1.1.

Clinical Background: Role of Alveolar Mechanics in Ventilator Induced Injury

The acute respiratory distress syndrome (ARDS) is a severe form of acute lung injury resulting from sepsis, trauma, or severe pulmonary infections. Intensive care management of ARDS and acute lung injury (ALI) frequently requires endotracheal intubation and mechanical ventilation of the patients to ensure sufficient blood oxygenation.² The mechanical ventilation itself, although life saving, eventually has negative side-effects on the lung. As soon as there was a consensus that the ventilation contributes to or exacerbates ARDS by causing ventilator induced/associated lung injury (VILI/VALI),³ an ongoing search for more protective forms of ventilation started. As demonstrated in clinical studies, ventilator settings with lower tidal volume (VT) and higher positive end-expiratory pressure (PEEP) than traditionally used lead to a better outcome and lower mortality of ARDS patients.⁴

To understand the positive effects of low volume and high PEEP ventilation and to develop new protective forms of ventilation, the mechanisms responsible for VILI/VALI have to be clarified in detail. The amount of stress (tensile strain and shear forces) put on the alveolar walls during ventilation is thought to be a key factor for the development of VILI.⁵ Two main initial stress mechanisms have been suggested: 1. volutrauma caused by overdistension of the alveolar walls, and 2. atelecttrauma associated with alveolar recruitment and derecruitment (R/D). The R/D mechanisms are controversially discussed. There is no agreement on the exact deformation of acini and alveoli during volume uptake and the corresponding stress put on the alveolar walls, even when looking at healthy lungs. A variety of mechanisms is discussed that are partly contradicting or hypothetic. Some researchers favor a balloon-like expansion of the alveoli, some suggest an expansion of alveolar ducts without changes of alveolar form and size in the normal lung, some describe R/D as repetitive alveolar collapse and opening or uncrumbling of folded alveoli and acini, others state that air bubbles being pressed through edemic (water filled) airways might induce the atelecttrauma.^{6, 7}

1.2.

Challenge: Imaging Alveolar Mechanics

The reason for the controversy about alveolar mechanics is mainly caused by the difficulties that arise when trying to image the alveoli in situ during ventilation. The interpretation of pressure-volume curves,⁸ arterial fixation and histological examination of lungs at different tidal volumes,⁹ inductance plethysmography, electrical impedance tomography,¹⁰ and computed tomography scans¹¹ have all contributed to a better understanding of whole lung mechanics. However, they all do not provide the real-time information and spatial resolution that are necessary to analyze mechanics of aerated lungs on an alveolar scale during ventilation. Since the mid-1970s, videomicroscopy of subpleural alveoli has provided high-resolution information about lung mechanics on an alveolar scale.¹² More recently, the group of Nieman captured real-time in-vivo microscopic images of subpleural alveoli and presented a series of quantitative studies on alveolar mechanics in healthy and pathological lungs modeling ARDS.^{6, 13} But there is a principle drawback in using microscopy: it provides top views of alveoli through a glass plate sucked to the pleura. Thus, the information is only 2-D. This restricted view leaves questions open about the interconnectivity of the alveoli, about a possible anisotropy of volume expansion, and about such complex processes like the unfolding of acini. Because this depth information is missing in microscopy, there is still a search for “the ideal investigative tool,” which was defined as one “that can measure 3-D changes that occur in the alveolus and alveolar duct continuously throughout tidal ventilation.”⁶

1.3.

