Cardiovascular development is a sublimely complicated process that involves precise timing of a multitude of molecular events, and a myriad of subtle three-dimensional conformational changes. Traditional developmental biology techniques have been used extensively to detail cardiovascular development in the mouse, as well as non-mammalian species including zebrafish (Isogai et al., 2001, 2003) and chick (Coffin and Poole, 1988; Poole and Coffin, 1989), providing vast quantities of qualitative knowledge as to how these complex organ systems develop. However, the major drawback of current developmental biological imaging is that it is two-dimensional (2D) in nature. Thus, a full understanding of how blood vessels track in three-dimensional (3D) space during development is not fully appreciated, nor is it clear how tightly controlled vascular patterning events are between different developing individuals. What is crucially needed is a method to both characterize and co-localize, in three-dimensional space, the vasculatures of a population of genetically identical mice. This will enable the generation of average vascular maps of the growing arborized blood vessel trees of populations of mice, but will also allow for the identification of small differences in cardiovascular development at discrete developmental time points of the different individuals within the population. Similar methods have been described for the Drosophila adult brain vasculature (Peng et al., 2011), which has a tightly regulated vascular pattern, but to our knowledge this has not been described for whole mount mammalian specimens.
Optical projection tomography (OPT) is a powerful imaging modality that is ideally suited to visualize the complex three-dimensional structure of the mouse embryo. We have shown that we can image early mouse embryos at high resolution to obtain 3D images of their developing vasculature (Walls et al., 2008), and, furthermore, that we can detect cardiovascular abnormalities in mutant mouse models (Lickert et al., 2004). But OPT has far more potential than simply being a tool to generate elegant 3D pictures; the digital nature of data collected by OPT allows it to be used for quantitative analysis. Since it is well known that using a population of mice rather than one individual allows for the capture of within-strain developmental variance (Zamyadi et al., 2010; Wong et al., 2012b), presented here is a similar method of vascular registration for early mouse embryos. Autofluorescence (for general anatomy) and vascular-specific PECAM-1 OPT scans of several mouse embryos within two different background strains are collected at specific developmental time points. PECAM-1, the major constituent of the endothelial cell intercellular junction, has been extensively used as a molecular marker of mature endothelial cells (Drake and Fleming, 2000; Gerety and Anderson, 2002; Chaturvedi and Sarkar, 2006; Suchting et al., 2007), and thus identifies the growing vascular tree. Using autofluorescence scans as an initial guide, individual mouse vasculature images are then co-registered in three-dimensional space, thus superimposing the datasets to create a population average image of all samples. Once fully registered in 3D, the similarities and differences between each individual vascular map at homologous points in anatomy can be visualized. This allows for the investigation of temporal differences in cardiovascular development of specific embryos at discrete developmental time points, and will allow for the mapping of developmental differences caused by genetic mutations as well.
Here we outline a novel method for registering the vasculatures of several mouse embryos into a common three-dimensional space, which allows for comparisons of high-resolution vascular structures. Whether they come from a C57Bl/6 or CD-1 background, embryos at a specific developmental stage have vascular maps that are both rapidly patterned and tightly controlled; embryos that differ by as little as one to three somites have small developmental differences in their vasculature that can be visualized. The ability to visualize such precise developmental processes of the cardiovascular system allows for more accuracy in identifying and describing the steps that take place during this complex process, and provides more information about the individual genetic and hemodynamic requirements necessary to create a functional mammalian heart and vascular network. This method will be amenable to many developmental biology questions in both other organ systems and other species, and will be of great use in future studies of genetically modified organisms.
