2.1.

Optical Setup

Three-dimensional simultaneous optical coherence tomography and 2-D confocal fluorescence microscopy is provided in one setup. The design was chosen to be modular for flexible interchangeability and adjustment to suit different experimental demands. Three main parts can be specified (Fig. 1): first, the OCT light source and detection system; second, the scanner head; and third, the fluorescence excitation and detection setup. The expansion to a 3-D fluorescence image acquisition was done by introducing a piezo-driven objective stage to the scanner head. The system uses two distinct numerical apertures for the different imaging modalities to assure high depth of field in the case of OCT and sufficiently high resolution for fluorescence microscopy to resolve micrometer sized elastic fiber bundles in the lung tissue.

Fig. 1

Combined OCT and confocal fluorescence microscopy setup for simultaneous 3-D data acquisition. The system contains three major parts: OCT light source and spectrometer, the laser coupling and fluorescence detection system, and the scanner head. The OCT uses a light source centered at 843 nm and provides a 11.88 kHz A-scan rate due to the line scan rate of the CCD line detector. Fluorescence excitation is provided for dyes excitable at 488 and 561 nm utilizing diode-pumped solid-state (DPSS) lasers, where the detection channels are centered at 525 nm (BP1) and 641 nm (BP2) with a spectral bandwidth of 50 and 75 nm, respectively. The fluorescence signal is detected using avalanche photodiodes (APD). Near-infrared light from the super luminescent diode (SLD) as well as visible light from the DPSS lasers is combined within the scanner head at a dichroic mirror (DC2). For OCT, a Michelson interferometer, consisting of a beam splitter (BS) and a reference as well as sample arm, is inserted into the beam path. Data acquisition is performed simultaneously by using the same scanners (SC) and the sample beam path of the OCT modality. Within the zoomed area the utilization of two different numerical apertures (N.A.) is shown as a prerequisite for the combination of both imaging methods. Via underfilling of the objective’s back aperture, the OCT modality uses an N.A. of 0.08, resulting in an extended depth of field and therefore higher imaging depth than the CFM modality utilizing an N.A. of 0.45. The higher N.A. results in an increased resolution for the fluorescence system. Fluorescence light is recollected behind the dichroic mirror at the glass fiber, where the aperture of the fiber serves as a pinhole to establish the confocality of the CFM method. Abbreviations: APD (alanche photodiode), BP1 (band-pass emission filter $525/50$), BP2 (band-pass emission filter $641/75$), BS (beam splitter), C (circulator), CCD (line sensor), DC1 (dichroic mirror [long-pass 491 nm]), DC2 (dichroic mirror [short-pass 790 nm]), G (diffraction grating), GP (glass plates for dispersion adjustment), LP (long-pass mirror 585 nm), M1 (silver mirror), M2 (reference mirror), O (microscope objective), P (glass prisms for dispersion adjustment), PC (polychroic mirror with 488 and 561 nm reflection), PE (piezo element), S (sample), SC ($x\text{-}y$-scanner), SLD (super luminescent diode at 840 nm and spectral width of 50 nm).

The OCT system is based on a spectral domain setup. Broadband near-infrared light at [center wavelength 843 nm, full-width-half-maximum (FWHM) bandwidth 50 nm] with a power of 1 mW is emitted by a superluminescent diode (Superlum Moscow, Russia) and coupled into a circulator. Behind the circulator, the light is guided via a connected single mode fiber to the scanner head. Passing a beam splitter, a separation occurs into a reference beam and a sample beam. The reference arm carries additional glass plates (N-BK7 of 5 mm thickness, SF6 of 6.35 mm thickness) and a prism box (N-BK7 of 13.95 mm thickness), allowing dispersion compensation for the optical components located in the sample arm. Since second-order dispersion decreases the axial resolution in an OCT system, the dispersion between the reference arm and the sample arm was balanced according to the minimization of the second-order derivative of the optical path difference between the two traveling paths.¹⁹ As mentioned before, this was carried out by inserting additional glass into the reference arm with suitable dispersion and thickness. Upon splitting, the sample beam gets reflected by a dichroic NIR/VIS mirror passes an $x\text{-}y$-scanning unit and, after beam expansion through a telescope configuration of lenses, is imaged onto the sample using a $20\times 0.45$ N. A. objective (Olympus, Japan). Since the OCT beam radius amounts to 1.5 mm at the rear aperture of the objective, the N.A. used for this imaging modality is reduced to a value of 0.08, providing a slim focus and a high imaging depth. After reflection at the reference mirror and the sample, respectively, reference and sample light are recombined at the beam splitter where an interference of the two light beams takes place. Recollected at the single mode fiber, the light passes the circulator and is imaged onto a diffraction grating, which splits the light into its wavelength constituents. The interference spectrum is recorded using a CCD line scan detector (Teledyne Dalsa, Canada) with an A-scan rate of 11.88 kHz.