Feasibility of Fourier Domain Optical Coherence Tomography for Imaging Alveoli

Our study aims at proving that optical coherence tomography (OCT), and especially its spectral domain variant Fourier domain OCT (FD-OCT),^{14, 15} is a potent tool for studying the dynamic mechanics of subpleural alveoli and acini during ventilation. In contrast to microscopy, OCT has the advantage of providing depth information from the first subsurface $500\mu \mathrm{m}$ of highly scattering tissue with a resolution better than $10\mu \mathrm{m}$ . This information is gained by analyzing the reflection of infrared probe light scattered by structures within the tissue. To extract the depth information of the scatterers, the echoed probe light is brought to interference with a reference light beam originating from the same low coherence light source. The principle of operation and the different variants of OCT, such as time domain OCT (TD-OCT), spectral/Fourier domain OCT (FD-OCT), linear OCT, and swept source OCT, are all described in review articles.¹⁶ For high-speed applications, Fourier domain OCT seems most suitable because the analysis of the interference by a spectrometer requires neither mechanically moving parts in the reference arm nor frequency tuning of the light source. In fact, high-speed FD-OCT systems have recently achieved frame rates of $29.3∕\mathrm{s}$ for 2-D cross sectional imaging.¹⁷ More and more, 3-D imaging OCT systems are reported to produce volume scans of complex 3-D structures like the spiral sweat glands in human skin.¹⁸ The optical setup of the compact prototype FD-OCT system used in this feasibility study is described in Sec. 2.3.

1.4.

Optical Coherence Tomography Imaging of the Lung

Its ability to provide depth information makes OCT a promising tool for tissue classification and tumor investigation. Thus, OCT has already been used in other feasibility studies for imaging excised lung tissue,^{19, 20} introducing OCT as a potential real-time alternative to histological investigation during thorax surgery. In another study, the dynamics of laser ablation of ex-vivo lung tissue, followed by the deflation of alveoli, was captured by OCT imaging with a $0.1\text{-}\mathrm{s}$ time resolution.²¹ Very recently, endoscopic OCT scanner heads of $1\mathrm{mm}$ diam were used to detect abnormalities in the diseased airway epithelium during in-vivo fiber optic bronchoscopy.²² In the same study, ex-vivo cross sectional OCT imaging of the subpleural alveoli was performed on fixated lung tissue excised from pigs. To our knowledge, there has been no report so far on OCT imaging of alveolar mechanics during ventilation.

2.1.

Isolated Rabbit Lung

The model of the ventilated and blood-free perfused isolated rabbit lung has been already used by our group for studies on ventilator-induced injury, and is therefore an appropriate model for the testing of OCT imaging.²³ The protocol of this study is in accordance with the National Institutes of Health guidelines for animal use. The preparation of the isolated, perfused rabbit lung model has been described in detail by our group elsewhere.²⁴ Briefly, female rabbits (Orticolagus caniculus) weighing $2.2\text{to}3.2\mathrm{kg}$ were anesthetized with ketamine $50\mathrm{mg}∕\mathrm{kg}$ (CuraMED, Karlsruhe, Germany) and xylazine hydrochloride $4\mathrm{mg}∕\mathrm{kg}$ (Bayer, Leverkusen, Germany) after cannulation of the auricular vein. Anticoagulation was achieved by means of IV administration of heparin $1000\mathrm{U}∕\mathrm{kg}$ (Liquemin, Hoffman–La Roche, Grenzach-Wyhlen, Germany). After pretracheal infiltration with $8\mathrm{mL}$ of lidocaine hydrochloride 1% (Jenapharm, Jena, Germany), a tracheotomy was performed, and the trachea was cannulated by using a $10\text{-}\mathrm{cm}$ -long catheter with a diameter of $0.4\mathrm{cm}$ (endotracheal tube) (B. Braun, Melsungen, Germany). Animals had the lungs ventilated with room air by use of a mechanical ventilator equipped with a serial interface (Cato, Dräger, Lübeck, Germany). The real-time pressure data were transmitted to the OCT controlling personal computer via RS232 serial bus using a real-time protocol (MEDIBUS, Dräger, Lübeck, Germany). The initial ventilator settings were $\mathrm{Vt}=8\mathrm{mL}∕\mathrm{kg}$ , respiratory $\text{frequency}=30\text{breaths}∕\min$ , $\mathrm{PEEP}=5\mathrm{cm}$ ${\mathrm{H}}_2\mathrm{O}$ , and inspiratory-expiratory ratio 1:1. After a median sternotomy, the pulmonary artery was cannulated, and perfusion with Krebs-Henseleit hydroxyethyl starch buffer solution (buffered with $20\text{-}\mathrm{mM}$ HEPES at $\mathrm{pH}=7.35$ ) started with a roller pump at $50\mathrm{mL}∕\min$ (Masterflex L/S, Cole-Parmer, Manufacturing, Barnant, Barrington, Illinois). Simultaneously, the heart was opened to permit the exsanguination of the lungs. The lungs and trachea were carefully dissected, removed en bloc, and suspended from a weight transducer (Hottinger Baldwin Messtechnik, Darmstadt, Germany) in a temperature-controlled $(37°\mathrm{C})$ double-walled chamber, which could be moved down to free the lung surface for OCT measurement. The lung perfusate dropping from the lungs into the chamber was collected into a reservoir and redirected to the roller pump as a closed recirculation system. The total volume of perfusate in the system was $200\mathrm{mL}$ . The temperature of the perfusate was maintained at $37°\mathrm{C}$ with a water bath, and weight gain was continuously recorded.

