The protocol was approved by our Institutional Review Board for the care of animal subjects. The care and handling of the animals were performed in accordance with the National Institutes of Health guidelines for ethical animal research.

Six female piglets (mean ± standard deviation (SD) = 28 ± 3 kg; Table 1) were premedicated with an intramuscular injection of xylazine (20 mg), droperidol (10 mg), and ketamine (500 mg). The animals were tracheotomized and mechanically ventilated (Avea; Viasys Healthcare, Höchberg, Germany) in volume-controlled mode using V_T 10 ml/kg, fraction of inspired oxygen (FiO₂) 0.21 during the part of the experiment on non-injured lungs, and zero end-expiratory pressure (ZEEP) (Table 1). Right internal jugular vein and carotid artery were cannulated. Anesthesia-analgesia was maintained with intravenous infusion of propofol 200 to 300 mg/hour and fentanyl 2 to 4 mcg/kg/min, and paralysis with pancuronium bromide 3 mg/hour.

The experiments were carried out in the experimental research imaging facility of the University of Lyon (CERMEP, Lyon, France).

The EIT device used was the Goettingen Goe-MF II System (Viasys Healthcare, Höchberg, Germany). A single array of 16 electrodes (Blue Sensor, BR-80-K, AMBU, Denmark) was placed on the mid-chest circumference of the animal. Electrical currents (50 kHz, 5 mA) were injected through adjacent pairs of electrodes in a rotating mode. During each electrical current injection, the resulting potential differences were measured at adjacent electrodes pairs and the resulting impedance (Z) distribution was calculated. The EIT recordings were sampled at a rate of 13 Hz, that is, 13 scans/second.

The PET study was performed using an ECAT EXACT HR+ scanner (Siemens, CTI, Knoxville, Tennesse, USA).

Piezoresistive pressure transducers (Gabarith 682002, Becton Dickinson, Sandy, UT, USA) were calibrated at the mid-chest level and connected to a A/D card (MP 100; Biopac Systems, Santa Barbara, CA, USA). Systemic arterial blood pressure, airway pressure and airflow (Fleish 2, Lausanne, Switzerland) were continuously recorded, sampled at 200 Hz, and analyzed with Acknowledge software (Biopac MP100 Systems, Santa Barbara, CA, USA). The value of V_T was obtained from the numerical integration of the airflow signal.

Once preparation was completed the animal was installed into the PET camera in a supine position. Two sets of experiments were performed in each animal. First, from its baseline value of 10 ml/kg, V_T was randomly changed to 6, 8, and 15 ml/kg on ZEEP. Second, while V_T was kept constant at 10 ml/kg, positive end-expiratory pressure (PEEP) was randomly changed from 5 to 15 cmH₂O by a 5 cmH₂O-step procedure. Each step was applied for five minutes (Figure 1).

In three animals, acute lung injury (ALI) was subsequently created by injecting 3 ml/kg hydrochloric acid 0.1 M via the endotracheal tube, after having increased FiO₂ to 100%. The target was to obtain partial pressure of arterial oxygen (PaO₂) less than 300 mmHg 10 minutes after inhalation. Additional doses of 1 ml/kg each were allowed to be used to reach this objective. Reinjection of HCl was needed once in only one animal. Once the target was reached, PEEP was set to 3 cmH₂O for two hours to obtain stabilization. At the end of the stabilization period, two sets of experiments were performed. First, at PEEP 5 cmH₂O, V_T was randomly changed to 8 and 12 ml/kg for 10 minutes each from the baseline of 10 ml/kg. Second, PEEP of 10 and 15 cmH₂O were applied in a random order for 10 minutes, at V_T 10 ml/kg. The respiratory rate was titrated to keep arterial pH above 7.20 and intrinsic PEEP lower than 1 cmH₂O.

Arterial blood gas was obtained from 2 ml of arterial blood injected into a cartridge (BG Cartridge, Gamida, Eaubonne, France) for immediate pH, partial pressure of carbon dioxide (PCO₂) and partial pressure of oxygen (PO₂) analysis using blood gas analyzer (IRMA Trupoint™, ITC, Edison, NJ, USA). At the end of each step, the following measures were assessed in this order: mean systemic arterial blood pressure; total PEEP (PEEPt) and end-inspiratory elastic recoil pressure of the respiratory system (Pplat, rs) by occluding the airways at the end of expiration for three seconds and at the end of the immediately following inspiration for four seconds, respectively; and lung ventilation.

The EIT signals were recorded continuously from the onset to the end of each experimental condition. PET assessment of ventilation was performed as follows (Figure 1). First, a transmission scan was made within 10 minutes. Then, the ¹³N-N₂ tracer continuously produced by the cyclotron fed the ventilator and was washed-in into the lungs through the endotracheal tube, and administered synchronously with the mechanical insufflations from the activation of an electronic valve 4. Once the activity of the tracer monitored from the camera screen plateaued, entry function of the tracer, that is, the amount of activity entering the lung, was measured at the endotracheal tube and equilibrium PET images were taken for three minutes. Then, the administration of the tracer was stopped at the very onset of inspiration and the tracer was washed-out from the lungs. Emission scans were taken for four minutes from the onset of washout to measure the tracer activity inside the lung.

