Electrical impedance tomography applied to assess matching of pulmonary ventilation and perfusion in a porcine experimental model

Electrical impedance tomography (EIT) is a non-invasive, non-radiant, continuous technique that can be used to perform bedside measurements of impedance changes related to the thoracic content of air and blood [1, 2]. Several clinical and experimental studies have demonstrated that EIT can be applied to monitor respiration and tailor the settings of mechanical ventilation [3‐5]. The pulmonary circulation can also be monitored by EIT as reported both in normal conditions and during changes induced by disease or pharmacological interventions [6‐10]. However, simultaneous measurements of pulmonary ventilation and circulation by EIT have previously been only scarcely reported [11‐13]. The concurrent determination of ventilation-related and perfusion-related changes in impedance is complicated by the two signals differing in amplitude by up to an order of magnitude and being superimposed on each other with variable frequency harmonics.

We recently described how EIT can be used to monitor beat-to-beat changes in global pulmonary perfusion with adequate precision during acute variations in stroke volume related to changes in cardiac preload of pulmonary and extra-pulmonary origin [14]. This was performed by applying short periods of apnoea. The present investigation was designed to assess the matching of ventilation and perfusion, as reflected by the ventilation/perfusion (V/Q) ratio derived from EIT measurements. In addition to global measurements, the left and right lung ventilation and perfusion were studied separately and regionally in order to provide a comprehensive assessment of overall and regional V/Q matching.

The EIT algorithm generates values for impedance changes relative to a baseline measurement in every element of the reconstructed tomographic slice [15]. Significant changes in baseline conditions, for example sudden reductions in blood volume or the occurrence of pneumothorax or haemothorax, may confound the interpretation of impedance changes in absolute terms at different time points [16]. Therefore, the relation and distribution of impedance changes within and between defined regions of interest for any given time point, as utilised in the present experimental design, may provide a more robust estimate of ventilation and circulation in different parts of the lung. The V/Q results may potentially be helpful in optimising ventilator and haemodynamic therapy, taking into account the complex cardiorespiratory interactions, and could significantly extend the clinical applicability of EIT.

The aim of the experiment was to investigate the feasibility of EIT to describe the dynamic distribution of pulmonary ventilation and perfusion at resting conditions and during interventions designed to generate a wide range of V/Q ratios. The application of 20 cmH₂O of positive end-expiratory pressure (PEEP) was used to primarily increase pulmonary aeration with secondary reduction of cardiac preload. Inflation of a balloon in the inferior caval vein was used to acutely reduce cardiac preload with minimal effects on pulmonary ventilation.

We hypothesised that combined measurements by EIT of lung ventilation and circulation could be used to assess V/Q matching including the extremes of venous admixture and wasted ventilation. It was further hypothesised that V/Q ratios by EIT would correlate to shunt and dead space fractions calculated using standard formulae.

Six Swedish Landrace pigs (weighing 32 to 34 kg) were included in the study following approval by the Ethics Committee for Animal Experiments at the University of Göteborg. Animal care conformed to the principles set forth in the Guide for the Care and Use of Laboratory Animals [17]. A limited dataset from two of the animals included in this study has previously been reported [14].

Baseline cardiac output was 3.0 ± 0.8 L/minute and decreased to 1.8 ± 0.3 L/minute during PEEP20 and to 1.4 ± 0.3 L/minunte during B_infl, corresponding to stroke volumes of 27 ± 4 mL at baseline, 14 ± 4 mL during PEEP20 and 12 ± 4 mL during B_infl. Mixed venous oxygen saturation decreased from 70 ± 3% at baseline to 41 ± 9% during interventions. Baseline Q_s/Q_t was 11 ± 2% and decreased to 5 ± 1% during PEEP20, whereas baseline V_D/V_T increased from 9 ± 4% to 27 ± 11% during B_infl.

The greatest impedance amplitudes for Z_V and Z_Q at baseline were observed in the mid-ventral region (ROI 2) with no significant differences between the left and right lung. B_infl reduced Z_Q in both left and right ROI 2, although an increase was noted in left ROI 1. PEEP20 also decreased Z_Q in the left and right ROI 2 and increased Z_V in ROI 3.

