Impact of positive end-expiratory pressure during heterogeneous lung injury: insights from computed tomographic image functional modeling

Multi-resolution convolutional neural networks for fully automated segmentation of acutely injured lungs in multiple species

Segmentation of lungs with acute respiratory distress syndrome (ARDS) is a challenging task due to diffuse opacification in dependent regions which results in little to no contrast at the lung boundary. For segmentation of severely injured lungs, local intensity and texture information, as well as global contextual information, are important factors for consistent inclusion of intrapulmonary structures. In this study, we propose a deep learning framework which uses a novel multi-resolution convolutional neural network (ConvNet) for automated segmentation of lungs in multiple mammalian species with injury models similar to ARDS. The multi-resolution model eliminates the need to tradeoff between high-resolution and global context by using a cascade of low-resolution to high-resolution networks. Transfer learning is used to accommodate the limited number of training datasets. The model was initially pre-trained on human CT images, and subsequently fine-tuned on canine, porcine, and ovine CT images with lung injuries similar to ARDS. The multi-resolution model was compared to both high-resolution and low-resolution networks alone. The multi-resolution model outperformed both the low- and high-resolution models, achieving an overall mean Jacaard index of 0.963 ± 0.025 compared to 0.919 ± 0.027 and 0.950 ± 0.036, respectively, for the animal dataset (N=287). The multi-resolution model achieves an overall average symmetric surface distance of 0.438 ± 0.315 mm, compared to 0.971 ± 0.368 mm and 0.657 ± 0.519 mm for the low-resolution and high-resolution models, respectively. We conclude that the multi-resolution model produces accurate segmentations in severely injured lungs, which is attributed to the inclusion of both local and global features.

Impact of ventilation frequency and parenchymal stiffness on flow and pressure distribution in a canine lung model

To determine the impact of ventilation frequency, lung volume, and parenchymal stiffness on ventilation distribution, we developed an anatomically-based computational model of the canine lung. Each lobe of the model consists of an asymmetric branching airway network subtended by terminal, viscoelastic acinar units. The model allows for empiric dependencies of airway segment dimensions and parenchymal stiffness on transpulmonary pressure. We simulated the effects of lung volume and parenchymal recoil on global lung impedance and ventilation distribution from 0.1 to 100 Hz, with mean transpulmonary pressures from 5 to 25 cm H2O. With increasing lung volume, the distribution of acinar flows narrowed and became more synchronous for frequencies below resonance. At higher frequencies, large variations in acinar flow were observed. Maximum acinar flow occurred at first antiresonance frequency, where lung impedance achieved a local maximum. The distribution of acinar pressures became very heterogeneous and amplified relative to tracheal pressure at the resonant frequency. These data demonstrate the important interaction between frequency and lung tissue stiffness on the distribution of acinar flows and pressures. These simulations provide useful information for the optimization of frequency, lung volume, and mean airway pressure during conventional ventilation or high frequency oscillation (HFOV). Moreover our model indicates that an optimal HFOV bandwidth exists between the resonant and antiresonant frequencies, for which interregional gas mixing is maximized.

Optimizing positive end-expiratory pressure by oscillatory mechanics minimizes tidal recruitment and distension: an experimental study in a lavage model of lung injury

Introduction

It is well established that during mechanical ventilation of patients with acute respiratory distress syndrome cyclic recruitment/derecruitment and overdistension are potentially injurious for lung tissues. We evaluated whether the forced oscillation technique (FOT) could be used to guide the ventilator settings in order to minimize cyclic lung recruitment/derecruitment and cyclic mechanical stress in an experimental model of acute lung injury.

Methods

We studied six pigs in which lung injury was induced by bronchoalveolar lavage. The animals were ventilated with a tidal volume of 6 ml/kg. Forced oscillations at 5 Hz were superimposed on the ventilation waveform. Pressure and flow were measured at the tip and at the inlet of the endotracheal tube respectively. Respiratory system reactance (Xrs) was computed from the pressure and flow signals and expressed in terms of oscillatory elastance (E_X5). Positive end-expiratory pressure (PEEP) was increased from 0 to 24 cm H₂O in steps of 4 cm H₂O and subsequently decreased from 24 to 0 in steps of 2 cm H₂O. At each PEEP step CT scans and E_X5 were assessed at end-expiration and end-inspiration.

Results

During deflation the relationship between both end-expiratory and end-inspiratory E_X5 and PEEP was a U-shaped curve with minimum values at PEEP = 13.4 ± 1.0 cm H₂O (mean ± SD) and 13.0 ± 1.0 cm H₂O respectively. E_X5 was always higher at end-inspiration than at end-expiration, the difference between the average curves being minimal at 12 cm H₂O. At this PEEP level, CT did not show any substantial sign of intra-tidal recruitment/derecruitment or expiratory lung collapse.

Conclusions

Using FOT it was possible to measure E_X5 both at end-expiration and at end-inspiration. The optimal PEEP strategy based on end-expiratory E_X5 minimized intra-tidal recruitment/derecruitment as assessed by CT, and the concurrent attenuation of intra-tidal variations of E_X5 suggests that it may also minimize tidal mechanical stress.