Acquisition of OPT Data and Comparison With Histology
OPT was developed to fill a need for 3D cellular resolution imaging of biological specimens that are up to a few cubic centimeters in size (Sharpe et al., 2002). Its ability to support the use of molecular markers such as PECAM-1, and other endothelial cell-specific markers (Drake and Fleming, 2000), makes it an extremely valuable tool for the study of mouse cardiovascular development, since the growing vasculature can be imaged at early developmental stages, anywhere from 5 to 30 somites (Walls et al., 2008). Our in-house built OPT system (Fig. 1A and B) (Wong et al., 2012a) was used to capture projection images of mouse PECAM-stained embryos, which were then reconstructed using a standard parallel ray filtered backprojection reconstruction algorithm, which resulted in 3D datasets that can be manipulated, digitally sliced, and viewed from any angle (Slaney and Kak, 1988). Figure 1 illustrates the high-resolution data that OPT is capable of producing when imaging a 21-somite C57Bl/6 embryo. Autofluorescence data is acquired using GFP excitation and emission filters to capture the intrinsic signal of all tissues, thus supplying us with general anatomical information (Fig. 1C–F). Vascular data are acquired using Cy3 excitation and emission filters to capture the PECAM-1 signal of endothelial cells, thus supplying us with vascular anatomy information (Fig. 1C'–F'). Due to the digital nature of the data and its simultaneous acquisition, dual channel images of both the autofluorescence and vascular data can be obtained (Fig. 1C''–F''), illustrating where the blood vessels reside in relation to the tissue as a whole. Sagittal (Fig. 1C–C'', D–D''), transverse (Fig. 1E–E'', F–F''), coronal/frontal (data not shown), as well as arbitrary oblique sections can be viewed as desired, in order to characterize organs such as the growing heart or the budding intersomitic blood vessels in their most beneficial orientation. Figure 1D'', for instance, illustrates the atrioventricular canal and the endocardial cushions of the developing heart (red arrows). Additionally, the trabeculations of the ventricle are visible due to the PECAM-1 stain (blue asterisks). The PECAM-1 stain is specific to the endocardium of the heart, as illustrated by the fact that the myocardium is not stained (green arrowheads). Similar observations can be made from the paired transverse sections of the same embryo (Fig. 1F''). Histological sections of a littermate control embryo (Fig. 1C'''–F''') are shown for comparison.
Registration of Seven E9.5 Embryos
Registration is a process that attempts to manipulate two or more 3D datasets into a common 3D space, so that similarities and differences between individual samples at equivalent locations can be recognized. This has been used to great success in neuroanatomy studies (Spring et al., 2007; Dorr et al., 2008; Lerch et al., 2008). It has recently been applied to mouse embryo studies of general anatomy, though this work has tended to focus on older, larger embryos that have been imaged using magnetic resonance imaging (MRI) (Zamyadi et al., 2010; Cleary et al., 2011) or micro-computed tomography (micro-CT) (Wong et al., 2012b). The embryo registration method employed here is based upon the current mouse brain registration algorithm employed in our laboratory (Kovacevic et al., 2005), but has been modified with additional steps to make it sensitive to the very fine features of the mouse embryonic vasculature. We first looked at C57Bl/6 embryos. We collected autofluorescence and vascular OPT scans of somite stage–matched embryos (20–23 somites; n=7). During the registration algorithm, the autofluorescence data (Fig. 2A–G, represented as volume textured samples) are first used to register the embryos together, where first a set of 3D affine transformations (translations, rotations, scaling factors, and shearing factors), and secondly non-linear, voxel-by-voxel transformations are used to generate a representative consensus average image of all embryo inputs (Fig. 2H). These transformations are encoded in a deformation field for each 3D input image, which, when applied to the original autofluorescence image of each embryo, result in registered images that look very similar (Fig. 2A'–G', A''–G'', and A'''–G''', shown in three orientations). The corresponding deformation fields are then applied to the vascular scans of the individual embryos (Fig. 3A–G, represented as volume-textured images in their original orientations), which approximately registers the individual vascular trees together. However, since the vascular structures are so fine, there are evident registration mismatches (data not shown). Therefore, the vascular images are then subjected to an additional registration algorithm that consists of another set of non-linear, voxel-by-voxel transformations. Thus, a representative consensus average of all vascular maps is calculated (Fig. 3H), and the same process as previously described results in registered vasculature images that look very similar (Fig. 3A'–G', A''–G'', A'''–G''', shown in three orientations).
We next performed the same procedure on a set of somite stage–matched mouse embryos on the CD-1 background (20–23 somites; n=7). Autofluorescence scans and vascular scans were obtained in the same fashion as for the C57Bl/6 embryos. Registration of the autofluorescence data produced a representative consensus average image of all embryo inputs as before, and we were similarly able to register the vascular scans of the embryos to produce a representative consensus average image, and thus register the vascular maps in 3D to a high degree (data not shown). The genetic background of mice, therefore, does not appear to have any bearing on whether or not registration of their anatomy or vascular trees is possible.
