Effect of regional lung inflation on ventilation heterogeneity at different length scales during mechanical ventilation of normal sheep lungs

Abstract

Heterogeneous, small-airway diameters and alveolar derecruitment in poorly aerated regions of normal lungs could produce ventilation heterogeneity at those anatomic levels. We modeled the washout kinetics of ¹³NN with positron emission tomography to examine how specific ventilation (sV̇) heterogeneity at different length scales is influenced by lung aeration. Three groups of anesthetized, supine sheep were studied: high tidal volume (Vt; 18.4 ± 4.2 ml/kg) and zero end-expiratory pressure (ZEEP) (n = 6); low Vt (9.2 ± 1.0 ml/kg) and ZEEP (n = 6); and low Vt (8.2 ± 0.2 ml/kg) and positive end-expiratory pressure (PEEP; 19 ± 1 cmH₂O) (n = 4). We quantified fractional gas content with transmission scans, and sV̇ with emission scans of infused ¹³NN-saline. Voxel ¹³NN-washout curves were fit with one- or two-compartment models to estimate sV̇. Total heterogeneity, measured as SD[log₁₀(sV̇)], was divided into length-scale ranges by measuring changes in variance of log₁₀(sV̇), resulting from progressive filtering of sV̇ images. High-Vt ZEEP showed higher sV̇ heterogeneity at <12- (P < 0.01), 12- to 36- (P < 0.01), and 36- to 60-mm (P < 0.05) length scales compared with low-Vt PEEP, with low-Vt ZEEP in between. Increased heterogeneity was associated with the emergence of low sV̇ units in poorly aerated regions, with a high correlation (r = 0.95, P < 0.001) between total heterogeneity and the fraction of lung with slow washout. Regional mean fractional gas content was inversely correlated with regional sV̇ heterogeneity at <12- (r = −0.67), 12- to 36- (r = −0.74), and >36-mm (r = −0.72) length scales (P < 0.001). We conclude that sV̇ heterogeneity at length scales <60 mm increases in poorly aerated regions of mechanically ventilated normal lungs, likely due to heterogeneous small-airway narrowing and alveolar derecruitment. PEEP reduces sV̇ heterogeneity by maintaining lung expansion and airway patency at those small length scales.

ventilation heterogeneity is present in normal lungs (10) and increased in disease (41, 66). Anatomical structures from large central airways to those at the microscopic subacinar level are thought to contribute to total ventilation heterogeneity (10, 17, 21). During mechanical ventilation of normal lungs, partial, complete, or intermittent airway closure and alveolar collapse have been implicated in changes in regional ventilation distribution (14, 16, 55, 60).

Images from intravital microscopy have shown that heterogeneous alveolar expansion (39) and collapse (6) can occur at the microscopic level. In areas of low lung inflation, loss of parenchymal tethering forces due to alveolar collapse may facilitate the reduction in diameter of peripheral airways (3), potentially in a heterogeneous manner (67). The resulting heterogeneity of regional airway resistance (Raw) and alveolar expansion would be expected to generate ventilation heterogeneity at small length scales. Measurements based on fluorescent microspheres implied existence of ventilation heterogeneity down to length scales of 2 cm³ (57), and recent findings from synchrotron computed tomography (CT) suggested that heterogeneity of specific ventilation (sV̇; defined as ventilation per unit gas volume) can occur at lung volumes as low as 1 mm³ (52). However, it is unknown to what extent different length scales contribute to total ventilation heterogeneity in mechanically ventilated normal lungs, or how those components are influenced by regional lung inflation.

Mechanisms producing ventilation heterogeneity at small length scales, such as heterogeneous alveolar collapse and expansion and cyclic small airway closure, are frequently cited as leading to ventilator-induced lung injury (VILI) (12, 37, 42). Thus knowledge of the length scales contributing to ventilation heterogeneity may provide relevant information on the lung structures involved in the development of VILI. In fact, injury of peripheral airways of diameter <1 mm (12) and parenchymal attachments in small airways (13) has been demonstrated during mechanical ventilation of normal lungs with zero positive end-expiratory pressure (PEEP). That VILI is relevant in normal human lungs has been emphasized by findings of acute lung injury (19) and pulmonary inflammation (43, 74) in mechanically ventilated patients without preexisting lung injury.

We have developed methods using positron emission tomography (PET) to assess heterogeneity of sV̇ at length scales below the effective image resolution of 12 mm (66, 68, 69), as well as at length scales above this resolution (65, 70). Subresolution sV̇ heterogeneity is reflected by a multicompartment voxel-level tracer kinetics during washout of [¹³N]nitrogen (¹³NN) from alveolar air space. Heterogeneity of sV̇ at length scales above the image resolution is evidenced by changes in sV̇ dispersion with progressive low-pass filtering. We hypothesize that the degree of ventilation heterogeneity at small length scales is inversely related to regional lung inflation. Accordingly, in supine normal lungs ventilated with zero PEEP (ZEEP) and normal tidal volumes (Vt), regions of low inflation should demonstrate increased ventilation heterogeneity at small-length scales as a result of heterogeneous small airway closure and alveolar derecruitment. High Vt in the presence of ZEEP could further increase such heterogeneity in regions of low inflation by promoting intratidal recruitment of small airways. Conversely, high PEEP levels would be expected to reduce small-length-scale ventilation heterogeneity by maintaining alveolar inflation and airway patency in all lung regions. To test these hypotheses, we used a sheep model to investigate in normal lungs whether 1) components of ventilation heterogeneity at several length scales are influenced by changes in regional aeration produced with distinct levels of PEEP and Vt; and 2) patterns of ventilation heterogeneity compatible with alveolar collapse and airway narrowing or closure occur in poorly aerated lung regions.

METHODS

Animal Preparation and Study Groups

The experimental protocols were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care (Boston, MA). Sixteen sheep (22.5 ± 5.0 kg) were fasted overnight and premedicated with intramuscular ketamine (4 mg/kg) and midazolam (2 mg/kg). After intravenous induction of anesthesia with ketamine (4 mg/kg), an endotracheal tube was inserted. A femoral artery and jugular vein were cannulated. General anesthesia was maintained with a continuous infusion of propofol and fentanyl titrated to heart rate and blood pressure. Pancuronium (0.1 mg/kg) was used for muscle paralysis at induction and repeated every 90 min (0.02–0.04 mg/kg).