For the confocal fluorescence imaging the light of two excitation lasers (Coherent, USA), 488 nm and 561 nm with each having a maximum power of 20 mW, is combined and coupled into a single mode fiber. At the scanner head it passes the NIR/VIS dichroic mirror and uses the same beam path as the OCT sample beam. The beam diameter for the fluorescence excitation amounts to 9 mm, thereby overfilling the rear aperture of the microscope objective and causing a full N.A. of 0.45. Fluorescence light is recollected, traverses the VIS/NIR mirror and is imaged onto the aperture of the single mode fiber, which is serving as a pinhole for establishing the confocal functionality. Behind the fiber the fluorescence signal is split into a red and green detection channel and is recorded via two avalanche photo diodes (Laser Components, Germany) where additional emission filters, with transmission from 500 to 550 nm and 603 to 679 nm, respectively, block additional stray light from the laser excitation.

2.2.

Mouse Model

A recently developed mouse model introduced by Tabuchi et al.²⁰ was used post mortem according to the guidelines of the local government’s animal care authorities. Briefly, mice were anesthetized and intravenously stained with 50 μl of a $10\mathrm{mg}/\mathrm{ml}$ solution of sulforhodamine B (SRB) in saline water. The fluorescent dye SRB was previously shown to stain elastin fibers and is suited for future in vivo applications.²¹ After sacrificing the animal, a tracheal tube was introduced to the trachea for pressure-controlled ventilation. The chest was opened to gain access to the upper surface of the lung, thereby causing a collapse of the organ. To re-establish the physiological environment within the chest, an intrathoracic catheter was pushed from the inside to the outside of the thorax, and the previously opened chest was closed using a transparent wrapping film glued to the ribcage. Afterward, the negative pressure in the thorax was reconstituted by removal of excess air.

For applying different pressure values to the mouse model, a 1 ml syringe was connected to one arm of a $y$-piece tubing. The second arm carried a differential pressure sensor, and the third arm was connected to the trachea tube of the mouse.

2.3.

Image Acquisition and Processing

Image acquisition was performed using analogue to digital converters (DAQmx, National Instruments, USA) and custom-built software developed with LabVIEW 2009 (National Instruments, USA). The acquired spectrum was transformed to be linear in wavenumber (k) space since the light distribution on the CCD is approximately linear in wavelength space due to the diffraction grating and further influenced by the chromatic aberrations of the applied spectrometer optics. With an evenly $k$-spaced spectrum applying fast Fourier transformation (FFT) is now possible.²² To reduce the side lobes of the Fourier transformed signal, the detected spectrum was shaped to a Hanning distribution by dividing the detected spectrum by the spectrum of the light source and multiplication with a Hanning window prior to the transformation. Subsequently the modified signal underwent the FFT revealing the intensity over depth profile for each scan. Fluorescence intensity images of the focal plane were recorded simultaneously. By driving the piezo objective stage in steps of 3 μm, subsequent planes were obtained to form a 3-D datacube. For both modalities the lateral step size is mainly determined by the lateral resolution of the fluorescence modality, the selected field of view and the desired scanning duration. The value was set to 1.2 μm of lateral displacement to investigate a $(384\times 384){\mu \text{m}}^2$ tissue area within 9.5 s.