The respiratory frequency was lowered to $12∕\min$ and the inspiratory-expiratory ratio was changed to 1:1.5 before a series of six different ventilator settings, each lasting a time period of $10\min$ , was applied to the isolated lung (Table 1 ). OCT measurements were performed at the upper and lower lobe of the left lung at the second half of each time period.

2.2.

Formalin-Fixated Rabbit Lung and Scanning Electron Microscopy

The lung was prepared as described in the previous section, disconnected from ventilation, and connected to a formalin reservoir. The lung was filled endobronchially with 4% formalin buffered solution with a pressure of $20\text{-}\mathrm{cm}$ ${\mathrm{H}}_2\mathrm{O}$ , leading to an extension that is equivalent to the end-inspiratory size during ventilation. 20 minutes after starting the fixation and keeping the filling pressure constant, a series of 3-D OCT scans was taken.

Approximately $1\mathrm{h}$ after starting the fixation, the lungs were ligated at the trachea, disconnected from the reservoir, and placed in 4% formalin solution for several days. Small probes of the lungs at the sites corresponding to OCT measurements were excised and washed three times with $0.1\text{-}\mathrm{M}$ phosphate buffer solution (PBS). Dehydration of the lung was performed by a series of ethanol baths with increasing concentration (70% ethanol overnight, 80% for $1\mathrm{h}$ , 90% for $1\mathrm{h}$ , 96% for $30\min$ , and $3\times 20\min$ 100%). After critical point drying with liquid $\mathrm{C}{\mathrm{O}}_2$ (E3000, Gala Instruments, Schwalbach, Germany) and sputtering with gold in a Balzers MED010 sputtercoater $(90\mathrm{s},50\mathrm{mA})$ , specimens were studied with a LEO 5430 scanning electron microscope.

2.3.

Setup of the Common Path Fourier Domain Optical Coherence Tomography System

The optical setup [Fig. 1a ] of the FD-OCT system used for alveolar imaging is based on a low coherent light source, which is a broadband near-infrared superluminescent diode (SLD) with a $50\text{-}\mathrm{nm}$ -wide spectrum centered at $830\mathrm{nm}$ . One milliwatt of the SLD output power is coupled into a single-mode optical fiber. After passing a fiber coupler joining the pathways to the SLD and to the spectrometer, the light is guided to a handheld scanner head. In the scanner head, the light is collimated to a free space beam and deflected by a dichroitic mirror, reflective for infrared and transparent for visible light. This mirror is mounted on a galvanometer scanner providing fast beam deflection. The deflected beam is focused on the probe by telecentric optics. Close to the focal plane, the beam passes a partially reflective glass window. The reflective side of this window serves as the reference plane for the OCT interferometer. Between the reference window and the probe, an index matching gel is used to minimize first surface reflections by the probe. The light reflected by scattering structures within the probe, together with the reference light, is focused back into one common fiber and guided via the fiber coupler into the grating spectrometer, where the dispersed light is focused on a silicon line detector with a width of $1024\text{pixels}$ . The resulting interference spectrum is read out to a personal computer at a maximum rate of $5\mathrm{MHz}$ by a $12\text{-}\text{bit}$ analog-digital converter.

