The ‘acute respiratory distress syndrome’ (ARDS) is characterized by an inhomogeneous pattern of lung damage, with atelectasis primarily occurring in the dependent lung regions. This syndrome with a complex constellation of symptoms from different etiologies has a high mortality (Ashbaugh et al 1967, Cressoni et al 2014, Bellani et al 2016, Thompson et al 2017).

Protective mechanical ventilation with ‘positive end-expiratory pressure’ (PEEP) to treat and prevent gravity-dependent atelectasis (Gattinoni et al 2018) has been demonstrated to improve patient survival (Briel et al 2010, Brochard et al 2017). However, currently no specific guideline for the application of PEEP exists (Grasselli et al 2023). Thus, PEEP can be set in different ways: one established method is to set the ‘fraction of inspired oxygen’ (FiO2) and PEEP according to the ARDSnet tables (NIH NHLBI ARDS Clinical Network 2008). Another method is to find the open lung PEEP with the best respiratory system compliance determined during a decremental PEEP titration (Suarez-Sipmann et al 2007).

The individual setting of PEEP above the alveolar closing pressure can also be achieved by employing an esophageal catheter to measure a surrogate of pleural pressure (Fichtner et al 2018, Williams et al 2019). It is placed semi-invasively as a gastric tube (Chiumello et al 2011). Talmor’s work suggests that the optimal level of PEEP is found at the transition from open to closed alveoli, where ‘transpulmonary pressure’ (TPP), defined as the pressure difference between end-expiratory ‘airway pressure’ (PAW) and the corresponding ‘esophageal pressure’ (PESO), reaches values between 0 and 10 mbar (Talmor et al 2008). Although one study was unable to demonstrate any significant benefits (Beitler et al 2019), it has been shown that PEEPs corresponding to this range, particularly a PEEP close to a TPP of 0 mbar, improved compliance, oxygenation and survival rates in ARDS patients on mechanical ventilation (Bastia et al 2021, Sarge et al 2021). Nevertheless, the methods mentioned so far measure a global lung condition, even though lung damage caused by ARDS is regionally heterogeneous.

In recent years the setting of PEEP at the crossing point of the curves of overdistension and of collapse determined by ‘electrical impedance tomography’ (EIT), a real-time imaging modality of regional ventilation in health and disease, has been advocated (Adler et al 2009, Costa et al 2009b, Frerichs et al 2017, Liu et al 2019, Dalla Corte et al 2020). Thoracic EIT uses an electrode belt at the level of the ventral 5th intercostal space to apply electrical currents and to measure voltages, which are the basis for reconstructing real-time tomographic images of impedance changes within the thorax (Brown 2003, Hinz et al 2008, Adler et al 2012, Costa et al 2009a, Muders et al 2010).

As previously described, volume-controlled (Ranieri et al 1994) and flow-controlled (Chen et al 2018) ramp maneuvers have been used to create global ‘pressure–volume’ (P/V) curves, regional P/V curves (Scaramuzzo et al 2019) and pressure-impedance curves (Sun et al 2020) to record the respiratory system compliance. The EIT-based determination of regional opening and closing pressures by P/V curves might help in setting appropriate ventilation parameters in ventilated patients (Pulletz et al 2012). However, identifying these pressures has been difficult (Cannon et al 2000) as the time-dependent analysis of changes in regional lung volumes resulting from pressure changes might not be sensitive enough to detect the opening and closing of smaller lung units.

Therefore, in this study novel EIT-derived pressure-flow curves are described, in which changes in regional flow provide surrogates for the regional opening and closing pressures of the lung. Thus, the specific objective of this study was to use EIT in conjunction with a custom-built pressure ramp maneuver to identify the airway pressures at which lung units in the dependent region, which are supposed to include the level of the esophagus, begin to collapse, and to correlate these pressures with the PEEP at a transmural pressure of zero.

The present study was performed at the Institute for Experimental Surgery, University Medical Center Rostock, with the approval of the State Ethics Committee for Animal Research (Landesamt für Landwirtschaft, Lebensmittelsicherheit und Fischerei, Mecklenburg-Vorpommern, Germany; LALLF, M–V–No: 7221.3.-1-014/20) and in accordance with the ‘Animal Research: Reporting of In vivo Experiments’ (ARRIVE) guidelines (Du Percie Sert et al 2020).

