2.1.

Animal Preparation

All animal procedures were performed according to the approved protocols of the Institutional Animal Care and Use Committee at the University of Illinois at Urbana-Champaign. Swiss Webster mice were purchased (Harlan, Indianapolis, Indiana) at $6\text{to}8\text{weeks}$ of age. The paired breeding system was used and the presence of a vaginal plug indicated copulation.¹⁷ Fertilization was assumed and gestational dating of day 0.5 was assigned. Females were subsequently placed in separate cages for the remainder of gestation.

2.2.

Embryo Preparation for Static Imaging

Pregnant females were euthanized by $\mathrm{C}{\mathrm{O}}_2$ inhalation at desired gestational age and the maternal abdomen was incised. The uterus was gently retracted out of the abdominal cavity and embryos were dissected from the uterus and placed in a petri dish. For E14.5 and E17.5 embryos, the chest cavity was opened and the pericardium was removed to expose the heart. During OCT imaging, whole embryos (E10.5) or embryos with exposed hearts (E14.5 and E17.5) were immersed in saline. At each gestational age, three embryo specimens were imaged and multiple stacks, comprising 3-D volumes of OCT image data, were recorded.

2.3.

Embryo Preparation for Dynamic Imaging

Immediately following euthanasia of the pregnant mouse, the entire uterus was gently retracted from the abdominal cavity and one half of the uterus was opened to expose embryos still in intact amnion sacs. Partially exposed embryos were immersed in warmed saline for real-time spectral-domain OCT imaging. Embryo-placental continuity was maintained, and embryos remained in situ, encased in the amnion sac and filled with amniotic fluid. In situ embryos remained warm from the body temperature of the euthanized pregnant mouse. No additional effort was made to maintain normothermia of the embryos. This in situ approach was used to yield more physiologically relevant functional data.

2.4.

Optical Coherence Tomography

3-D OCT imaging was performed using a neodymium:vanadate diode-pumped titanium:sapphire laser as a low-coherence light source. The pulsed laser output had a center wavelength at $800\mathrm{nm}$ with a $100\text{‐}\mathrm{nm}$ spectral bandwidth at a pulse repetition rate of $90\mathrm{MHz}$ and an average output power of $350\mathrm{mW}$ , which enabled an axial imaging resolution of approximately $2\mu \mathrm{m}$ in tissue. This light was coupled into a fiber optic coupler that served as a beamsplitter in this fiber-optic-based OCT system (Fig. 1 ). The sample arm light from the fiber was collimated, reflected off a pair of orthogonal galvanometer-mounted mirrors that directed the $x\text{‐}y$ position of the beam on the tissue, and was focused by a $20\text{‐}\mathrm{mm}$ focal length achromatic lens to a spot size diameter (transverse resolution) of $15\mu \mathrm{m}$ , with a confocal parameter of $400\mu \mathrm{m}$ . Incident optical power on the tissue was measured to be $10\mathrm{mW}$ . The time-domain OCT instrument included a reference arm with a galvanometer-mounted retroreflecting mirror translated over an optical pathlength of $1.5\mathrm{mm}$ at a frequency of $40\mathrm{Hz}$ . Reflections from each interferometer arm were recombined in the fiber optic coupler and the interference signal was sensed by a photodiode. A second reference photodiode was used to monitor laser power fluctuations and was subtracted from the signal photodiode to cancel these fluctuations. The electrical output of the photodiode was bandpass filtered, amplified, and digitized with $12\text{‐}\text{bit}$ accuracy. OCT volume scans were performed by acquiring sequential cross sectional $\left(x\text{‐}z\right)$ scans at $10\text{‐}\mu \mathrm{m}$ intervals along the $y$ axis. Custom software assembled digital data for display as 2-D or 3-D images. 2-D images $(512\times 512\text{pixels})$ were acquired in $17\mathrm{s}$ with the slower time-domain OCT method, and 3-D datasets consisting of 256 2-D images were acquired in about $70\min$ . Faster acquisition rates are possible using advanced time-domain scanning methods ( $64\mathrm{ms}$ per image, $16.4\mathrm{s}$ per 3-D dataset)¹⁸ at the expense of lower signal-to-noise ratios, or by using swept-source¹⁹ or spectral-domain OCT techniques. Custom previously published image processing algorithms were applied to enhance the visualization of morphological features at high resolutions.²⁰

Fig. 1

Schematic of the time-domain optical coherence tomography (TD-OCT) and the spectral-domain optical coherence tomography (SD-OCT) systems. The two systems use the same source and sample arm. TD-OCT uses a photodetector and a moving reference-arm mirror, while SD-OCT uses a spectral detector and a stationary reference-arm mirror.