Three mechanical ventilation protocols were used to study separately the effects of Vt and PEEP. For all groups, the inspired O₂ fraction was initially set at 0.3 and adjusted to achieve an arterial O₂ saturation > 0.88, inspiratory-to-expiratory time ratio = 1:2, and respiratory rate (RR) = 18 beats/min or higher to maintain the arterial carbon dioxide pressure between 32 and 45 Torr. In the first group, Vt was adjusted to achieve airway plateau pressure = 30–32 cmH₂O with ZEEP (Vt = 18.4 ± 4.2 ml/kg), aiming at maximizing Vt while maintaining clinically acceptable pressure limits and avoiding hyperinflation (high-Vt ZEEP, n = 6). A variable dead space was added to the breathing circuit if arterial carbon dioxide pressure was <32 Torr with RR = 18 beats/min. The second group was ventilated with Vt = 8 ml/kg and ZEEP (low-Vt ZEEP, n = 6). The last group was ventilated with a currently established lung protective ventilation strategy (38), with Vt = 8 ml/kg and PEEP adjusted for plateau pressure = 30–32 cmH₂O (PEEP = 19 ± 1 cmH₂O) (low-Vt PEEP, n = 4). In each animal, a recruitment maneuver (40 s at airway pressure = 35 cmH₂O) was performed before initiation of the mechanical ventilation protocols, to standardize lung volume history. Ventilation protocols were begun ∼30 min before acquisition of images. Respiratory mechanics were continuously monitored in four high-Vt ZEEP animals, five low-Vt ZEEP animals, and all low-Vt PEEP animals using a NICO cardiopulmonary management system (Philips Respironics, Murrysville, PA).

PET Imaging Protocol

For all PET scans, sheep were positioned supine in the camera (Scanditronix PC4096, GE Healthcare, Milwaukee, WI) with the most caudal slice of the field of view adjacent to the diaphragm dome. The camera collected 15 transverse slices of 6.5-mm thickness over a 9.7-cm-long axial field, providing three-dimensional data for an estimated 70% of the total lung volume (69). Transmission scans of 10 min were obtained using a rotating pin source of ⁶⁸Ge to correct PET emission scans for tissue attenuation and to measure lung density (56).

Emission scans of a ¹³NN-saline bolus infusion, consisting of 22 frames (8 × 2.5, 10 × 10, and 4 × 30 s), were obtained as previously described (68, 69), to measure regional perfusion and ventilation. Simultaneous with the start of imaging, a bolus of ¹³NN (∼1,000 MBq), dissolved in 40-ml saline, was injected into the jugular vein at the beginning of 60 s of apnea. Due to its low solubility in water and tissue (partition coefficient water/air = 0.018 at 37°C), virtually all ¹³NN in the blood diffuses into alveolar gas spaces at the first pass in aerated lung units, and tracer activity accumulates during apnea in proportion to regional blood flow. After 60 s, ventilation was resumed, and 3 min of washout were imaged. Because washout images were acquired during continuous breathing, each frame represented the average tracer activity over several breaths.

PET images were reconstructed with voxel size of 2 × 2 × 6.5 mm using a convolution back-projection algorithm. Images were filtered in the image plane with a circular moving average filter of diameter 12 mm and along the z-axis with a two-point moving average filter, yielding a 128 × 128 × 14 matrix with an effective volumetric resolution of 1.66 cm³.

Definition of Lung Fields and Regions of Interest

Transmission scans were processed, as previously described (22), to construct images of average fractional gas content (F_GAS) over the respiratory cycle. Volumetric masks of lung fields were delineated by initially including all voxels with F_GAS > 0.5. Perfused but poorly aerated voxels were identified in the early apnea frames of the ¹³NN bolus infusion scans and included. Masks were manually refined to exclude the trachea and two main bronchi. Regions of interest (ROIs) were defined in the following ways: 1) individual voxels were analyzed to measure intravoxel sV̇; 2) voxels were grouped according to F_GAS level, with bins of width 0.1 from 0 to 1, to examine the influence of inflation level on sV̇ behavior; and 3) the lung field was divided along the ventral-dorsal axis into three equally spaced regions, comprising nondependent, middle, and dependent ROIs, to examine regional sV̇ heterogeneity.

Tracer Kinetics Modeling

Intravoxel sV̇.

The tracer kinetics curves of each voxel were analyzed to determine perfusion and sV̇. The activity in the last apnea frame was included in the washout curve for each voxel and assigned washout time zero. Four possible models were used to describe tracer washout. A single-exponential model (Eq. 1) was used to describe voxels with a homogeneous washout pattern, compatible with a single compartment. A complete gas-trapping model (Eq. 2) was used to describe voxels with no change in activity during washout. A double-exponential model (Eq. 3) or a partial gas trapping model (Eq. 4) was used to describe voxels with heterogeneous washout patterns, compatible with more than one compartment. In regions with gas trapping, a compartment with sV̇ = 0 described the observed tracer retention (Eqs. 2 and 4).

$$A\left(t\right)=A_1·e^{-\text{s}{\dot{\mathrm{V}}}_1·t}$$ (1)

$$A\left(t\right)=A_1\left(\text{s}{\dot{\mathrm{V}}}_1=0\right)$$ (2)

$$A\left(t\right)=A_1·e^{-\text{s}{\dot{\mathrm{V}}}_1·t}+A_2·e^{-\text{s}{\dot{\mathrm{V}}}_2·t}$$ (3)

$$A\left(t\right)=A_1·e^{-\text{s}{\dot{\mathrm{V}}}_1·t}+A_2\left(\text{s}{\dot{\mathrm{V}}}_2=0\right)$$ (4)

In these equations, A_i (i = 1, 2) represents the initial activity in each compartment, proportional to the perfusion to that compartment, sV̇_i(i = 1, 2) is the sV̇ of each compartment, and t is time in seconds.