SRB imaging was performed using the 561 nm laser at $0.3\mathrm{kW}/{\mathrm{cm}}^2$ and a single detection channel in the red spectral range.

To allow registration between OCT and fluorescence images, a correlation-based algorithm was applied sorting each single fluorescence plane into the 3-D OCT stack to the z position of maximum similarity between OCT en face and CFM slices. The algorithm was previously described for the mapping of OCT slices into the fluorescence stack of a static sample,²³ but can be performed also vice versa. For this reason, the planes of fluorescence, which had a distance of 4.2 μm in tissue, which is approximately the OCT pixel size in depth, were sorted into the 3-D OCT stack, and an interpolation was done to allow pixel-by-pixel mapping between both modalities along the axial direction. After mapping, the CFM and OCT stacks were rescaled providing voxel sizes of $(1.2\times 1.2\times 1.2){\mu \text{m}}^3$. Further image processing and 3-D visualization was performed using ImageJ (National Institute of Health, USA). Alveolar volume segmentation was done via convolution of the data with a $3\times 3$ Kernel, setting of a defined threshold, which allows a delineation of the alveolar lumen [see results section], and utilizing the ImageJ plugin particle analysis calculating the air-filled space within the alveoli for every $z$-position. The 3-D segmentation of fibers was performed using the ImageJ plugin simple neurite tracer.²⁴ The images of the alveoli as well as the fibers, used for quantification, are displayed and discussed in the results section.

3.1.

Setup Evaluation

The sensitivity of the OCT modality was measured using a silver mirror as a sample and a -14 dB attenuator in the sample arm. With a power of 260 μW at the focal position of the microscope objective, a signal-to-noise ratio of 62 dB was determined for a single A-Scan. Adding 28 dB of attenuation in the sample arm the sensitivity resulted in 90 dB.

The imaging performance of the combined OCT and CFM setup was tested using an USAF-1951-target. Since this target has a micrometer-sized metalized pattern, the lateral OCT resolution can be determined according to the specified distance of the line pairs, which appear bright in the OCT image [Fig. 2(a)]. An intensity profile plot along the group 7 elements 4, 5, and 6 of the USAF-1951-target [Fig. 2(d), data of the left intensity scale] visualizes the resolvability of the line pairs. The three line pairs are distinguishable in elements 4 and 5 but not in element 6 of group 7. The line pairs of group 7 element 5, corresponding with a distance of 4.9 μm, therefore provide the limit for the lateral resolution of the OCT System. For determining the lateral resolution of the CFM modality, the USAF-target was laid upside-down onto a red fluorescence slide (Chroma, USA). Scanning with an excitation wavelength of 561 nm yields an absorption image where the pattern appears dark [Fig. 2(b)]. Merging both signals into an RGB image, OCT in the red channel and the inverted CFM image in the green channel, the overlay of both modalities is revealed by a yellow color [Fig. 2(c)]. Lateral mismatches between OCT and CFM were determined to be in the range of 3.6 μm [Fig. 2(e)]. Since this mismatch is smaller than the resolution of the OCT setup, the correlation algorithm, which is mapping the fluorescence slices into the 3-D OCT stack, can be applied without great disturbance.

Fig. 2

Lateral resolution determination using an USAF-1951-target. The OCT reflection image (a) as well as the fluorescence image (b) were obtained simultaneously by laying the target upside down onto a fluorescence slide. An RGB overlay with OCT in the red channel and the inverted CFM image in the green channel (c) showed good lateral alignment of both imaging modalities. For the determination of the lateral resolution the vertical lines of the group 7 elements 4, 5, and 6 were used to generate an intensity profile plot along these structures (d) for both OCT (left intensity scale) and CFM (right intensity scale). The resolution limit for OCT was determined by element 5, resulting in a resolution of 4.9 μm. For CFM a lateral resolution of better than 4.4 μm could be confirmed, since this is the smallest distance within the target. For a better evaluation of the lateral CFM resolution, the diffraction limited dimensions of 100 nm-sized fluorescent nanoparticles (Fig. 3) were additionally determined. A maximum lateral mismatch between both imaging modalities can be quantified within the group 6 element 1 structure shown in (c). An intensity profile plot perpendicular to the horizontal lines (e) reveals a line distance between OCT and CFM of 3.6 μm. Lines between the data points of (d) and (e) are guides to the eye.