2.5.

Spectral-Domain Optical Coherence Tomography

Spectral-domain optical coherence tomography (SD-OCT) is an emerging technique that rapidly generates an OCT image from the spectrum of the light backscattered off of the sample.^{21, 22} The advantage of this technique is an increased signal-to-noise ratio (SNR) at a higher acquisition rate.²¹ Figure 1 shows features of our SD-OCT system. The source and sample arm are equivalent to those used in the time-domain (TD) OCT system described before. The main differences found in the SD-OCT system compared to the TD-OCT system are that the reference arm is held stationary rather than scanned for depth imaging, and that a spectrometer (linear detector array) is used to capture the optical spectrum of the returning light, rather than a single element photodiode. In the SD-OCT system, the light returning from the sample and reference arms was directed into the spectrometer. Light was dispersed off of a diffraction grating $(830.3\text{grooves}∕\mathrm{mm})$ and focused using a pair of achromatic lenses which have a combined focal length of $150\mathrm{mm}$ . The focused light was incident on a line-scan camera (L104k-2k, Basler, Incorporated, Ahrensburg, Germany) containing a 2048-element array of detection elements. This camera has a maximum readout rate of $29\mathrm{kHz}$ , thus one axial (depth) scan can be captured in $34\mu \mathrm{s}$ . Digital processing of the detected signal included a spline interpolation to make the signal more uniform, and a discrete Fourier transform on each set of 2048 $10\text{‐}\text{bit}$ values to transform the signal from the frequency (spectral) domain into the spatial (depth) domain. Each scan contained data for one axial (depth) scan. Adjacent axial scans were assembled into 2-D cross sectional images for visualization on a personal computer.

It should be noted that in the present instrument, the axial resolution is superior to the transverse resolution. High axial resolution can be used to ameliorate speckle by averaging the amplitudes of several speckles over a particular feature size. Alternatively, one could use a higher transverse resolution at the expense of a shorter depth of field, and then employ focus tracking or acquire several depth scans to stitch together an image. Our resolution dimensions were chosen on the basis of available instrumentation and murine cardiac feature sizes.

2.6.

Tissue and Image Analysis

During OCT imaging, approximately 256 images per specimen were acquired. Following OCT imaging, specimens were fixed in 10% neutral buffered formalin, followed by standard paraffin preparation. Since histological sectioning can only represent one plane per sample, two additional littermates of the imaged E10.5 embryo were histologically processed and sectioned at planes that corresponded to the computationally extracted 2-D OCT image planes. Sections ( $6\mu \mathrm{m}$ thick) were stained with H and E according to standard protocol for light microscopy observation (BH-2, Olympus America, Incorporated, Melville, NY) and digital image capture (Spot RT Slider, Diagnostic Instruments, Sterling Heights, MI)). Comparisons were made between matched OCT and light microscopy images, correlating microstructural features.

OCT images contained within 3-D datasets were first processed using custom dispersion compensation algorithms,²⁰ then processed using the linear burn algorithm (Photoshop 7.1, Adobe Systems, Incorporated, San Jose, CA) to compensate for depth-dependent signal attenuation, and then viewed and analyzed using commercial software (Analyze^®, Biomedical Imaging Resource, Mayo Clinic, Rochester, MN and Amira^®, ZIB, Mercury Computer Systems, Berlin, Germany) on a personal computer. 3-D projections and volume-rendered datasets were generated and rotated along $x\text{‐}y\text{‐}z$ coordinates for visualization of microstructural features. Using volume data, three different planes of images were formed, consisting of a transverse plane (cross sectional depth plane), a sagittal plane, and an oblique plane, relative to the long axis of the embryo. Acquired data are presented in three formats: 1. extracted 2-D images at selected planes, 2. flipbook movies (QuickTime Version 6.5.2), and 3. volumetric renderings of the whole mouse embryo (E10.5) and embryonic hearts (E14.5 and E17.5). All movie files and captions can be viewed at: http:∕∕biophotonics.uiuc.edu/movies/mouseembryo∕.

3.1.