Model selection for each voxel was based on the Akaike Information Criterion (AIC) (32), which was computed as:

$$\text{AIC}=N·\text{In}\left(\text{MSE}\right)+2P$$ (5)

where N is the number of points in the washout curve, P is the number of parameters of the model, and MSE is the mean squared error of the fit. All four models were fit to the voxel washout data using an integral-based approach (18) and least squares linear regression. Models whose fits returned nonphysiological values (i.e., A_i < 0 or sV̇_i < 0) were rejected, and, of the remaining models, the one with the lowest AIC was chosen for each voxel. Use of this criterion ensured selection of the simplest model that provided satisfactory description of washout kinetics in each voxel.

To fit the models, extraneous points with zero activity at the end of each voxel washout curve were discarded; only the first point with activity <2% of the initial washout activity (i.e., end-apnea activity) was retained. This was done to avoid a disproportionately large influence of a zero-valued tail of the washout curve on the mean squared error of the model fits. Additionally, in voxels where a two-compartment model yielded the lowest AIC, but the initial activity in one compartment was <2% of the total initial activity, the voxel was treated as a single compartment with parameters equal to those of the dominant compartment. Only a small fraction of voxels (mean = 0.023% for all animals, maximum = 0.34%) did not return reasonable fits for any of the models and were excluded from the data analysis.

Right-to-left shunt.

Right-to-left shunt fraction was estimated in ROIs based on F_GAS intervals of width 0.1 from 0 to 1. Those ROIs with volume <5 ml were excluded due to low signal-to-noise ratio and minimal contribution to overall tracer exchange. In each ROI, the tracer kinetics during apnea were modeled as previously described (46), with compartments for shunt and gas-exchanging lung. The shunt fraction was computed from the model parameters as the ratio of blood flow to the shunting compartment, divided by total blood flow to the region.

Length-Scale Analysis of Ventilation Heterogeneity

Histograms of sV̇ were constructed from the computed voxel sV̇. For voxels described by two compartments, the fractional contribution of each compartment (f_i) was computed from its relative perfusion [i.e., f_i = A_i/(A₁+ A₂), i = 1,2]. Dispersion of sV̇ was measured after a logarithmic transformation, consistent with previous studies on sV̇ (31, 52, 63, 72). In each animal, total sV̇ heterogeneity (σ_TOT) was quantified as the standard deviation of the log₁₀(sV̇) distribution. It is important to note that this measure of dispersion is independent of the mean value of sV̇ as a result of the logarithmic transformation (33) and is directly related to the coefficient of variation of sV̇ on a linear scale (34). Values of sV̇ < 10⁻³ were set to 10⁻³ to avoid extremely negative numbers, resulting from log₁₀ transformation of small values.

The component of sV̇ heterogeneity occurring below the effective resolution following filtering (12 mm), i.e., the subresolution sV̇ heterogeneity (σ_{<12 mm}), was derived by measuring the change in the dispersion of log₁₀(sV̇), resulting from averaging subvoxel compartments, to obtain a single sV̇ value for each voxel. Because sV̇ was log-normally distributed, we used the geometric mean (i.e., log-average or mean on log scale) rather than the arithmetic mean (mean on linear scale) to average subvoxel compartments. Thus, in all two-compartment voxels, a perfusion-weighted geometric mean sV̇ (sV̇_VOXEL) was derived according to Eq. 6:

$$\text{s}{\dot{\mathrm{V}}}_{\text{VOXEL}}={10}^{\left[A_1·{\log }_{10}\left(\text{s}{\dot{\mathrm{V}}}_1\right)+A_2·{\log }_{10}\left(\text{s}{\dot{\mathrm{V}}}_2\right)\right]/\left[A_1+A_2\right]}$$ (6)

Heterogeneity in voxel-level sV̇ (σ_VOXEL) was computed as the standard deviation of the log₁₀(sV̇_VOXEL) distribution. Finally, σ_{<12 mm} was derived as the square root of the difference between total variance and voxel-level variance: ${\sigma }_{<12\text{mm}}=\sqrt{{\sigma }_{\text{TOT}}^2-{\sigma }_{\text{VOXEL}}^2}$.

The resulting images of sV̇_VOXEL were also used for analysis of larger length scales. Because the images had a different resolution along the cephalo-caudal axis (z-axis), we first separated heterogeneity occurring along that axis into a distinct component. For this, each slice was mean normalized, and the standard deviation of the resulting images was computed, representing exclusively in-plane sV̇ heterogeneity at length scales ≥12 mm (σ_{≥12 mm}). The sV̇ heterogeneity corresponding to the z-axis (σ_Z-AXIS) was computed as: ${\sigma }_{\text{Z-AXIS}}=\sqrt{{\sigma }_{\text{VOXEL}}^2-{\sigma }_{\ge 12\text{mm}}^2}$.

To partition the remaining in-plane sV̇ heterogeneity into distinct length-scale ranges > 12 mm, we applied a previously described progressive filtering technique (65, 70). Starting with images of sV̇_VOXEL mean-normalized by slice, we applied circular geometric moving average filters (taking geometric mean of all voxels in averaging window) with diameters of 12, 36, 60, 84, and 108 mm. For each filtered image, we computed the variance of log₁₀(sV̇_VOXEL) (e.g., σ_{12 mm}², σ_{36 mm}², etc.). Finally, the heterogeneity within each length-scale range was computed as the square root of the difference in variance between progressive filter sizes (e.g., ${\sigma }_{12-36\text{mm}}=\sqrt{{\sigma }_{12\text{mm}}^2-{\sigma }_{36\text{mm}}^2}$). This analysis yielded values for heterogeneity occurring over five length-scale ranges at and above resolution (σ_{12–36 mm}, σ_{36–60 mm}, σ_{60–84 mm}, σ_{84–108 mm}, and σ_{>108 mm}), in addition to the subresolution heterogeneity previously derived (σ_{<12 mm}).

