1.

Introduction

The development of protective ventilation strategies to improve mechanical ventilation of patients in intensive care is of great importance. In particular, the ventilation of acute respiratory distress syndrome (ARDs) patients has to be improved to decrease the mortality of patients with lung injuries. Furthermore, artificial ventilation has to be optimized to prevent worsening of pre-existing lung injuries.^{1, 2, 3} However, approaches to improve artificial ventilation are hampered by our lack of understanding of their effects on the actual units of gas exchange, i.e., the alveoli. One approach to circumvent this limitation is the development of numerical models of the lung on an alveolar scale.⁴ With the help of such models, alveolar dynamics can be simulated and different protective ventilation strategies can be characterized concerning the mechanical stress exerted on the alveolar structures. The basis for such models is detailed knowledge of the alveolar behavior in an in vivo situation, especially information about alveolar volume change during the breathing cycle. However, adequate in vivo 3-D information about the parenchymal dynamics at the level of individual alveoli is not available in the current literature. The most established modality to image in vivo lung tissue at the alveolar level is intravital microcopy (IVM), which allows 2-D, high resolution, and real-time imaging of the subpleural lung surface. Yet IVM does not provide 3-D information on alveolar dynamics during ventilation.⁵ Additionally, there are focusing problems during the ventilation cycle for this technique, requiring the positioning of cover glasses on the lung surface for stabilization, which in turn may influence the dynamic behavior of the imaged alveoli beneath. 3-D imaging techniques like computed tomography or magnetic resonance tomography, on the other hand, do not provide a sufficient spatial resolution to image alveoli in vivo.^{6, 7} To investigate the alveolar behavior during ventilation in vivo, an imaging technique is hence required that provides high resolution, allows for 3-D imaging without influencing the alveolar mechanics, yields high acquisition rates to perform real-time imaging, and can be applied in vivo. Optical coherence tomography (OCT)⁸ is an optical imaging technique allowing contactless 3-D imaging of strongly scattering samples like biological tissue. With a spatial resolution of less than $10\mu \mathrm{m}$ and measurement ranges of up to $2\mathrm{mm}$ , OCT is suitable to image subpleural alveoli that have diameters of approximately $100\mu \mathrm{m}$ .⁹

2.

Methods and Materials

3.

Results

Evaluation of the acquired image stack demonstrates that virtual 4-D, airway pressure-gated imaging of subpleural lung tissue in vivo is in principle feasible by OCT. Figure 1 presents an exemplarily volume rendering taken from an injured lung at an airway pressure of $8\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ . The lung tissue is bordered by the visceral pleura (part b), which can be seen on the upper side of the tissue block. Furthermore, single alveoli (part a) are recognizable beneath the pleura. The dashed line represents the plane that was used to extract the OCT en-face images for analysis at a depth of approximately $45\mu \mathrm{m}$ beneath the pleura. The picture series shown in Fig. 2 displays the OCT en-face images taken from the 3-D datasets. In the upper two rows are images of healthy lung tissue, and in the lower rows are images of injured lung tissue from the same animal. The image sequences show alveolar changes with increasing airway pressures during the inspiratory phase. Black structures in the en-face images represent the air-filled alveoli, and white structures are the alveolar walls. Identical structures can be observed in all images, demonstrated by the landmark in the OCT en-face images. However, in the image taken at $8\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ airway pressure in the healthy lung, it was impossible to identify this structure. Furthermore, the margin between two lung lobes is recognizable and denoted by the black arrow. The main difference between the image sequences obtained in healthy and injured lungs is the reduced movement of the injured lung during the inspiratory phase, as demonstrated by the lack of dislocation of the lobe margin. This notion is supported by the quantitative analysis of alveolar distension in healthy and injured lungs. Figure 3 shows the alveolar area change during the increase of airway pressure from $2\text{to}11\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ in the inspiratory phase. Alveolar area was normalized to the area measured at $2\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ . The black line presents the area change in an exemplarily healthy lung, the dashed line in an injured lung. Both graphs show an increase in alveolar area with increasing airway pressure during the inspiratory phase. In healthy lungs, the major part of expansion occurs between 2 and $4\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ airway pressure, whereas the larger part of expansion in injured lungs happens at airway pressures higher than $8\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ . Compared to alveolar size at $2\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ , the increase of airway pressure to $11\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ increased alveolar area in healthy lungs by 12%, while alveoli only expanded by 8% in injured lungs. Both the right shift of the area-pressure curve and the overall reduced alveolar expansion reflect the reduced alveolar compliance and the increased stiffness of the injured lung.