2.1. Preparation

The animals were raised by local pig farmers and brought to the stables one week before the start of the experiments to allow them to acclimatize. Free access to standard laboratory feed and water was available to them. Fifteen minutes after premedication with 7–8 mg kg−1 Azaperone (Stresnil®, Medistar Arzneimittelbetrieb GmbH, Ascheberg, Germany), 17,5–20 mg kg−1 Ketamin (Medistar Arzneimittelbetrieb GmbH, Ascheberg, Germany) and 33,5 mg kg−1 Midazolam (Dormicum®, Ratiopharm GmbH, Ulm, Germany) in a dark, quiet stable, 14 healthy, juvenile and female Landrace pigs (25–40 kg, 12–15 weeks of age) were transported to the operating theater and monitored. They were preoxygenated and anesthesia was induced with 100 mg of Propofol 2% at 20 mg ml−1 (B.Braun, Melsungen, Germany), 0.2 mg of Fentanyl (Fentadon®, Dechra Veterinary Products, Aulendorf, Germany) and 4 mg of Pancuronium (Pancuronium-Deltaselect®, Inresa Arzneimittel GmbH, Freiburg, Germany), followed by endotracheal intubation (ID 7.5 mm). The ELISA 800 respirator (Loewenstein Medical SE & Co. KG, Bad Ems, Germany) was used for ventilation in a hybrid ventilation mode (BiLevel IV) with the following default settings: a respiratory rate (RR) of 12–23 min−1, a tidal volume (TV) of 10–15 ml kg−1 body weight (BW) to achieve normocapnia (35–45 mmHg arterial carbon dioxide partial pressure (PaCO2)), a fraction of inspiratory oxygen (FiO2) of 0.5, an inspiratory–expiratory ratio (I:E) of 1:2 and an initial PEEP of 5 mbar.

Anesthesia was maintained with Propofol 2% at 4–8 mg kg−1 BW h−1, Fentanyl at 5–10 µg kg−1 BW h−1, Midazolam at 0.1 mg kg−1 BW h−1 and Pancuronium at 10 ml h−1. The depth of anesthesia was assessed by observing the animals’ vegetative signs, such as heart rate and blood pressure, to minimize the stress they experience during the experiment.

An arterial line was placed in the femoral artery to determine mean arterial pressure (MAP), cardiac index (CI), stroke volume variation (SVV) and pulse pressure variation (PPV) using the PulsioFlex Monitor (Pulsion Medical Systems, Feldkirchen, Germany).

Arterial blood samples were taken from a femoral arterial catheter using BGA monovettes (Sarstedt AG& CO KG, Nürmbrecht, Germany) and analyzed using the ABL 800 FLEX analyzer (Radiometer Medical Aps, Krefeld, Germany).

The impedance measurements were made with a porcine EIT Pioneer Set (Swisstom, Landquart, Switzerland) and a customized belt with 32 integrated electrodes (EIT-branch, SenTec AG, Landquart, Switzerland).

The 7.5 FR NutriVent esophageal balloon catheter (Sidam medical devices, San Giacomo Roncole, Italy) was placed to measure the PESO using the Baydur occlusion test (Baydur et al 1982). The TPP was calculated as the pressure difference between end-expiratory PAW and the corresponding PESO. This calculation was used to determine the Talmor PEEP based on an end-expiratory TPP of 0 mbar (Talmor et al 2008).

2.2. Lung injury and ramp maneuver

The protocol was divided into two sections—a healthy and a sick lung condition—each comprising the same ventilation maneuvers without randomization. A pre-post comparison could then be performed for each animal, comparing the ‘healthy lung’ control condition with the ‘ARDS’ intervention condition. No blinding was performed. All participants were aware of the group allocation at all stages of the experiment.

The pressure ramp started at 0 mbar at a FiO2 of 1.0 reaching a maximum pressure of 50 mbar with a pressure increase of 1 mbar s−1, and ending at 0 mbar decreasing the pressure at 1 mbar s−1. After the ramp, inspiratory hold maneuvers were performed for 30 s to calculate static compliance.

An ARDS-like condition was induced by repeated endotracheal lung lavage with 35 ml kg−1 BW of body-warm isotonic sodium chloride (NaCl) solution 0.9% and subsequent two hours of ventilator-induced lung injury (VILI) ventilation with a FiO2 of 1.0, PEEP of 0 mbar and TV of 15 ml kg−1 at an RR of 12 breaths per minute and an I:E ratio of 1:2.

The animals were euthanized at the end of the experiment using Pentobarbital at 45 mg kg−1 (Release®, Wirtschaftsgenossenschaft Deutscher Tierärzte eG, Garbsen, Germany).