Fig. 1

Setup of the FD-OCT system used for (a) (top) 3-D imaging with an OCT scanner head mounted on a $y$ -translation stage. The scanner head contains a beamsplitter cube and a reference arm. The translation stage is controlled synchronously to image acquisition (b) (bottom) real-time 2-D imaging with a scanner head containing a partially reflective reference outlet window. Index matching gel is put between pleura and outlet window. The pressure of the mechanical ventilation is synchronously monitored.

The depth information of the scattering structures in the probe along the axial $\left(z\right)$ direction (A scan) is derived from the fast Fourier transformation of the spectrum. The axial measuring depth into the lung is limited by scattering to approximately $500\mu \mathrm{m}$ , and the axial resolution in air is $8\mu \mathrm{m}$ . One B-mode OCT image of $1200\times 512\text{pixels}$ is acquired by collecting successive A scans and translating the logarithmized intensity values into grayscale levels while scanning a lateral extension of $6\mathrm{mm}$ .

A small complementary metal-oxide semiconductor (CMOS) camera behind the dichroitic scanner mirror allows monitoring of the surface of the probe and the moving beam spot, so disturbing air bubbles in the index matching gel can be detected and removed. Scanner control, data acquisition of OCT spectra, real-time acquisition of the ventilator pressure, and image processing are all implemented in a LabView® (National Instruments) program.

2.4.

Setup of the Contact-Free 3-D Optical Coherence Tomography

For the acquisition of 3-D OCT scans of the fixated lung, a modified OCT scanner has been used [Fig. 1b]. Instead of the partially reflective window at the outlet of the scanner head, a beamsplitter cube is introduced into the free space beam between the collimator and scanner mirror. The cube deflects 50% of the beam intensity into a reference arm with a focusing lens and a reference mirror. The reference mirror is adjusted at a fixed length in such a way that the optical pathway in the reference arm is approximately $1\mathrm{mm}$ longer than the optical pathway to the first surface in the probe arm. The depth of scattering structures is measured relative to the virtual reference plane within the tissue. In this geometry, first surface reflections are the most distant scattering signals producing the highest modulation frequencies in the spectrum. Since the modulation transfer function of the line detector drops toward higher frequencies, the first surface reflections are damped compared to the scattering signals from deeper within the tissue. The focal plane can be adjusted independently, preferably in-between the surface and virtual reference plane. Because of the intrinsic reduction of first surface reflections, the use of index matching gel is not necessary in this setup.

To add the 3-D $\left(y\right)$ perpendicular to the $\left(x\right)$ deflection of the scanner mirror, the scanner head is mounted on a motorized translation stage with a maximum of $25\text{-}\mathrm{mm}$ travel. The translation stage steps between successive B-mode scans, allowing the capture of a full 3-D volume scan of the tissue.

Since fixated lungs were used for 3-D imaging, measuring speed was not as crucial as in the real-time application. Consequently, a higher lateral resolution $\left(16\mu \mathrm{m}\right)$ than in the 2-D measurements could be chosen and a $1250\text{-}\mathrm{kHz}$ $16\text{-}\text{bit}$ analog/digital converter was used to collect the data. The B-mode image stacks were saved to the harddrive and processed off-line with a 3-D volume rendering program.

3.1.