The relative distribution of Z_V and Z_Q between the individual ROIs in relation to the total, cumulative signal in ROI 1 to 4 is illustrated in Figures 2 and 3. At baseline conditions, a parallel distribution of Z_V and Z_Q was observed, with the greatest proportion of ventilation and perfusion appearing in ROI 2 followed by ROI 3 and then ROI 1, while only a minor, although equal, fraction was observed in ROI 4. B_infl decreased the proportion of Z_Q distributed to ROI 2 in both the left and right lung and increased the proportional Z_Q in the left ROI 1. No changes were observed in the distribution of Z_V between the ROI 1 to 4. Hence, Z_V increased in relation to Z_Q in ROI 2 bilaterally, although the reverse was observed in the left ROI 1 (Figure 2). PEEP20 increased the proportion of Z_V in ROI 3 bilaterally while Z_V in ROI 2 tended to decrease. A reduction of relative Z_Q was observed in ROI 2 and thus Z_V increased in relation to Z_Q in ROI 2 and with a similar trend in ROI 3 (Figure 3).

The calculated Z_V/Z_Q ratios were plotted against Z_V and Z_Q before and after B_infl (Figures 4 and 5) and PEEP20 (Figures 6 and 7). At baseline, a close to ideal Z_V/Z_Q ratio of 1 was observed in ROI 2 and similarly in ROI 3 and 4 although with increased scatter, while a trend towards venous admixture was noted in ROI 1. B_infl increased the Z_V/Z_Q ratio in ROI 2 indicating dead space ventilation while increased venous admixture appeared in ROI 1. Changes in ROI 3 and 4 did not reach statistical significance because of considerable scatter. Following PEEP20, Z_V/Z_Q ratio increased in ROI 2 consistent with dead space ventilation and a similar trend was noted in ROI 3 while venous admixture appeared in ROI 1. No significant changes appeared in ROI 4 as a result of persistent scatter.

The frequency distribution of all EIT Z_V/Z_Q ratios calculated at baseline demonstrated a Gaussian distribution with a mean of 1.1 (95% confidence interval = 0.96 to 1.2). This distribution was markedly changed into a non-Gaussian pattern during B_infl and PEEP20 with a mean of 0.86 (95% confidence interval = 0.78 to 1.1; Figure 8). The calculated Z_V/Z_Q ratios based on global EIT correlated to venous admixture (r² = 0.48 [regression coefficient]) with a negative slope (y = -194 × +67) and to fractional dead space (r² = 0.63) with a positive slope (y = 92 × +32; Figure 9).

This study demonstrated the use of EIT to study pulmonary ventilation and perfusion in parallel. Using an experimental model to generate a wide range of stroke volumes for a constant tidal volume, we were able to investigate the dynamic relation between Z_V and Z_Q. By defining four ROIs along the ventral to dorsal axis for each of the left and right sides separately, an extensive dataset of Z_V/Z_Q ratios was collected. A largely parallel proportion of Z_V and Z_Q in all four ROIs during baseline conditions corresponded to a bell-shaped frequency distribution of Z_V/Z_Q ratios with only moderate scatter. Acute changes in stroke volume, established by inflating a balloon in the inferior caval vein or increasing PEEP from 5 to 20 cmH₂O, while tidal volume was maintained, significantly increased the mismatch of regional, proportional Z_V and Z_Q, the scatter of Z_V/Z_Q ratios and the heterogeneity of the Z_V/Z_Q frequency distribution. Significant positive and negative correlations were demonstrated between fractional alveolar dead space and venous admixture, respectively, and the global Z_V/Z_Q ratio.

Although an expanding body of evidence supports the use of EIT to monitor pulmonary ventilation and perfusion separately, only few studies have been published on their contemporaneous measurements. Deibele and colleagues, using similar EIT equipment, focused on technical aspects to separate the cardiac and respiratory related signals in two healthy, spontaneously breathing volunteers [12]. Kunst and colleagues investigated left-to right division of ventilation and perfusion in 14 spontaneously breathing patients scheduled for lobectomy or pneumectomy for lung cancer [13]. Frerichs and colleagues studied pulmonary perfusion in three pigs using a hypertonic saline solution as an electrical impedance contrast agent [21]. Thus, this study extends previous observations by adding regional separation and V/Q matching during anaesthesia and mechanical ventilation. Venous admixture was assumed to be an inherent feature of the model using anaesthetised animals studied in the supine position with low PEEP. The balloon inflation mimics a situation of acute onset hypovolaemia and increased PEEP is commonly applied in mechanically ventilated patients to prevent or reduce lung collapse.