2.3. Data acquisition

Pulse oximetry, ECG, arterial blood pressure, central venous pressure and pulmonary artery pressure were recorded synchronously, saved as an AD instruments file and sent to the AD instruments Powerlab (ADInstruments, Dunedin, New Zealand). Three minutes of measurement recording immediately before and after the ramp were used for analysis. Arterial blood gas analyses were performed before and after the ramp maneuver.

Ventilation parameters such as PAW and PESO were recorded at an acquisition rate of 200 Hz in csv format by the VIT Scientific Unit (Löwenstein Medical Innovation GmbH & Co. KG, Steinbach, Germany). Data were processed using Lab Chart 8 (AD Instruments, Dunedin, New Zealand).

Using STEM software (EIT-branch, SenTec AG, Landquart, Switzerland), raw EIT data (voltage values) were recorded at an acquisition rate of 47.68 Hz and EIT images (difference images from inspiration–expiration) were recorded with each breath and during the pressure ramp. EIT image-synchronized reference signals were sent from the EIT monitor to the AD instruments Powerlab and the Bridge Amp (Pioneer Interface Module, EIT-branch, SenTec AG, Landquart, Switzerland) and stored for offline analysis.

2.4. Data analysis

The EIT raw data and tidal images were generated for each pixel and reconstructed to ‘zero reference images’ (ZRI), which were adapted to the anatomy of the pigs using the ‘TicEmulatorForPigs’ (EIT-branch, Sentec AG, Landquart, Switzerland), based on the GREIT algorithm (Adler et al 2009).

Each EIT image was divided into 32 by 32 pixels, with pixel row 1 representing the most ventral slice and pixel row 32 representing the most dorsal slice of the image. As shown in figure 1 the ‘regions of interest’ (ROIs) of the lung were located from ventral to dorsal in pixel rows 7–23.

Figure 1. Estimated position of the esophageal catheter in EIT tidal image. The outer frames of the pixel rows (7–23) derived by electrical impedance tomography (EIT) are coded by a color gradient from gravity-non-dependent (bright orange) to gravity-dependent (dark gray) and the estimated esophageal level is marked by the arrow. Heart (1), thoracic aorta (2) and esophagus (3) are indicated by the respective numbers.

Download figure:

Standard image High-resolution image

The ZRI data were frequency adjusted, time-synchronized and normalized to the PAW in Matlab (The MathWorks, Natick, MA, USA) using the pre- and post-ramp breath peaks. The selection of the pressure ramp to be analyzed was made manually in the data set. The heart rate was detected within the frequency spectrum of the global EIT signal using a Fourier transform (Bracewell 2000). Interpolation was then performed using a linear time-resampling algorithm. Due to the presence of cardiac oscillations, both data sets were filtered using a finite impulse response (FIR) low-pass filter with a Kaiser window (Oppenheim and Schafer 2014). The filter was set to the following cutoff frequencies: 0.075 times the heart rate (in Hz) determined from the EIT frequency spectrum for the EIT signal (Kaiser window: passband edge frequency: , stopband edge frequency: , passband ripple: 0.5 dB, stopband attenuation: 30 dB), 0.15 times the heart rate for the derived EIT flow signal (zero-phase, order: 300), and 0.5 times the heart rate for the pressure signal (Kaiser window: passband edge frequency: , stopband edge frequency: , passband ripple: 0.5 dB, stopband attenuation: 30 dB). Furthermore, the EIT flow signal was smoothed using a Savitzky–Golay filter (order: 1, window length: 260 times the heart rate) (Savitzky and Golay 1964).

By plotting the regional relative impedance change per time (ΔŻ), as a surrogate parameter for the regional flow within each pixel, against the PAW, regional pixel-by-pixel P/ΔŻ curves were generated. The opening of the lungs was estimated during the ascending pressure ramp where the positive flow increased. In the descending pressure ramp, deflation parameters were determined in the range where negative flows increased in magnitude indicating the onset of alveolar collapse. To further characterize the deflation behavior of the regional P/ΔŻ curves, five characteristic points (CP1–CP5; see figure 2) were identified mathematically using the expiratory part of the abovementioned flow curves as follows: CP1 and CP5 were defined as the transition points between concave and convex curve segments (2nd derivative crossed 0). CP4 was defined as the point on the pressure-flow-curve with the maximum negative curve acceleration (local minimum of 2nd derivative). CP2 was defined as the intersection of the tangents at CP1 and CP4, while CP3 was defined as the tangent intersection of CP1 and CP5.

Figure 2. Representative regional pressure-EIT-derived-flow-curve during the ramp maneuver (a) and corresponding first and second derivative (b). Characteristic points (CP1–CP5) mathematically defining the deflation limb of the curve are plotted against the respective airway pressure (PAW). EIT-derived flow (ΔŻ) in arbitrary units (AU).