Imaging of 3-D Alveolar Structure by Fourier Domain Optical Coherence Tomography

3-D OCT scans were taken from different spots at all lobes of a fixated rabbit lung immediately after fixation. In Figs. 2a, 2b, 2c different views of a 3-D OCT scan of the upper right lobe are shown. The dimensions of the volume shown are $640\times 640\times 1024\mu \mathrm{m}$ divided into $256\times 256\times 256$ voxels with a size of $2.5\times 2.5\times 4\mu \mathrm{m}$ (in air). The aspect ratio in the figures has been corrected for a refractive index of $n=1.3$ to produce a realistic view. The pleural surface is tilted with respect to the $x\text{-}y$ plane of the scan. Figures 2a and 2b show volume-rendered projections of the OCT scan. In Fig. 2b, a subpleural cut through the image data allows a view into the 3-D alveolar structure. The alveoli are filled with formalin, which produces much less scattering (black) than the lung parenchyma (light gray). In Fig. 2c, three perpendicular cross sections are shown. The contour of individual subpleural alveoli can be clearly seen. The cross sections in the $x\text{-}y$ direction provide a homogeneous gray-level distribution, which is useful for further image processing (threshold conversion to black and white, tissue-surface to volume ratios, and 3-D segmentation).

Fig. 2

3-D images of the formalin fixated rabbit lung. Volume rendered $640\times 640\times 1024\mu \mathrm{m}$ (corrected with refractive index 1.3) scan in (a) (top left) full perspective view and (b) (top right) cut open parallel to an $x\text{-}y$ plane. (c) (bottom left) Three perpendicular cut planes in perspective view; lung parenchyma light, formalin filling dark. (d) (bottom right) SEM image for comparison.

3.2.

Correspondence of Fourier Domain Optical Coherence Tomography Images to Scanning Electron Microscope Images

To validate the quality of the 3-D OCT images, scanning electron microscope (SEM) images are compared to the OCT scans of the same lobe of the fixated lung. In Fig. 2d, a SEM image shows a cut through the lung parenchyma. The OCT images of the fixated lung in Figs. 2a, 2b, 2c qualitatively reproduce the shape and distribution of alveoli observed in the SEM image.

Because a different OCT scanner head was used for real-time imaging, a cross sectional OCT image (Fig. 3 ) taken of the fixated lung by this scanner is also compared to a SEM image (in the inset of Fig. 3) of a probe that was cut close to the pleura. The OCT image shows the pleura and cross sectional cuts through acini with individual alveoli. Again, the measured OCT depth data have been corrected for a refractive index of 1.3 to obtain a realistic aspect ratio in the image. The smaller average diameter of alveoli and alveolar ducts in the SEM image compared to the OCT image is due to a shrinking process taking place during the preparation procedure for the SEM imaging.

Fig. 3

OCT cross sectional image ( $6\times 2.1\mathrm{mm}$ in air) of (a) (top) formalin fixated rabbit lung and in the inset a SEM image of the same scale. The height of the OCT image has been corrected for a refractive index of 1.3 to provide a realistic aspect ratio. The pleura, some individual alveoli, and an acinus (alveolar sack) are labeled. This image was acquired with the apparatus described in Fig. 1b.

3.3.

Real-Time Fourier Domain Optical Coherence Tomography Imaging of Alveoli throughout Tidal Ventilation