Quantification of Slow Washout Regions

To quantify the effect of slow washout in poorly aerated regions on σ_TOT, we computed the fraction of lung with slow washout (F_SW). This parameter seeks to express the relative volume of lung whose properties are consistent with the presence of airway narrowing or closure due to de-recruitment. We defined the threshold for slow washout as one-tenth of the median sV̇ for each animal (sV̇_MEDIAN/10) to account for differences in the sV̇_MEDIAN among animals, and to use a parameter robust to outliers. The factor 10 requires that values be at least one order of magnitude (one unit on the log₁₀ scale) smaller than the median. Poor aeration was defined as F_GAS < 0.5 (7). Thus F_SW was computed as:

$${\text{F}}_{\text{SW}}=\frac{{\displaystyle \sum {\text{V}}_i\forall \text{s}{\dot{V}}_i<\text{s}{\dot{\mathrm{V}}}_{\text{MEDIAN}}/10and{F}_{\text{GAS},\text{i}}<0.5}}{{\displaystyle \sum {\text{V}}_i}}$$ (7)

where V_i is the volume of each compartment, equal to f_i times voxel volume. To estimate the specific effect of slow washout occurring below the image resolution, we computed the component of F_SW occurring at the subresolution (<12 mm) level (F_{SW<12 mm}). For this parameter, only subvoxel compartments with sV̇ < sV̇_MEDIAN/10 and F_GAS < 0.5 were included in the summation of the numerator of Eq. 7. Finally, to quantify the relation between sV̇ heterogeneity and slow washout, we performed linear regressions of σ_TOT with F_SW and σ_{<12 mm} with F_{SW<12 mm}.

Dependence of Ventilation Heterogeneity on Lung Aeration

The association between sV̇ and aeration was explored by examining distributions of sV̇ in regions with distinct F_GAS. For this, ROIs based on F_GAS bins of width 0.1 from 0 to 1 were used, again excluding those with volume <5 ml. For each ROI, the distribution of log₁₀ (sV̇) was grouped into 15 bins from 10⁻³ to 10⁰, and bin amplitudes were normalized to sum to 1, to observe sV̇ behavior independent of ROI size. To eliminate voxels with decreased F_GAS along the edges of the lung, resulting from partial volume effects and motion artifacts, lung field masks were eroded by one to two voxels in each plane for this analysis.

The relationship between regional ventilation heterogeneity and lung aeration was also explored by calculating components of ventilation heterogeneity at <12-, 12- to 36-, and >36-mm length scales in each of the three isogravitational ROIs. We did not partition length scales >60 mm in this analysis, because such dimensions would approach or exceed the largest dimensions of the ROIs along the ventral-dorsal and latero-lateral axes. The linear regression of each length-scale component against the mean F_GAS of the ROIs was computed for all ROIs from all animals.

Statistical Analysis

Statistical analysis was performed using Matlab software (MATLAB R2011a Statistics Toolbox, The Mathworks, Natick, MA). Data are expressed as means ± SD. Parameters were compared among groups using either one-way ANOVA for normally distributed data or Kruskal-Wallis test otherwise, with Tukey-Kramer correction for multiple comparisons in post hoc tests. The Pearson product-moment correlation coefficient was used to quantify linear correlations. Significance was set at P < 0.05.

RESULTS

Global Physiological Variables

The designed ventilatory settings lead to clearly distinct Vt and RRs in the high-Vt ZEEP group and PEEP levels in the low-Vt PEEP group (Table 1). The low-Vt PEEP group had lower cardiac output and higher blood oxygen and carbon dioxide levels than the low-Vt ZEEP group.

Table 1.

Global respiratory and hemodynamic variables

	High-Vt ZEEP	Low-Vt ZEEP	Low-Vt PEEP
n	6	6	4
Weight, kg	25 ± 7	22 ± 3	20 ± 1
Vt, ml/kg	18.4 ± 4.2##	9.2 ± 1.0	8.2 ± 0.2
RR, beats/min	19 ± 2###	26 ± 3	26 ± 1
V̇a, ml·min−1·kg−1	147 ± 44**	104 ± 24	66 ± 7
Raw, cmH2O·l−1·s	23.0 ± 11.5	12.7 ± 6.5	15.1 ± 4.4
Cdyn, ml/cmH2O	16.7 ± 6.9	11.5 ± 1.8	18.6 ± 3.2
PEEP, cmH2O	0 ± 0***	0 ± 0***	19 ± 1
MAP, mmHg	92 ± 12	97 ± 4	88 ± 14
CO, l/min	4.4 ± 1.1	5.4 ± 1.1*	3.1 ± 0.9
FiO2	0.33 ± 0.08	0.35 ± 0.08	0.33 ± 0.05
PaO2, Torr	93 ± 36	85 ± 34*	128 ± 14
PaCO2, Torr	33 ± 9	34 ± 3*	40 ± 3

Values are means ± SD; n, no. of animals. Vt, tidal volume; ZEEP, zero end-expiratory pressure; PEEP, positive end-expiratory pressure (ZEEP, PEEP = 0); RR, respiratory rate; V̇a, alveolar minute ventilation; Raw, airway resistance; Cdyn, dynamic compliance; MAP, mean arterial pressure; CO, cardiac output; Fi_O2, inspired oxygen fraction; Pa_O2, arterial oxygen pressure; Pa_CO2, arterial carbon dioxide pressure.

^##P < 0.01 and

^###P < 0.001 vs. other groups.

^*P < 0.05,

^**P < 0.01, and

^***P < 0.001 vs. low-Vt PEEP. Measurements of V̇a, Raw, and Cdyn were available for 4 high-Vt ZEEP, 5 low-Vt ZEEP, and all low-Vt PEEP animals.

Lung Aeration

Use of PEEP had notable influence on aeration of the lungs. The low-Vt PEEP group had the highest mean F_GAS (P < 0.01 vs. high-Vt ZEEP, P < 0.001 vs. low-Vt ZEEP) and lower standard deviation of F_GAS compared with high-Vt ZEEP (P < 0.01; Table 2). Global F_GAS properties were similar between the ZEEP groups (Table 2). Regional F_GAS measurements revealed that dependent lung regions were the main source of reduced global F_GAS in both ZEEP groups (Fig. 1). Of note, the low-Vt PEEP group had significantly higher mean F_GAS in middle and dependent regions than either low-Vt ZEEP or high-Vt ZEEP (Fig. 1).

Table 2.