Download figure:

Standard image High-resolution image

Figure 3 shows the complete process of curve generation and parameter calculation, including all derivatives and filtering.

Figure 3. Process of generating curves and calculating deflation parameters. Finite impulse response Filter (FIR), Characteristic points (CP1–CP5) in mbar, Electrical impedance tomography (EIT), airway pressure (PAW) in mbar.

Download figure:

Standard image High-resolution image

To determine the position of the esophagus in the EIT images, CT images of previous studies were used in which both the aorta as the leading structure and the esophagus could be identified using an esophageal catheter (Wodack et al 2018, Brauer et al 2011). Based on the finding of Ferrario et al (2012), the pixels horizontal to the vertical gravity vector around EIT pixel rows 20 most closely represent the ventro-dorsal position of the esophagus. The pre-specified evaluation of pixel rows 18–23 was performed based on anatomical considerations. These rows are candidate rows of the dependent region and are expected to include the esophageal level at pixel row 20.

2.5. Statistics

The data underwent statistical analysis using the software program ‘IBM SPSS Statistics 27’ (IBM, Armonk, New York, United States of America). Linear regression analysis was performed to evaluate the association between the PEEP corresponding to an end-expiratory TPP of 0 mbar and the EIT-derived ‘characteristic points’ (CP2–CP5) identified in the EIT pixel row representing the dependent lung region and in the adjacent pixel rows (18–23). Analyses were conducted separately for each EIT pixel row as well as for the pooled dataset comprising all six pixel rows, to investigate the effect that a larger data set would have on the results. CP1 was used merely to calculate the other characteristic points CP2–CP5 and was excluded from subsequent correlation and regression analyses because it appears too early in the expiration limb to be relevant in connection with the alveolar collapse under investigation. Pressures are given in mbar with 1 mbar = 1,02 cmH2O.

For each analysis, the Talmor PEEP corresponding to an end-expiratory TPP of 0 mbar served as the independent variable, whereas the pressure values associated with the EIT-derived parameters we considered dependent variables. Pearson’s correlation coefficient (r), the regression slope and intercept together with their 95% confidence intervals and the coefficient of determination (R2) are reported. Continuous variables are expressed as mean value (MW) ± standard deviation (SD). For categorical variables, both the absolute frequency (n) and the relative frequency in % are given. P-values less than 0.05 were considered statistically significant. Agreement between TPP-derived and EIT-derived PEEP values was assessed using Bland–Altman analysis. For each characteristic point, mean bias and limits of agreement (bias ± 1.96 SD) were calculated. Predictive performance was evaluated using the mean absolute error (MAE) and the root mean square error (RMSE).

The PEEP at a TPP of 0 mbar could be determined in all 14 animals in both lung conditions using data from the descending pressure ramp.

The EIT data and ventilation parameters from one animal could not be analyzed due to technical issues. Therefore, 13 of 14 animals could be used for the calculations of lung mechanics and parameters.

Two animals died after lavage while inducing ventilation-induced lung injury. The lung condition of ARDS was reached by 11 of which two further animals died before the end of the protocol. Thus, complete measurements of the entire protocol were available for 9 of 14 animals. Following the unexpected death of three animals during the ARDS condition, the number of animals planned for the experiment was increased by this amount to reach the original number, which had been calculated to achieve sufficient statistical power. Table 1 presents the data on Talmor PEEP at TPP 0 mbar and lung mechanics together with the respective number of animals analyzed.

Table 1. Talmor-PEEP at TPP 0 mbar and parameters of lung mechanics.

 HealthyARDSParameterNMean ± SDNMean ± SDTalmor-PEEP (mbar) at TPP 0 mbar1410,1 ± 1,91214,3 ± 2,8Plateau pressure PPLAT (mbar)1424,0 ± 3,41424,0 ± 3,4Driving pressure ΔP (mbar)1424,0 ± 3,41424,0 ± 3,4CSTAT (ml mbar−1)1352,0 ± 8,1814,0 ± 8,7TV (ml)12451 ± 1049171 ± 123LIP (mbar)1416,2 ± 10,1928,6 ± 7,8UIP (mbar)1431,2 ± 11,0940,4 ± 7,7PMC (mbar)1420,1 ± 2,11025,1 ± 3,3

Positive end-expiratory pressure (PEEP), transpulmonary pressure (TPP), static compliance (CSTAT), tidal volume (TV), lower inflection point (LIP), upper inflection point (UIP), point of maximum curvature (PMC), number of animals analyzed (N), standard deviation (SD), Acute respiratory distress syndrome (ARDS).