In Figs. 4a, 4b, 4c, 4d two pairs of cross sectional B-mode OCT images of the upper left lobe of an isolated and blood-free perfused rabbit lung are shown. The acquisition time interval of one B scan $\left(123\mathrm{ms}\right)$ is short compared to the inspiratory and expiratory phases of ventilation, allowing us to assign the images to either of the ventilator phases and to omit images during transitional phases. In each pair there is one image captured at the end-inspiratory phase of ventilation and the other at the end-expiratory phase of mechanical ventilation. From pair to pair, different ventilator settings are applied. In Figs. 4a and 4b, ventilator settings were $\mathrm{VT}=15\mathrm{mL}∕\mathrm{kg}$ and $\mathrm{PEEP}=0\mathrm{cm}$ ${\mathrm{H}}_2\mathrm{O}$ . Figure 4a shows the end-inspiratory phase and Fig. 4b the end-expiratory phase of one ventilation cycle. In Figs. 4c and 4d, ventilator settings were $\mathrm{VT}=10\mathrm{mL}∕\mathrm{kg}$ and $\mathrm{PEEP}=5\mathrm{cm}$ ${\mathrm{H}}_2\mathrm{O}$ . Figure 4c shows the end-inspiratory phase and Fig. 4d the end-expiratory phase. For both ventilator settings, peak inspiratory airway pressure was $16\text{-}\mathrm{cm}$ ${\mathrm{H}}_2\mathrm{O}$ . The images cover a cross sectional area of $6\times 2.1\mathrm{mm}$ with a resolution of $20\times 8\mu \mathrm{m}$ (in air). The simultaneously recorded pressure curves are plotted in the inserted bottom-left boxes with small circles indicating the point in time of the image acquisition.

Fig. 4

Cross sectional image series (top to bottom) of the ventilated isolated rabbit lung during (a) end-inspiratory phase of ventilation with $\mathrm{VT}=15\mathrm{mL}∕\mathrm{kg}$ and zero PEEP; (b) corresponding end-expiratory image; (c) during end-inspiratory phase with $\mathrm{VT}=10\mathrm{mL}∕\mathrm{kg}$ and $\mathrm{PEEP}=5\mathrm{hPa}$ ; and (d) corresponding end-expiratory phase. The airway pressures in (a) and (c) are both $\mathrm{Paw}=16\mathrm{hPa}$ . In the insets, airway pressure versus time curves are shown; the circles indicate the time points of image acquisition. In the non-PEEP case, an acinus undergoing recruitment/derecruitment is labeled.

Table 1

Ventilator settings for the isolated rabbit lung.

In the images, light structures are assigned to lung parenchyma and dark areas are interpreted as cross sections of air filled volumes, i.e., alveoli and alveolar ducts.

In Fig. 4a, air-filled (recruited) subpleural alveoli are present at end-inspiration and they are interconnected in the $z$ direction to alveolar ducts. In Fig. 4a, a pulmonary acinus with recruited alveoli is labeled. During the expiratory phase [Fig. 4b], the nonair-filled volume is increased, indicating that alveoli are derecruited and alveolar walls are folded [labeled in Fig. 4b] and there is low interconnectivity to alveolar ducts. These observations suggest that during inspiration, air flows from the distal airways into the alveoli, increasing the inner volume and expanding the walls of the acini. Looking at the change in shape and curvature of the walls surrounding the acini, one portion of the expansion might be due to expansion of individual alveolar walls, another part might be due to opening of additional alveoli (recruitment/derecuitment on an alveolar scale), and another portion might be due to an increase of the interconnecting volume (alveolar ducts). In a cross sectional view, however, the 3-D interconnectivity of the sliced volume is unknown, so the question remains if entire acini or only single alveoli are folded or crumbled during the non-PEEP expiratory phase and reinflated during inspiration.

When PEEP is applied, the FD-OCT images of alveolar mechanics dramatically change. The corresponding images are shown in Figs. 4c and 4d. In the inspiratory phase, more acini are aerated than in the non-PEEP case. In the expiratory phase, alveoli and acini stay open, indicating that there is only little or no alveolar recruitment/derecruitment throughout the ventilatory cycle. In addition, there is more 2-D interconnectivity to alveolar ducts in the $z$ direction compared to ventilation without PEEP.

In conclusion, the images in Figs. 4a, 4b, 4c, 4d show that with the help of FD-OCT imaging, the influence of variations in VT and the effect of PEEP on the mechanics of the isolated lung can be visualized on an alveolar scale.