Global aeration, ventilation heterogeneity, and slow washout

	High-Vt ZEEP	Low-Vt ZEEP	Low-Vt PEEP
Mean FGAS	0.55 ± 0.12**	0.51 ± 0.05***	0.76 ± 0.03
SD FGAS	0.20 ± 0.06**	0.15 ± 0.02	0.10 ± 0.03
σTOT	0.62 ± 0.17*	0.42 ± 0.22	0.24 ± 0.03
σz-axis	0.18 ± 0.13#	0.10 ± 0.04	0.06 ± 0.03
FSW	0.09 ± 0.05*	0.06 ± 0.07	0.00 ± 0.00
FSW<12 mm	0.06 ± 0.04*	0.04 ± 0.05	0.00 ± 0.00

Values are means ± SD. F_GAS, fractional gas content (average over respiratory cycle); σ_TOT, total specific ventilation heterogeneity; σ_z-axis, specific ventilation heterogeneity occurring along the z-axis; F_SW, fraction of lung volume with slow washout; F_{SW<12 mm}, subresolution component of F_SW.

^#P < 0.10,

^*P < 0.05,

^**P < 0.01, and

^***P < 0.001 vs. low- Vt PEEP.

Fig. 1.

Mean regional fractional gas content (F_GAS; average over respiratory cycle) in three isogravitational regions of interest. Values are shown as means ± SD for each of 3 groups, ventilated with high tidal volume (Vt) and zero end-expiratory pressure (ZEEP; High Vt ZEEP), low Vt and ZEEP (Low Vt ZEEP), or low Vt and high positive end-expiratory pressure (Low Vt PEEP). **P < 0.01. ***P < 0.001.

Ventilation Heterogeneity

Patterns of tracer washout were distinct between the ZEEP and PEEP groups. Several animals in the ZEEP groups exhibited regions of slow washout (tracer retention) in the poorly aerated, dependent lung regions, whereas PEEP animals showed minimal tracer retention (Fig. 2). Low sV̇ regions contributed significantly to the greater dispersion in sV̇ distributions in those ZEEP animals (Fig. 2). Following this trend, σ_TOT was significantly higher in high-Vt ZEEP than in low-Vt PEEP (P < 0.05), with low-Vt ZEEP falling in between (Table 2). σ_Z-AXIS was not significantly different between groups, although there was a trend (P < 0.10) toward higher σ_Z-AXIS in the high-Vt ZEEP group than in the low-Vt PEEP group (Table 2).

Fig. 2.

Images of average F_GAS, tracer retention (percentage of end-apnea activity remaining at end of washout), specific ventilation (sV̇), voxel washout pattern, and sV̇ distributions. One representative animal is shown for each group. Voxel washout patterns are single-exponential (1 Exp), double-exponential (2 Exp), partial gas trapping (PGT), and complete gas trapping (CGT). Note the homogeneous distributions of F_GAS and sV̇ and predominance of 1 Exp washout patterns in the Low Vt PEEP animal.

The frequencies of the different voxel washout patterns were affected by the mode of ventilation (Fig. 3). The low-Vt PEEP group had a significantly greater fraction of voxels with single-exponential washout (P < 0.05) and a significantly lower fraction of voxels with partial gas trapping (P < 0.05) than the high-Vt ZEEP group. There was notable variability in the voxel classification of animals in the low-Vt ZEEP group, with some presenting mostly single-exponential voxels, while others showed higher fractions of double-exponential and partial gas trapping voxels. In both ZEEP groups, the fractions of double-exponential (P < 0.001) and partial gas trapping (P < 0.05) voxels were significantly greater than zero, whereas neither of those two-compartment voxel types were significantly different from zero in the low-Vt PEEP group.

Fig. 3.

Fractions of voxels classified according to each washout pattern for the 3 studied groups. Bars represent median values. The High Vt ZEEP group had significantly less single exponential voxels and significantly more partial trapping voxels than Low Vt PEEP. Animals in the Low Vt ZEEP group showed variable behavior. *P < 0.05.

Length-scale components of ventilation heterogeneity.

Ventilation heterogeneity was present at all length scales studied (Fig. 4). The distributions of sV̇ heterogeneity across length scales were different among the groups. The high-Vt ZEEP group had significantly higher sV̇ heterogeneity than the low-Vt PEEP group at length scales of <12 (P < 0.01), 12–36 (P < 0.01), and 36–60 mm (P < 0.05). Animals in the low-Vt ZEEP group showed considerable variability in patterns of length-scale components.

Fig. 4.

Components of sV̇ heterogeneity occurring over different length scales. The High Vt ZEEP group had higher heterogeneity at length scales of <12 and 12–36 mm compared with the larger length scales, and higher heterogeneity than the Low Vt PEEP group at length scales of <12, 12–36, and 36–60 mm. Animals in the Low Vt ZEEP group showed variable patterns, with some similar to High Vt ZEEP animals and others similar to Low Vt PEEP animals. *P < 0.05. **P < 0.01. ***P < 0.001. ^#P < 0.05 and ^##P < 0.01 vs. High Vt ZEEP.

In the high-Vt ZEEP group, sV̇ heterogeneity was significantly higher at length scales of <12 and 12–36 mm than at the larger length scales (Fig. 4). There were no significant differences between length scales for the low-Vt ZEEP or low-Vt PEEP groups, with low-Vt PEEP showing consistently low sV̇ heterogeneity at all length scales.

Relation of ventilation heterogeneity with slow washout, aeration, and shunt.

The F_SW and the F_{SW<12 mm} were both larger in the high-Vt ZEEP group than in the low-Vt PEEP group (P < 0.05, Table 2). We found a strong linear correlation (r = 0.95, P < 0.001) between σ_TOT and F_SW in each animal (Fig. 5A). Additionally, we examined the relationship between the components of these parameters occurring at the subresolution length scale, i.e., σ_{<12 mm} and F_{SW<12 mm}. These parameters were also correlated (r = 0.84, P < 0.001) (Fig. 5B), showing that the observed relationship between these variables measured over all length scales was still present exclusively at the <12-mm length scale. Those correlations were still significant when animals from the low-Vt PEEP group, which had zero F_SW and F_{SW<12 mm} (Table 2), were removed from the analysis (r = 0.94, P < 0.001 for Fig. 5A; r = 0.78, P < 0.01 for Fig. 5B). Interestingly, we also found correlations of global Raw with both F_SW (r = 0.81, P < 0.001, n = 13) and σ_TOT (r = 0.75, P < 0.01, n = 13).

Fig. 5.