Fig. 1

A volume rendering of subpleural alveolar tissue. The tissue block has dimensions of $1500\times 1500\times 1500\mu {\mathrm{m}}^3$ . The visceral pleura (part b) and single alveoli (part a) can be detected. The dashed line shows the cutting plane, which was used to extract the en-face images for alveolar area quantification. The shift between single B-scans is attributable to motion artifacts caused by the heartbeat.

Fig. 2

Image sequences showing OCT en-face images of healthy (upper rows) and injured (lower rows) lung of the same animal. The images were extracted from the 3-D OCT image stacks acquired during the inspiratory phase from $2\text{to}11\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ . An alveolar landmark is highlighted in all images by the black circle. The black arrow denotes the margin between two lung lobes. A large dislocation of this margin can be recognized with increasing airway pressures in the healthy lung, in contrast to the injured lung, where the location of the interlobar margin remains unchanged, suggesting that alveolar expansion is decreased. Scale bar is $200\mu \mathrm{m}$ .

Fig. 3

Line graph showing exemplarily pressure-dependent changes in alveolar area during the inspiratory phase normalized to the area at $2\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ for a healthy (solid) and after induced lung injury (dashed).

4.

Discussion

In this study, we demonstrate the feasibility of in vivo virtual 4-D imaging of subpleural alveolar structures by optical coherence tomography in healthy and injured lungs of ventilated mice. This new method allows for imaging of alveoli during the dynamic respiratory cycle without applying constant positive airway pressure (CPAP)¹² or gating to the inspiratory or expiratory plateau phases.¹¹ During the entire data acquisition process, the mouse is regularly ventilated, warranting that the recorded data realistically reflect the physiological situation in the dynamically ventilated lung. Off-line assembly of the single B-scans acquired in sequent respiratory cycles at different pressure levels allows for virtual 4-D imaging of the lung parenchyma. We termed the acquired image stacks virtual 4-D, since the reconstructed 3-D image stacks at different airway pressures in fact represent the tissue at the identical phase of the ventilation cycle, because the acquisition of each stack was carried out over a series of successive breathing cycles. The observed expansion of the alveolar area in healthy lungs and the reduced alveolar compliance in lung injury corresponds to our previously reported data from OCT and IVM imaging in vivo ¹¹ and in isolated lungs.⁹ Our results are thus in accordance with the notion of decreased alveolar expansion in lung injury due to stiffening of the lung parenchyma.¹² However, the alveolar expansion measured in this study is smaller than comparable data on alveolar expansion published previously.^{11, 12} This quantitative discrepancy is likely attributable in part to different experimental conditions. In our study, fast data acquisition during the regular inspiratory phase allowed us to refrain from the application of CPAPs over a longer period of time, like in studies with isolated lungs.

Taken together, the presented imaging technique allows to our knowledge the first virtual 4-D imaging of subpleural alveoli in vivo. The method can be further improved by using faster and higher spatially resolving OCT setups to obtain more detailed knowledge about the alveolar dynamics during the ventilation cycle.^{14, 15} A faster OCT system will also allow minimizing motion artifacts, while high resolution setups can reduce the apparent underestimation of the alveolar area by the OCT technique.¹²