Oscillation mechanics of the respiratory system: applications to lung disease

Since its introduction in the 1950s, the forced oscillation technique (FOT) and the measurement of respiratory impedance have evolved into powerful tools for the assessment of various mechanical phenomena in the mammalian lung during health and disease. In this review, we highlight the most recent developments in instrumentation, signal processing, and modeling relevant to FOT measurements. We demonstrate how FOT provides unparalleled information on the mechanical status of the respiratory system compared to more widely used pulmonary function tests. The concept of mechanical impedance is reviewed, as well as the various measurement techniques used to acquire such data. Emphasis is placed on the analysis of lower, physiologic frequency ranges (typically less than 10 Hz) that are most sensitive to normal physical processes as well as pathologic structural alterations. Various inverse modeling approaches used to interpret alterations in impedance are also discussed, specifically in the context of three common respiratory diseases: asthma, chronic obstructive pulmonary disease, and acute lung injury. Finally, we speculate on the potential role for FOT in the clinical arena.

Emergent behavior of regional heterogeneity in the lung and its effects on respiratory impedance

The ability to maintain adequate gas exchange depends on the relatively homogeneous distribution of inhaled gas throughout the lung. Structural alterations associated with many respiratory diseases may significantly depress this function during tidal breathing. These alterations frequently occur in a heterogeneous manner due to complex, emergent interactions among the many constitutive elements of the airways and parenchyma, resulting in unique signature changes in the mechanical impedance spectrum of the lungs and total respiratory system as measured by forced oscillations techniques (FOT). When such impedance spectra are characterized by appropriate inverse models, one may obtain functional insight into derangements in global respiratory mechanics. In this review, we provide an overview of the impact of structural heterogeneity with respect to dynamic lung function. Recent studies linking functional impedance measurements to the structural heterogeneity observed in acute lung injury, asthma, and chronic obstructive pulmonary disease are highlighted, as well as current approaches for the modeling and interpretation of impedance. Finally, we discuss the potential diagnostic role of FOT in the context of therapeutic interventions.

Analysis of regional mechanics in canine lung injury using forced oscillations and 3D image registration

Acute lung injury is characterized by heterogeneity of regional mechanical properties, which is thought to be correlated with disease severity. The feasibility of using respiratory input impedance (Z_rs) and computed tomographic (CT) image registration for assessing parenchymal mechanical heterogeneity was evaluated. In six dogs, measurements of Z_rs before and after oleic acid injury at various distending pressures were obtained, followed by whole lung CT scans. Each Z_rs spectrum was fit with a model incorporating variable distributions of regional compliances. CT image pairs at different inflation pressures were matched using an image registration algorithm, from which distributions of regional compliances from the resulting anatomic deformation fields were computed. Under baseline conditions, average model compliance decreased with increasing inflation pressure, reflecting parenchymal stiffening. After lung injury, these average compliances decreased at each pressure, indicating derecruitment, alveolar flooding, or alterations in intrinsic tissue elastance. However, average compliance did not change as inflation pressure increased, consistent with simultaneous recruitment and strain stiffening. Image registration revealed peaked distributions of regional compliances, and that small portions of the lung might undergo relative compression during inflation. The authors conclude that assessments of lung function using Z_rs combined with the structural alterations inferred from image registration provide unique but complementary information on the mechanical derangements associated with lung injury.

Measurement of regional specific lung volume change using respiratory-gated PET of inhaled 13N-nitrogen

Regional specific lung volume change (sVol), defined as the regional tidal volume divided by the regional end-expiratory gas volume, is a key variable in lung mechanics and in the pathogenesis of ventilator-induced lung injury. Despite the usefulness of PET to study regional lung function, there is no established method to assess sVol with PET. We present a method to measure sVol from respiratory-gated PET images of inhaled (13)N-nitrogen ((13)NN), validate the method against regional specific ventilation (sV), and study the effect of region-of-interest (ROI) volume and orientation on the sVol-sV relationship.

METHODS

Four supine sheep were mechanically ventilated (tidal volume V(T) = 8 mL/kg, respiratory rate adjusted to normocapnia) at low (n = 2, positive end-expiratory pressure = 0) and high (n = 2, positive end-expiratory pressure adjusted to achieve a plateau pressure of 30 cm H(2)O) lung volumes. Respiratory-gated PET scans were obtained after inhaled (13)NN equilibration both at baseline and after a period of mechanical ventilation. We calculated sVol from (13)NN-derived regional fractional gas content at end-inspiration (F(EI)) and end-expiration (F(EE)) using the formula sVol = (F(EI) - F(EE))/(F(EE)[1 - F(EI)]). sV was computed as the inverse of the subsequent (13)NN washout curve time constant. ROIs were defined by dividing the lung field with equally spaced coronal, sagittal, and transverse planes, perpendicular to the ventrodorsal, laterolateral, and cephalocaudal axes, respectively.

RESULTS

sVol-sV linear regressions for ROIs based on the ventrodorsal axis yielded the highest R(2) (range, 0.71-0.92 for mean ROI volumes from 7 to 162 mL), the cephalocaudal axis the next highest (R(2) = 0.77-0.88 for mean ROI volumes from 38 to 162 mL), and the laterolateral axis the lowest (R(2) = 0.65-0.83 for mean ROI volumes from 8 to 162 mL). ROIs based on the ventrodorsal axis yielded lower standard errors of estimates of sVol from sV than those based on the laterolateral axis or the cephalocaudal axis.

CONCLUSION

sVol can be computed with PET using the proposed method and is highly correlated with sV. Errors in sVol are smaller for larger ROIs and for orientations based on the ventrodorsal axis.