A: total sV̇ heterogeneity vs. fraction of lung with slow washout (F_SW). A strong correlation (r = 0.95, P < 0.001) was found between these parameters, with animals from all groups following the same trend. B: subresolution (<12 mm) sV̇ heterogeneity was also correlated (r = 0.84, P < 0.001) with the subresolution component of F_SW (F_{SW<12 mm}). The positive sV̇ heterogeneity in both A and B for animals with little or zero slow washout and the positive y-intercepts of the regression lines support the presence of sV̇ heterogeneity not associated with slow washout.

Two-dimensional binning of regional aeration and ventilation revealed that poorly aerated regions contributed significantly to the increased sV̇ heterogeneity observed in most animals in the ZEEP groups (Fig. 6). In particular, the emergence of a bimodal distribution in the sV̇ vs. F_GAS plane was noted in all animals in the high-Vt ZEEP group. This bimodality was characterized by the presence of one mode corresponding to regions of normal sV̇ and aeration and another mode exhibiting reduced sV̇ in poorly aerated regions. There was variability in these patterns among animals in the low-Vt ZEEP group, with some showing a normal range of sV̇ values for F_GAS < 0.5, and others showing patterns similar to the high-Vt ZEEP group. None of the low-Vt PEEP animals demonstrated bimodal behavior, instead showing a single mode with normal aeration (F_GAS > 0.5) and sV̇ values in a normal range.

Fig. 6.

Distributions of sV̇ and shunt fraction computed in regions of interest (ROIs) based on average F_GAS bins of width 0.1 from 0 to 1, in the same representative animals from Fig. 2. Within each F_GAS ROI, distributions of sV̇ are normalized to sum to 1; thus gray-scale indicates fraction of ROI in each sV̇ range. Dashed lines outline regions with poor aeration (F_GAS < 0.5) and slow washout (sV̇ <sV̇_MEDIAN/10). In plots of shunt fraction, solid lines show means ± SD for all animals in the group, with open circles denoting value for the representative animal. Note the bimodal distributions of sV̇ for F_GAS < 0.5 in the High Vt ZEEP and Low Vt ZEEP groups and similar trends of progressively increasing shunt for F_GAS < 0.5. Low Vt PEEP animals had minimal lung with F_GAS < 0.5.

A similar relationship between shunt fraction and F_GAS was found for the ZEEP groups in those F_GAS ROIs (Fig. 6). Shunt was consistently low in ROIs with F_GAS between 0.5 and 1 and increased progressively for F_GAS < 0.5 in the ZEEP groups. Importantly, for ROIs with F_GAS < 0.5, the appearance of shunt coincided with the emergence of reduced sV̇ regions. The low-Vt PEEP group had very little lung with F_GAS < 0.5, although a small increase in shunt was apparent for F_GAS < 0.5.

Analysis based on three isogravitational ROIs confirmed the association between sV̇ heterogeneity and regional aeration. Negative correlations were found between regional mean F_GAS and sV̇ heterogeneity at length scales of <12 mm (r = −0.67, P < 0.001), 12–36 mm (r = −0.74, P < 0.001), and >36 mm (r = −0.72, P < 0.001) (Fig. 7A). Additionally, there was a negative correlation between the standard deviation of F_GAS and mean F_GAS (r = −0.71, P < 0.001) (Fig. 7B). Interestingly, dependent regions in PEEP animals exhibited higher aeration and lower sV̇ heterogeneity at all three length scales than dependent regions of ZEEP animals, implying that regional sV̇ heterogeneity is determined by aeration and not anatomic location.

Fig. 7.

A: sV̇ heterogeneity at length scales of <12, 12–36, and >36 mm. B: standard deviation of F_GAS (F_GAS SD) vs. mean F_GAS in three isogravitational ROIs in each animal. Nondependent, middle, and dependent ROIs are depicted in solid, shaded, and open symbols, respectively. Dashed lines are fit to all data points. Note the high mean F_GAS and low sV̇ heterogeneity at each length scale in all ROIs of the Low Vt PEEP group.

DISCUSSION

In mechanically ventilated, paralyzed supine sheep with normal lungs, we found the following: 1) sV̇ heterogeneity was present at all length scales studied, including below the image resolution. During mechanical ventilation with high Vt and ZEEP, sV̇ heterogeneity was higher at length scales of <12 and 12–36 mm compared with both larger length scales in the same group and the same length scales in the low-Vt PEEP group. 2) σ_TOT was highly correlated with the F_SW, with higher sV̇ heterogeneity reflecting the emergence of bimodal sV̇ patterns in poorly aerated lung regions. 3) Components of regional sV̇ heterogeneity at length scales of <12, 12–36, and >36 mm were inversely related to regional lung aeration. 4) Application of high PEEP with low Vt led to lower regional sV̇ heterogeneity at length scales of <12, 12–36, and >36 mm by maintaining normal aeration in all lung regions.

Length Scales of Ventilation Heterogeneity

Previous studies during anesthesia, mechanical ventilation, and muscle paralysis observed changes in ventilation distribution (71), with partial or complete airway collapse and atelectasis (20, 24, 54, 59, 60). However, those studies were mostly based on whole lung measurements, providing limited information on the length scales over which ventilation heterogeneity occurs. Although studies using fluorescent microspheres showed increasing ventilation heterogeneity down to 2 cm³ in pigs (2), it has been presumed that homogenization occurs within that volume in normal lungs (57). In contrast, preliminary measurements from high-resolution synchrotron CT in rabbits, whose alveolar mean linear intercept is only ∼38% smaller than that of pigs (35), suggested that ventilation heterogeneity occurs down to volumes of 1 mm³ (52).

Our results demonstrate a significant contribution of length scales <60 mm to total ventilation heterogeneity during mechanical ventilation of supine normal lungs. In fact, heterogeneity at length scales of <12, 12–36, and >36 mm was present, even in well-inflated lung regions. These findings are consistent with studies in spontaneously breathing normal humans in the sitting position showing heterogeneous intraregional mixing (9, 10, 17, 50). Such heterogeneity inherent to the lung structure has been attributed to asymmetry in airway and alveolar geometry (10, 49) and likely contributes to the isogravitational heterogeneity observed in previous studies (52, 58).

Mechanisms of ventilation heterogeneity.

