma: critical roles of collagen and mechanical forces. J. Appl. Physiol. 98, 1892–1899. Zitovà, B., Flusser, J., 2003. Image registration methods: a survey. Image Vision Comput. 21, 977–1000.