Optical Coherence Tomography Imaging of Static E10.5 Murine Embryo

A fresh whole embryo (E10.5) was imaged by OCT. Images from three different cross sectional planes from a 3-D OCT dataset are shown in Figs. 2, 3, 4 . OCT is capable of resolving the internal structures of the embryo with high resolution ( $2\text{‐}\mu \mathrm{m}$ axial, $15\text{‐}\mu \mathrm{m}$ transverse) and as deep as $1\mathrm{mm}$ in the highly scattering tissue of the embryo. A 3-D OCT dataset containing 256 2-D OCT images was computationally sectioned along the sagittal plane of the embryo, and shown with corresponding histological sections in Fig. 2. The optical sectioning capability and the high resolution are apparent. Despite the optically dense and highly scattering structures of the murine embryo (estimated at $10\text{to}20\text{‐}\mathrm{dB}∕\mathrm{mm}$ attenuation), features are well preserved at depths up to $1\mathrm{mm}$ in the specimen and strongly correlated with the H and E-stained histological sections at comparable planes. The representative segittal images in Fig. 2 present many detailed internal structures including the branchial arch, hepatic primordial, fourth ventricle, and subcardinal vein. Figures 2e, 2f, 2g, 2h are zoomed-in regions of acquired OCT images and show the anatomic structures of the embryonic heart including the bulbus cordis, the cushion tissue lining the atrio-ventricular canal, and an atrial and ventricular chamber of the heart. The high resolution and structural representation through the whole embryo clearly illustrate the developing cardiovascular and nervous system, among others. In Figs. 2i and 2j, a 3-D volume-rendered image from the 3-D OCT dataset is compared with a digital image of the murine embryo (E10.5) used to generate this dataset. A sequence of 256 computationally sectioned OCT images along the sagittal plane is presented as the flipbook Movie I.

Fig. 2

Sagittal OCT images of the E10.5 mouse embryo. (a) and (c) Computational sections along sagittal planes extracted from the 3-D OCT dataset. (b) and (d) Corresponding H and E-stained histological sections. (e) through (h) Zoomed OCT images of the embryonic heart, showing details of internal structures at varying cross sectional planes. (i) 3-D volume rendering of the OCT dataset. (j) Digital photograph of the E10.5 mouse embryo that was imaged with OCT. The entire set of 256 2-D sagittal OCT images can be viewed in Movie I. Abbreviations: ba, branchial arch; hp, hepatic primordia; fv, fourth ventricle; sv, subcardinal vein; vch, ventricular chamber; cav, cushion tissue lining the atrio-ventricular canal; ach, atrial chamber; bc, bulbus cordis; and uc, umbilical cord. Scale $\text{bars}=200\mu \mathrm{m}$ (a) through (d) and (i); $100\mu \mathrm{m}$ (e) through (h): and $1\mathrm{mm}$ (j).

Fig. 3

Oblique OCT images of the E10.5 mouse embryo. (a), (c), (e), (g), and (i) Computational sections along oblique planes extracted from the 3-D dataset. (b), (d), (f), (h), (i) Corresponding H and E-stained histological sections. Abbreviations: nl, neural lumen; bc, bulbus cordis; vch, ventricular chamber; fv, fourth ventricle; mda, middle line dorsal aorta; uv, umbilical vein; and ua, umbilical artery. Scale $\text{bar}=200\mu \mathrm{m}$ .

Fig. 4

Transverse OCT images of the E10.5 mouse embryo. One section (out of 256 2-D OCT images) of a whole E10.5 murine embryo is shown, and represents the imaging plane at which the original 2-D OCT images were acquired serially to assemble the 3-D OCT dataset. (a) OCT image and (b) corresponding H and E-stained histological section. The entire 256 OCT image set is presented as a flipbook movie (see Movie II). Abbreviations: nl, neural lumen; mdl, midline dorsal aorta; vv, vitelline vein; hp, hepatic primordia; iecc, intra-embryonic coelomic cavity; and vch, ventricular chamber. Scale $\text{bar}=200\mu \mathrm{m}$ .

To illustrate and identify other anatomical structures, the 3-D OCT dataset was also computationally sectioned at an oblique plane and shown in Fig. 3, along with corresponding histological sections. From these oblique images, structures identified include the neural lumen, bulbus cordis, a ventricular chamber of the heart, fourth ventricle, and the mid-line dorsal aorta. Finally, one representative transverse OCT section from the set of 256 is shown in Fig. 4, along with the corresponding H and E-stained histological section. The section was acquired transverse to the long axis of the embryo and represents the original plane at which 2-D OCT images were acquired in series to assemble the 3-D OCT dataset. A sequence of all 256 images can be viewed as a flipbook (Movie II). Depth-dependent attenuation of the OCT signal intensity is evident in these images. Relative to the orientation of these OCT images, the imaging beam was incident left to right. These depth-dependent attenuation effects appear as lighter, faded regions in the right half of this OCT image. This effect is dependent on the direction of the incident OCT imaging beam, relative to the regional optical properties of the specimen, and can be mediated by imaging at varying angles around structures of interest.