From the intensity profile plots of the fluorescence image [Fig. 2(d), data of the right intensity scale] the three smallest elements (group 7, elements 4, 5, and 6) could easily be resolved. The smallest structure corresponds with a line pair distance of 4.4 μm, the resolution capability of the target was therefore not sufficient for an exact estimate. Hence the point spread function of the CFM modality was determined using fluorescent nanoparticles (nano-screenMAG/R-ARA, Chemicell, Germany) with a diameter of 100 nm dried on a glass substrate [Fig. 3(a)]. The lateral and axial intensity profiles of 10 distinct nanoparticles were approximated by a Gaussian function revealing in average a lateral FWHM of $(0.75\pm 0.02)\mu \text{m}$ and an axial FWHM of $(5.9\pm 0.2)\mu \text{m}$ (errors expressed as SEM). Data are provided for one exemplary nanoparticle [Fig. 3(b) and 3(c)]. With this, the dual lateral resolution modality was verified, providing a large depth of field for OCT yielding higher imaging depth and a small depth of field but strong focusing for increased CFM resolution.

Fig. 3

Lateral and axial resolution determination of the CFM modality using fluorescent nanoparticles with a diameter of 100 nm. A monolayer of particles (a) was imaged three-dimensionally revealing a Gaussian intensity distribution into the lateral (b) and axial (c) direction, respectively. The lateral FWHM was determined to be 0.75 μm, the axial FWHM amounts to 5.9 μm. Data are shown for the nanoparticle surrounded by the white box in (a). $\text{Scale}\text{bar}=5\mu \text{m}$ for the large field of view of (a) and 1 μm for the enlarged region, respectively.

To determine the axial resolution, the OCT setup has been calibrated using a target with a single back reflection plane (glass plate with 5 mm thickness). By changing the focus and measuring the traveling distance, a scaling of 5.4 μm per depth pixel could be obtained. The full width at half maximum of the back reflection signal of the glass plate showed a value of approximately 2 pixels. The axial free space resolution therefore amounts to 10.8 μm. Taken into account the refractive index of lung tissue being close to water (1.333), the axial resolution in the tissue amounts to 7.5 μm for the OCT modality.

3.2.

In Situ Imaging of Mouse Lung

To monitor the dynamical behavior of the lung, we used the combined setup for simultaneous imaging of alveolar geometries and elastin fibers with SRB as a fluorescent marker. Figure 4 shows the 3-D data of the structure at an airway pressure of 6 mbar. OCT data is displayed as gray values whereas CFM data is false colored in red. Three-dimensional registration between OCT and CFM was performed using the above-mentioned mapping algorithm. In Fig. 4 a cross-sectional image of the alveolar tissue is demonstrated with the OCT modality whereas fiber distributions are given by the CFM. The upper serous membrane of the lung, the visceral pleura, is visible as a bright reflecting line. Subpleural alveoli are following underneath. For three distinct depth planes–0, 12, and 24 μm under the middle of the visceral pleura–the corresponding CFM, OCT, and overlay images were extracted, Fig. 4 right-hand side. The first depth (Fig. 4 first row) shows details of the visceral pleura. OCT data shows the bright reflex coming from the dense serous membrane whereas the confocal fluorescence microscopy reveals an intertwined meshwork of elastin fibers, which is covering the alveolar tissue. Approximately 12 μm under the pleura (Fig. 4 middle row) the alveolar structures become apparent in the optical coherence tomography image. The walls of the alveoli are intersected by elastin fibers, as it is visible in the fluorescence data. Furthermore, dense elastin concentrations appear at the edges of neighboring alveoli together with interconnecting fibers suggesting coiled elastin fibers, which could expand during the recruitment. In a depth of 24 μm the OCT still shows alveolar structures, where the CFM merely reveals blurred outlines of elastin fibers. Following analysis of fiber length, changes are therefore restricted to the pleural region and the upper half of subpleural alveoli. For analysis, 12 alveoli were chosen from every pressure step as well as 14 fibers surrounding those alveoli. The selection is displayed in Fig. 5(a) for the alveoli and 5(b) for the fibers, respectively. The quantification of the alveolar volume change and the fiber stretching during increase of ventilation pressure revealed a similar behavior for both parameters [Fig. 5(c)]. Data points of the volume change can be fitted with a sigmoid curve, which is found as a good approximation in studies of macroscopic lung function²⁵ and intravital microscopy studies.⁵ Assuming similar behavior between the volume change and the fiber change, a sigmoid function can also be fitted to the fiber length distension data. The average alveolar volume at 0 mbar was determined as $(19\pm 14)\mathrm{pl}(\text{mean}\pm \mathrm{SEM})$, the average fiber length as $(65\pm 6)\mu \text{m}$. By extracting the slope at the inflection point of the volume’s sigmoid function, a normalized alveolar compliance of 0.19 $1\mathrm{mbar}$ can be obtained.