As the number of EIT pixel rows within the ROI of the lungs was adapted to the individual lung size, it was not possible to calculate characteristic points down to pixel row 23 for each animal and each lung condition. In healthy lungs, for all 13 animals EIT pixel rows down to row 20 could be calculated, but only for 6 animals down to row 23. In 6 of the 9 ARDS conditions pixel rows could also be determined down to row 23.

Regional P/ΔŻ curves for each EIT pixel row of one representative animal derived from the sum of all pixels in the respective row for both, healthy and ARDS lungs, are shown in figure 4.

Figure 4. Regional pressure-flow curves in lung-healthy subjects (A) and ARDS (B). Sum of regional relative impedance change per time (ΔŻ) derived by electrical impedance tomography (EIT) of all pixels comprising one pixel level (row) for all pixel levels for healthy (A) and acute respiratory distress syndrome (ARDS) (B) lungs with respective colors as defined in figure 1. Regional flow maxima and minima (maximal negative flow) were marked as circles in ascending and descending pressure ramp. Arbitrary Units (AU), Airway pressure (PAW).

Download figure:

Standard image High-resolution image

The results from correlating the parameters of the descending ramp with Talmor PEEP at TPP 0 mbar are shown in table 2.

Table 2. Correlations (r) of EIT-based parameters and Talmor-PEEP at TPP 0 mbar.

 EIT Characteristic point 2Characteristic point 3Characteristic point 4Characteristic point 5pixel levelNrPrPrPrPHealthy18130,3650,2210,3280,2740,603<0,050,2970,324 19130,3790,2010,2990,3210,597<0,050,2310,449 20130,3820,1980,3020,3160,574<0,050,2270,456 21120,801<0,050,722<0,050,784<0,050,4640,129 22100,906<0,050,814<0,050,868<0,050,5410,107 2360,929<0,050,7730,0710,908<0,050,3450,504 18-23670,544<0,050,454<0,050,697<0,050,320<0,05ARDS1890,738<0,050,697<0,050,6630,0520,698<0,05 1990,731<0,050,703<0,050,4880,2200,5590,150 2070,817<0,050,813<0,050,5670,1840,6170,140 2170,764<0,050,3700,4130,6370,1240,7380,058 2260,2560,6240,4330,3910,6200,1890,6500,163 2360,1390,7920,1640,7560,2320,6580,3040,558 18-23440,614<0,050,537<0,050,547<0,050,602<0,05

The best regional correlations (r) and significances (P) for each parameter and for both lung conditions were highlighted in bold. Acute respiratory distress syndrome (ARDS), Electrical impedance tomography (EIT), number of animals analyzed (N).

For each regional parameter and lung condition, the best correlation of the individual EIT pixel rows 18–23 and the sum of 18–23, as identified in the exploratory post-hoc analysis, is highlighted. For example, in healthy lungs, characteristic point 2 (CP2) showed increasing correlations from non-dependent (r = 0.365 in EIT pixel row 18) to dependent EIT pixel rows (r = 0.929 in EIT pixel row 23, highlighted). The correlation from the sum of these rows was weak (r = 0.544).

For the CP with the strongest correlations in both lung states, a more detailed analysis was thereafter conducted to evaluate the effect size of this parameter when comparing EIT pixel rows (table 3). These analyses are reported to generate hypotheses regarding regional differences in lung derecruitment between healthy and ARDS lungs.

Table 3. Detailed statistics of Characteristic point 2 and optimal Talmor-PEEP at TPP 0 mbar for EIT pixel rows 18–23.

 EIT  95%-CI 95%-CI    pixel rowNrrR2R2PSlope (B)InterceptHealthy18130,365−0,240,760,1330,000,480,2210,26715 080 19130,379−0,220,760,1440,000,490,2010,45620 322 20130,382−0,210,760,1460,000,500,1980,47220 200 21120,8010,420,940,6410,170,840,0020,66618 873 22100,9060,630,980,8200,400,92<0,0010,86716 756 2360,9290,420,990,8620,180,950,0071,00515 357 18-23670,5440,350,690,2960,110,46<0,0010,60519 055ARDS1890,7380,160,940,5450,030,820,0231,3017,637 1990,7310,140,940,5340,020,820,0251,2797,857 2070,8170,120,970,6680,040,890,0250,89413 548 2170,7640,000,970,5840,000,860,0460,85714 045 2260,256−0,710,910,0660,000,630,6240,20922 163 2360,139−0,770,890,0190,000,580,7920,15023 298