3.2.

Optical Coherence Tomography Imaging of Static E14.5 and E17.5 Murine Embryos

Exposed hearts from E14.5 and E17.5 embryos were prepared and imaged. Figure 5 illustrates oblique images extracted from the 3-D OCT volume and the corresponding histology. In these images, cardiac structures including an atrial chamber, a ventricular chamber, a ventricular outflow tract, and the aorta and pulmonary trunk can be clearly visualized. The heart from a relatively late developmental stage (E17.5) was also imaged. Figures 5f and 5h are OCT images that reveal defined cardiovascular structures such as the atria, ventricles, and the aorta, which are confirmed in the strongly correlating histology in Figs. 5g and 5i. Figures 5e and 5j are 3-D volume-rendered images of the embryonic murine hearts at days E14.5 and E17.5, respectively, which clearly show the major vasculature surrounding the heart, including the pulmonary trunk and the aorta.

Fig. 5

OCT of the embryonic mouse heart (E14.5 and E17.5). (a) and (c) OCT images, (b) and (d) corresponding H and E-stained histology, and (e) 3-D OCT volume rendering of the E14.5 murine heart. (f) and (h) OCT images, (g) and (i) corresponding histology, and (j) 3-D OCT volume rendering of the E17.5 murine heart. Abbreviations: ach, atrial chamber; vch, ventricular chamber; otv, outflow tract of ventricle; ao, aorta; and pt, pulmonary trunk. Scale $\text{bar}=200\mu \mathrm{m}$ .

3.3.

Optical Coherence Tomography Imaging of Dynamic E10.5 Murine Embryo

The SD-OCT system enabled continuous data acquisition, processing, and display at five frames per second, which was suitable for in vivo optical imaging. In vivo murine embryo (E10.5) heart rates have been reported to be approximately $120\text{to}210\text{beats}\text{per}\text{minute}$ ,^{23, 24} which corresponds to $4\text{to}7\text{frames}∕\sec$ imaging required to acquire two frames per beat, and $5\text{frames}∕\sec$ is sufficient to image up to $150\text{beats}∕\min$ with two frames per beat. In this study, semi-exposure of the E10.5 embryo maintained embryo-placental continuity, and the embryo remained in situ, encased in its amnion sac and amniotic fluid [Fig. 6h ]. This semi-invasive approach was taken to yield more physiologically relevant data. However, due to maternal-embryo euthanasia, specimen preparation, and the semi-invasiveness of this protocol, the heart rate slowed to $20\text{to}30\text{beats}∕\min$ prior to imaging.

Fig. 6

In vivo OCT imaging of the beating E10.5 murine heart. (a) through (d) and (i) through (l) Sequences of four serial frames extracted from movies captured using SD-OCT. Each sequence was acquired from a different angle (acquired $5\text{frames}∕\sec$ , see Movies III and IV). The dynamic motion of the E10.5 heart is evident: (a) early systole, (b) end-systole, (c) early diastole, and (d) end diastole. (e) and (f) Corresponding H and E-stained histology for (a) through (d) and (i) through (l) sequences, respectively. (g) Digital photograph of the highly scattering embryonic heart. (h) Digital photograph of the murine embryo encased in an intact amnion sac filled with amniotic fluid, the condition under which OCT images were acquired. Abbreviations: ach, atrial chamber; vch, ventricular chamber; ct, cushion tissue; pc, pericardial cavity; and as, amnion sac. Scale $\text{bars}=200\mu \mathrm{m}$ (a) through (g) and (i) through (l) and $1\mathrm{mm}$ (h).

Figures 6a, 6b, 6c, 6d are four serial frames extracted from a SD-OCT movie (Movie III). The dynamic motion of the E10.5 heart is shown at varying stages of the cardiac cycle, including: early systole [Fig. 6a] end systole [Fig. 6b], early diastole [Fig. 6c] and end diastole [Fig. 6d]. Figures 6i, 6j, 6k, 6l are four serial frames extracted from a SD-OCT movie (Movie IV) captured at a different position and angle. The dynamic motion of the cardiac cushion of the E10.5 heart was observed. The corresponding histology for the image sequences is also shown. Figure 6g is a digital photograph of the surface of the highly scattering embryonic heart, shown in the orientation in which the OCT imaging beam was incident. Figure 6h is a digital image of the whole murine embryo encased in its amnion sac and surrounded by amniotic fluid, through which OCT imaging was performed.