In addition to the sV̇ heterogeneity observed in well-inflated lung regions, our data show that components of sV̇ heterogeneity at length scales of <12, 12–36, and >36 mm increase at low levels of aeration. As lung inflation is reduced, increased ventilation heterogeneity may be due to changes in airway diameters, affecting local airway resistances, or in alveolar inflation, affecting local elastance. Our results suggest an important role of airways, particularly small airways, in the observed increases in sV̇ heterogeneity. First, we found strong correlations between σ_TOT and F_SW. Since airway narrowing or closure is the most likely cause of slow washout in poorly aerated regions of normal lungs (8), this finding implies that reduction in airway diameters is a key factor contributing to increased sV̇ heterogeneity. The fact that there was also a correlation between σ_{<12 mm} and F_{SW<12 mm} further suggests that small airways were involved [i.e., those underlying the volumetric resolution of 1.66 cm³, likely <1 mm in diameter (75)]. Second, we observed bimodal distributions of sV̇ in poorly aerated regions of several ZEEP animals. This finding is consistent with previous findings of airway instability due to local feedback mechanisms in a model of a single terminal airway (3) and the emergence of heterogeneous airway narrowing in an airway tree model (67). In our study, the emergence of airway narrowing or closure could have been triggered by low lung inflation and thus loss of parenchymal tethering. This mechanism would tend to affect the small (<2 mm) airways, since they are not supported by cartilage (44), and thus rely on interdependence to remain open. Finally, the presence of regions with complete tracer retention (represented by a plateau at the end of the time-activity curve) suggests that complete airway closure occurred in those regions. This result is consistent with data from excised normal rabbit lungs, suggesting airway closure in conditions of lung derecruitment underlying dimensions of 2 cm³ (27). Overall, our findings indicate a predominant role of the dependence between regional airway narrowing and closure and regional aeration in determining ventilation distribution in normal lungs.

Some of the observed slow washout rates could represent intermittent airway opening and closure. This mechanism has been invoked to explain regions of low ventilation-to-perfusion ratios in anesthetized patients with normal lungs (23). During ventilation with ZEEP, the presence of poor aeration in some regions would predispose the airways to closure as a result of low parenchymal tethering forces. In this condition, the use of high Vt would increase the likelihood of intermittent airway reopening because 1) higher end-inspiratory pressures and longer inspiratory periods associated with high Vt would increase the probability of airway reopening during inspiration (1, 4, 11, 51); and 2) mechanical ventilation with high Vt and ZEEP has been shown to increase surface tension after only 1 h (36, 64), thus promoting airway instability (30) and facilitating the formation of liquid bridges (26, 47). In the low-Vt ZEEP condition, lower peak inspiratory pressures, combined with shorter inspiratory times, would be less likely to reopen airways (4). Indeed, these presumptions are in line with our findings of higher 1) partial gas trapping (Fig. 3); 2) heterogeneity at small length scales (Fig. 4); and 3) F_SW (Table 2) in the high-Vt ZEEP condition compared with low-Vt PEEP.

Nonuniformity of tissue elastance, as may occur at low lung volumes, is another potential cause of increased ventilation heterogeneity. Measurements based on small metallic markers indicated considerable variability of lung compliance within small regions of excised dog lungs (45). More recently, imaging studies using intravital microscopy suggested the presence of heterogeneous expansion at the alveolar level (6, 39). Such local differences in alveolar inflation patterns would be expected to result in variations in sV̇ at the subresolution level, potentially contributing to the increased sV̇ heterogeneity in poorly aerated regions. Additionally, regions of high inflation may demonstrate increased elastance due to overdistension. This condition would be most likely in the low-Vt-PEEP group and may have contributed to sV̇ heterogeneity at large length scales.

The increased shunt fraction observed in poorly aerated regions provides direct evidence of alveolar collapse, since only nonaerated alveoli would allow ¹³NN to pass through the lungs without retention. Local alveolar collapse can lead to loss of parenchymal tethering forces on small airways, resulting in nonuniform narrowing or closure. In addition, the presence of alveolar collapse supports the concept of ventilation heterogeneity caused by heterogeneous alveolar inflation.

Interaction between diffusion and convection in the presence of asymmetric airway branching and cross-sectional areas likely contributed to a portion of the small length-scale ventilation heterogeneity in all lung regions (10, 49). It is possible that, in poorly aerated regions, asymmetry of airway and alveolar geometry increased as a result of heterogeneous inflation, a presumption that is supported by the negative correlation between F_GAS heterogeneity and mean F_GAS (Fig. 7B). Such increased asymmetry may have contributed to the observed increase in sV̇ heterogeneity in poorly aerated regions. In addition to airway resistance and tissue elastance, other mechanical properties, such as tissue hysteresivity, could affect ventilation heterogeneity. However, hysteresivity is thought to be homogeneously distributed in the normal lung (61) and thus should not promote significant local increase in heterogeneity.

Importantly, the discussed mechanisms producing ventilation heterogeneity, i.e., cyclic recruitment of airways (12, 42, 48) and concentration of mechanical forces due to heterogeneous lung expansion (37), have been proposed as relevant mechanisms of injury during mechanical ventilation. Future studies should address the relationship between length-scale components of ventilation heterogeneity and markers of injury in experimental models of lung injury.

Effects of PEEP on Length-Scale Components of Ventilation Heterogeneity

The benefits of PEEP for gas exchange have been attributed to its effect on the homogenization of air distribution across large lung regions (29). In mechanically ventilated normal lungs, changing PEEP from 0 to 6 cmH₂O in rabbits (62) or 3 to 6 cmH₂O in children (71) increased FRC and reduced total ventilation heterogeneity. However, because those findings were based on global physiological measurements, they do not characterize the topographical source of the effect of PEEP.

Our results imply a substantial contribution of smaller length scales for the beneficial effect of PEEP on ventilation distribution. The insignificant number of voxels showing two-compartment behavior (i.e., double-exponential or partial trapping kinetics) in this group, and the lower sV̇ heterogeneity at length scales of <12, 12–36, and 36–60 mm compared with the high-Vt ZEEP group, suggest an effect of peripheral homogenization of ventilation. Moreover, our data on bimodal sV̇ behavior in regions of low F_GAS (Fig. 6) and the inverse relationships between length-scale components of ventilation heterogeneity and F_GAS (Fig. 7) indicate a key role of the degree of regional lung inflation for the effect of PEEP on ventilation heterogeneity. By maintaining normal aeration in all lung regions, PEEP likely also preserved parenchymal tethering forces on small airways, preventing the generation of slow washout regions associated with airway narrowing or closure at the lower range of studied length scales. Such an effect of PEEP in reducing ventilation heterogeneity by increasing parenchymal tethering forces is consistent with findings in bronchoconstricted rabbit lungs (53). Because PEEP can also reduce cardiac output and change the distribution of pulmonary perfusion, the ultimate effect of PEEP on global gas exchange will depend, not only on the reduction in sV̇ heterogeneity, but also on its effects on hemodynamics and ventilation-perfusion distribution (15, 25).