The method to assess pulmonary perfusion used in this study has been described previously [14] but warrants some comments. Global Z_Q represents an approximation of stroke volume but, like any other method, has an inherent bias and variability, although within clinically acceptable limits. The physiology behind cardiac-related changes in Z_Q is still imperfectly understood and, for example, the relative influence of large pulmonary vessels compared with pulmonary capillaries needs further research. Furthermore, Z_Q relates to a change in air/blood content by the stroke volume rather than flow velocity and no separation of the latter in terms of alveolar capillary flow or shunt flow can be made.

The definition of ROI is an essential point in the data analysis. In this study, four equally sized ROIs were set by default, covering the full anterioposterior distance and the Z_Q and Z_V signals were analysed off-line using a stand-alone computer algorithm. It remains possible that setting the four ROIs within the pulmonary boundaries of the individual EIT image might have resulted in other Z_V and Z_Q amplitudes. In particular the most dorsal ROI 4 may be less representative because of the low impedance signal in this region [22]. The default full anterioposterior approach using equally sized ROIs was chosen in the design of this study to minimise operator-related bias. Initial pilot data indicated that placing the ROIs individually in each analysis significantly increased inter- and intraobserver variability.

The distribution of ventilation and perfusion at baseline differed along the ventral to dorsal distance with the greatest proportion delivered to the mid-ventral followed by the mid-dorsal regions. This is in line with the concept of gravitational influence on distribution and appeared even more clearly if all four ROIs were individually adjusted to the apparent pulmonary field of the EIT image (data not shown). Importantly, the distribution of Z_V and Z_Q were very closely matched, as would be expected in stable baseline conditions. Balloon inflation significantly reduced Z_Q in ROI 2 leading to apparent dead space ventilation in this region, while an apparent redistribution of Z_Q towards ROI 1 in the left lung would contribute to venous admixture. However, it should be noted that ROI 1 represents a moderate volume of tissue with the heart and lungs in close proximity.

The spatial resolution of EIT is inferior to that of other imaging techniques, for example computed tomography, and cannot exclusively separate impedance changes originating from the cardiopulmonary border. It is thus possible that the recorded increase in Z_Q in ROI 1 during balloon inflation results from an increased cardiac contribution rather than an increase in pulmonary blood flow. Increased PEEP to 20 cmH₂O shifted ventilation towards the more dorsal region, as previously reported [5]. At the same time, Z_Q was reduced, leading to a relative increase in dead space ventilation. Thus, the well-matched distribution of Z_Q and Z_V was partially lost following the interventions. The relative distribution of Z_Q and Z_V highlights general changes in V/Q matching. In order to gain a more detailed picture, the individual relative Z_Q and Z_V were plotted against the Z_V/Z_Q ratio. This approach to illustrate V/Q matching is less prone to artefacts induced by the selection of ROIs because all data comes from within the same ROI, as opposed to the distribution plot that relates one ROI to all others.

Again, ventilation and perfusion were closely matched at baseline, although a scatter of Z_V/Z_Q ratios appeared in the most ventral and dorsal parts of the lung. Balloon inflation increased the scatter in most ROIs and increased dead space ventilation in ROI 2, consistent with the acute reduction of blood flow in this region. PEEP ventilation increased dead space in ROI 2 and 3, whereas venous admixture appeared in ROI 1, which is in agreement with the dorsal redistribution of ventilation. Furthermore, the Z_V/Z_Q plots demonstrated the heterogeneity including both wasted ventilation and perfusion that turned out as significant correlations to shunt and dead space calculated independently of EIT. The relative ease of operation, low cost and lack of invasive procedures and radiation of EIT compares favourably with other methods to assess V/Q matching such as radionuclide scanning [23] or multiple inert gas elimination technology [24].

This was primarily a feasibility study of EIT to determine V/Q relations, and as such has some important limitations. The results are based on a small group of animals and pooled data from multiple observations in each animal was used, which might exaggerate the statistical significance demonstrated. All animals were without lung pathology and thus the results regarding PEEP should not be directly extrapolated to the clinical context when PEEP is applied during mechanical ventilation for respiratory failure. The interruption of ventilation when measuring perfusion-related impedance changes excludes the effects of tidal volume changes on pulmonary circulation. Although the acquisition of EIT data could be achieved on-line within minutes, the off-line analyses were more time consuming and would not be directly available at the bedside. No comparisons were made to other methods of assessing V/Q matching such as radionuclide scanning [23] or the multiple inert gas technique [24], and hence data on precision and bias of EIT in this respect is lacking.

We conclude that EIT might be used to monitor the distribution of pulmonary ventilation and perfusion making detailed studies of V/Q matching possible. The technique holds substantial potential although further work is necessary to position EIT in the context of clinical critical care.