The EIT signals retained in the comparison with the PET data were acquired for one minute at the time of transmission scan before tracer inhalation and during the wash-out period synchronously with emission scan (black squares in Figure 1). The wash-out period was selected because the modeling of the tracer kinetic with PET was performed from the data collected during the wash-out phase. The transmission frame was used to compare the effect of PEEP on lung volume while the emission frame was selected to compare the effect of changing V_T on lung ventilation. Therefore, this design has the unique feature of allowing the comparison between EIT and PET methods at the same time. To make the comparison between EIT and PET as accurate as possible, one of the most difficult issues to deal with was to match the same lung regions of interest (ROI) with each of the two techniques. An approximately 5 cm lung height was sampled with the 16-electrodes array 6. We selected as closely as possible the corresponding PET planes as follows. PET field of view was defined by laser projection onto the pig's thorax. Camera bed was then positioned so that the EIT electrodes were located at PET mid-field of view. The information contained in seven contiguous PET slices located at mid-field of view was then averaged, assuring an acceptable match between regions studied with both imaging techniques.

The investigators in charge of EIT (IF) and PET (JCR) analyses were blinded to the definition of each condition and, moreover, analyzed the data independently.

EIT scans were generated using the weighted backprojection reconstruction procedure along equipotential lines 7. EIT data was evaluated offline in terms of tidal volume (V_TEIT) and change in lung volume (V_L) in four ROIs corresponding to the anterior and posterior area of the right and left lungs, respectively. V_L reflected the shift in lung mid-capacity with PEEP relative to ZEEP 8.

ROIs were drawn around both lungs using PET transmission scans, on seven contiguous tomographic slices encompassing 5.1 cm of lung height. Lung volume measured with PET from density obtained on the transmission scan (VA_atten) was obtained from voxel-by-voxel values of lung attenuation in these ROIs, as previously described 5. ROIs were then superimposed on PET equilibrium and wash-out scans, and voxel-by-voxel time-activity curves were analyzed as previously described using a single compartment model 4. The modeling analysis enabled the determination of alveolar ventilation (V) expressed as ml/min/100 ml V_L and alveolar volume. Global analyses were performed on the whole set of voxels, while regional values were computed in four quadrants corresponding to the anterior and posterior area of the right and left lungs, respectively. In each of these regions, VA_atten and V_PET were computed as follows:

where i refers to the i^th voxel of the region and n to the total number of voxels of the corresponding region.

The values are presented as their mean ± SD. The relationships of V_TEIT (arbitrary units, a.u.) to V_PET (ml/min), in the first part of the experiment, were performed over the whole lungs from linear regression 9. Then, in each quadrant, the values of V_TEIT were computed as ml/min by using the following equation:

The same approach was used to compare VA_atten to V_L in the part of the study performed at different PEEP levels. The resulting predicted values of V_TEIT and V_L were henceforth expressed as ml/min and ml, respectively. Furthermore, since, by definition, V_L was 0 at ZEEP, the differences in VA_atten (ΔVA_atten) relative to ZEEP in normal condition and to PEEP of 5 cmH₂O in ALI condition were compared with the corresponding values of V_L across the PEEP levels.

Linear regression was performed by using the least square method. Bias and agreement were assessed from the Bland and Altman representation 10. The non-uniformity distribution of errors in regional measurements was checked by inspecting plots of residuals vs. predicted values. The statistical analysis was performed using SPSS statistical software (version 15.0 for Windows, SPPS Inc., Chicago, IL, USA). P < 0.05 was taken as the statistically significant threshold.

We found a strong correlation between global V_TEIT and V_PET (Figure 2a) over both conditions. The coefficients of determination were 0.95 and 0.91 (P < 0.001) in normal and ALI conditions, respectively. There were no bias and narrow limits of agreement (-37.42 to +37.42 ml/min) over both conditions (Figure 2b). The bias amounted to 5.77 and limits of agreement -24.49 to +36.03 ml/min for normal condition, and -16.59 and -55.26 to +22.08 ml/min for ALI condition. For regional ventilation, the correlation was slightly weaker but still significant (Figure 3a) over both conditions. The coefficients of determination were 0.63 in normal condition and 0.73 in ALI condition (P < 0.01). There were no fixed bias and narrow limits of agreement (-29.01 to +29.08 ml/min) over both conditions (Figure 3b). The bias was 1.47 and limits of agreement -29.71 to +32.66 ml/min for the normal condition, and 0.91 and -27.94 to +29.76 ml/min for ALI.

We found a strong correlation between global VA_atten and V_L over both conditions (Figure 4a). The coefficients of determination were 0.96 and 0.94 (P < 0.001) for normal and ALI, respectively. There were no bias and acceptable limits of agreement (-38.16 to +38.16 ml) over both conditions (Figure 4b). The bias (limits of agreement) were 0.28 (-30.17 to +29.61) ml for normal condition and 0.62 (-51.53 to +52.78) ml for ALI. At the regional level, the correlation was lower but still significant over both conditions (Figure 5a). The coefficients of determination were 0.76 (P < 0.01) and 0.54 (P < 0.05) for normal and ALI, respectively. There was no bias and limits of agreement ranged from -31.96 to +31.48 ml over both conditions. The bias (limits of agreement) were 0.21 (-26.17 to +26.58) ml for normal condition and -2.54 (-41.88 to +36.80) ml for ALI. The results pertaining to ΔVA_atten instead of VA_atten were similar (not shown).

Inspection of plots of residuals vs. predicted values disclosed that errors in measurements were uniformly distributed (Figure 6).

EIT analysis could be refined and extended further by implementing pixel-by-pixel analysis and by better defining atelectasis, so the functional lung recruitment should be assessed. Indeed, the lung recruitability 13 measured with the CT scan are anatomic features. However, for the lung mass recruited to be a relevant issue it should correspond to an increase in ventilation in those areas which continue to receive blood flow and, hence, should contribute to reduce the functional shunt. It has recently been shown that anatomic shunt and functional shunt do not correlate in ARDS patients 14. As lung perfusion could be assessed with EIT 15, this tool should be well suited to deal with these key issues. Further studies would be welcome to address these questions.