Effects of Vt on Length-Scale Components of Ventilation Heterogeneity

We found no significant differences in ventilation heterogeneity between high and low Vt during mechanical ventilation with ZEEP. Previous studies using multiple-breath washout techniques in spontaneously breathing subjects suggested a decrease in intraregional (small length scale) and increase in interregional (large length scale) ventilation heterogeneity with larger Vt (9, 21). In contrast, studies in mechanically ventilated normal lungs showed no effect of Vt on ventilation heterogeneity measured with multiple-breath washout (31) or synchrotron computed tomography (52), in line with our findings.

Methodological Aspects

The ¹³NN-PET technique has been demonstrated to be very sensitive for detection of ventilation heterogeneity, both below and above the imaging resolution (68, 69). Modeling of two compartments within each voxel allows for quantification of ventilation heterogeneity underlying the effective volumetric resolution. Because subresolution sV̇ heterogeneity is quantified through its functional properties (i.e., ¹³NN washout kinetics), we cannot establish what anatomical level (e.g., smaller or larger than the acinus) is responsible for this component of heterogeneity. It is theoretically possible to model a voxel with more than two compartments, but in practice such analysis is precluded by the presence of noise in voxel-level kinetics (68). This limitation was unlikely to influence our findings, as the use of two compartments to model subresolution ventilation heterogeneity has been sufficient to describe gas exchange in normal and disease states in previous studies (68, 69).

We quantified dispersion of sV̇ using the standard deviation of log₁₀(sV̇), consistent with previous studies on sV̇ (31, 52, 63, 72). Importantly, this measure of dispersion is independent of the mean sV̇ as a result of the logarithmic transformation (33). Such property is particularly important in this study, because there were significant differences between the groups in alveolar minute ventilation, one of the main determinants of mean sV̇.

Although ketamine has been shown to have bronchodilatory effects during bronchoconstriction (5, 28), we do not believe those effects influenced our results because 1) ketamine has been shown to have no effect in the absence of bronchoconstriction (5, 28); and 2) the same doses were used in each group, so differences between groups would not be explained by the use of ketamine.

Transmission and ¹³NN washout images were acquired during continuous breathing. As a result, our images of F_GAS represent average values over the entire respiratory cycle. Although it is possible to obtain images of F_GAS at end expiration and end inspiration with ¹³NN-PET (73), for the purpose of relating aeration to ventilation, images of average F_GAS allowed direct comparison with ¹³NN washout images, which were also obtained over several breaths. Lung motion has the effect of spatial averaging over the range of motion and has been shown to reduce estimates of sV̇ heterogeneity derived from PET images, presumably at length scales comparable to the distance of motion (40). Thus our estimates of sV̇ heterogeneity at smaller length scales (e.g., <12, 12–36 mm) may have been underestimated due to lung motion.

A fraction of the measured sV̇ heterogeneity was likely attributable to imaging noise. To reduce the contribution of noise to heterogeneity, we filtered images both in-plane, as well as along the z-axis. Residual noise would likely have influenced the smaller length scales more than the large ones. However, since the same methodology was used to estimate sV̇ heterogeneity for all animals, the differences between animals or groups should reflect real physiological differences rather than distinct noise levels. Additionally, our approach of fitting washout models to several data points (n = 10) allows for robust estimation of sV̇, since several images with uncorrelated noise characteristics are used for parameter estimation.

Delivery of tracer through perfusion provides the important advantage of being able to distinguish between regions of zero ventilation due to either alveolar collapse or airway closure, based on tracer kinetics patterns. In regions with alveolar collapse, tracer will pass through the lungs and demonstrate shunting kinetics (46), whereas, with airway closure, the tracer will diffuse into aerated alveoli and remain there during washout (68). The ability to distinguish these patterns is particularly important for characterization of ventilation at low lung volumes, where collapse of alveoli and airways may occur simultaneously. Other techniques to measure ventilation that administer tracer via inhalation (50, 57, 58) are unable to distinguish between regions of alveolar collapse or airway closure.

Conclusions

We found in paralyzed, supine normal sheep that increased ventilation heterogeneity during mechanical ventilation with high Vt and ZEEP occurs primarily at length scales <60 mm, with a significant component derived from subresolution (<12 mm) length scales. Components of sV̇ heterogeneity at length scales of <12, 12–36, and 36–60 mm were highest in poorly aerated regions. Total ventilation heterogeneity was highly correlated with the F_SW, suggesting a role of airway narrowing or closure in regions of increased sV̇ heterogeneity. Finally, high levels of PEEP reduced the <60 mm length-scale components of ventilation heterogeneity by increasing lung inflation in all regions to normal levels, with effects on both alveolar inflation and airway patency. In future studies, the topographical combination of regional aeration and length-scale analysis of ventilation heterogeneity may be helpful in identifying lung regions susceptible to VILI.

GRANTS

This work was supported by grant 5R01HL086827 from the National Heart, Lung, and Blood Institute. G. Musch was supported in part by grant R01 HL-094639.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: T.J.W., T.W., E.L.V.C., G.M., R.S.H., and M.F.V.M. performed experiments; T.J.W. and M.F.V.M. analyzed data; T.J.W., T.W., E.L.V.C., G.M., R.S.H., J.G.V., and M.F.V.M. interpreted results of experiments; T.J.W. and M.F.V.M. prepared figures; T.J.W., T.W., and M.F.V.M. drafted manuscript; T.J.W., T.W., E.L.V.C., G.M., R.S.H., J.G.V., and M.F.V.M. edited and revised manuscript; T.J.W., T.W., E.L.V.C., G.M., R.S.H., J.G.V., and M.F.V.M. approved final version of manuscript; J.G.V. and M.F.V.M. conception and design of research.