Fig. 4

Visualization of the combined structural and functional data revealed through the new setup. Fluorescence data was acquired using sulforhodamine B as an elastin fiber staining dye. Left-hand side: A cross-sectional impression of a mouse lung tissue. OCT data is displayed in gray scale, CFM data in red. The blue lines indicate the positions of the extracted en face images shown on the right-hand side. Right-hand side: Extracted CFM, OCT, and overlay images (columns) for different horizontal planes of the lung tissue (rows) at 0, 12, and 24 μm under the middle of the visceral pleura. The first row is attributed to the pleura itself where OCT shows a bright reflection, whereas the fluorescence microscopy reveals a dense elastin meshwork. The second and third row indicates the geometrical features of the tissue, which was obtained using the OCT modality. Complementary to this, confocal fluorescence imaging reveals the distribution of elastin fibers lining the alveolar walls. $\text{Scale}\text{bar}=50\mu \text{m}$.

Fig. 5

Delineation for the analyzed alveoli (a) and starting as well as ending point for the fiber length segmentation of the surrounding fibers (b) at the empirically found threshold level shown exemplarily for one plane of the OCT stack (18 μm below the pleura) and one plane of the CFM stack (3.6 μm below the pleura) at 0 mbar pressure. The stacks were rotated to allow optical sectioning parallel to the pleural surface. After rotation, the view was resized to fit the whole volume where empty areas were filled black. Numbers indicate the alveoli and fibers taken for analysis. More than one fiber per alveolus is indicated as *.1 and *.2 etc. The alveoli and fibers were chosen according to their presence in every pressure step, since some were traveling out of the field of view due to the movement of the lung at higher-pressure values. (c) Alveolar volume change (gray dots with a scale on the right) and fiber stretching (black squares with a scale on the left) normalized to the value given at 0 mbar. During expansion of the tissue both parameters show the similar behavior of a sigmoid increase. Error bars are displayed as standard error of the mean for 10 alveoli and 14 fibers within the tissue of one mouse lung. (d) Double logarithmic representation of the relative volume change over the relative fiber-length change. Experimental data is shown as black squares; data from the fitted sigmoid functions is shown as a gray line. Isotropic expansion would lead to a linear function with a slope value of 3 (dashed line). Since neither the data nor the fit based interpolation correspond with the isotropic function, a homogeneous distension of the alveolar tissue is unlikely. $\text{Scale}\text{bar}=50\mu \text{m}$.

For an isotropic expansion one would suggest a cubic relationship between the fiber-length changes and the volume expansion, which would result in a linear function with a slope value of 3 in a double logarithmic plot [Fig. 5(d)]. Utilizing this illustration, the data as well as the interpolated data from the sigmoid fitting curves was compared with the theoretical linear function. Since both the data and the interpolated curve do not exhibit the theoretical linear behavior, the result indicates a heterogeneous expansion of the tissue, which is underlining the need for a three-dimensional investigation of the alveolar volume and fiber dynamics.