Side-by-Side Embryo Visualization and Registration Evaluation
Registration of seven 20- to 23-somite C57Bl/6 or CD-1 embryos enables us to analyze the functional and relevant 3D data that the registration pipeline provides. To visualize this, two 21-somite embryos from the C57Bl/6 registration pipeline are shown side by side in Figure 4A–E, and two 21-somite embryos from the CD-1 registration pipeline are shown side-by-side in Figure 4F–J. Figure 4A and F are representative autofluorescence coronal/frontal slices through the two embryos (one embryo in red, the other in green, along with their corresponding overlap image) of C57Bl/6 and CD-1 embryo pairs, respectively. Figure 4B and G are the corresponding PECAM-1 images of the same slices through the same embryos. Figure 4C and H are representative autofluorescence sagittal slices through two embryos (again, the first embryo depicted in red, the second in green, along with their corresponding overlap image) of the two C57Bl/6 and CD-1 embryo pairs, respectively, while Figure 4D and I are the corresponding PECAM-1 images of the same slices through the embryos. For presentation here, each image is actually a 5-slice stack in the selected plane (thus corresponding to a 27.5-μm slab), in order to better represent the structures of the anatomical organs and the blood vessels of the embryos. What is apparent from the autofluorescence data is that the head, and structures within the heart such as the atrioventricular canal, register together to a high degree. Looking closely at the registered vascular data, findings are similar in that blood vessels in the head region, intersomitic regions, dorsal aorta, and the endocardium in the atrioventricular canals are registered to a high degree. In order to visualize the data in another fashion, surfaces can be rendered from the 3D dataset using an isosurface algorithm (using the visualization program Amira®, Visualization Sciences Group) that creates a digital surface corresponding to a selected intensity value threshold of the reconstructed data. Figure 4E and J illustrate the anterior-most intersomitic vessels of the left side of the body segmented away from the right side and from the rest of the embryo (for the C57Bl/6 and CD-1 embryo pairs, respectively) with the first embryo depicted in red and the second embryo depicted in green. What is clear is that the highly patterned intersomitic vessels are highly aligned in 3D space following registration, especially in these two cases where the two embryos are somite stage–matched. Although the feeding vessels of the developing somites have no discernible pattern, the intersomitic vessels themselves are precisely patterned and are highly conserved in mice that share identical genetic backgrounds.
Vasculature Registration Evaluation
We next wanted to get a better understanding of how much our registration algorithm was changing the embryos in our pipeline in order to properly align their vascular maps following the initial gross anatomical alignment. The average vascular image, displayed in gray scale, can be overlaid with the average voxel displacement map for the entire collection of embryos in the pipeline, displayed as a heat map (Fig. 5A–D). This allows for the identification of areas of both high similarity and dissimilarity, as it displays those voxels that have undergone large and small movements during the registration process. Figure 5A and C are representative coronal and sagittal sections from the C57Bl/6 registration pipeline, respectively, and Figure 5B and D are similar coronal and sagittal sections from the CD-1 registration pipeline. The voxel displacement maps display the exact regions in the registration volumes that have moved following registration, and to what degree (heat map scale bar has been converted to μm). What is evident is that, overall, the original registration of the vascular trees using the original autofluorescence deformation fields does not change too significantly with our additional non-linear registration steps, though certain areas do undergo fairly large realignment. There is a tightening of the blood vessel alignment throughout the vascular tree, particularly in the intersomitic vessel regions. In order to quantify the degree to which our vasculature registration more properly aligns the blood vessels, we decided to focus on the intersomitic vessel region of the embryos. Figure 5E and F are close-up maximum intensity projections (MIPs), through half of the embryo, of the middle of the intersomitic regions for a representative C57Bl/6 and CD-1 embryo, respectively. Overlaid on top of each MIP is a set of manually drawn centerlines corresponding to five sequential intersomitic blood vessels. This was done for all seven embryos in each pipeline, and the relative positions both before vascular registration (Fig. 5G and H) and after vascular registration (Fig. 5I and J) for both the C57Bl/6 and CD-1 registration pipelines are shown. In order to determine the relative positions of each intersomitic vessel, we next calculated the root mean square (RMS) displacement of the center point of each of the centerlines relative to one another in each of the intersomitic vessels. Prior to vascular registration, the average RMS displacement of the centerlines for the C57Bl/6 embryos was 6.00 voxels (32.73 μm) (Fig. 5G), and 7.28 voxels (56.21 μm) for the CD-1 vascular trees (Fig. 5H). Following vascular registration, the RMS displacement was 3.17 voxels (17.28 μm) for the C57Bl/6 vascular trees (Fig. 5I), and 1.30 voxels (10.06 μm) for the CD-1 vascular trees (Fig. 5J).
Vascular and Cardiovascular Developmental Differences at Sequential Developmental Stages
We next wanted to identify the useful information the registration algorithm was able to provide with respect to early developmental biology. As such, following registration, we looked at embryos side-by-side that differed in developmental stage by one or, at most, two somites for both the C57Bl/6 and the CD-1 embryos. Small developmental differences were detected, even though a difference of a few somites corresponds to only a few hours of gestational time. Figure 6A depicts slices of the registered vascular data of two C57Bl/6 embryos: a 23-somite embryo (left), a 21-somite embryo (middle), and the overlap of the two vascular maps (right), all shown in sagittal views (all images are depicted as an average of five slices through-plane, creating a 27.5-μm slab). When superimposed, the presence of an additional intersomitic vessel can be detected in the vascular tree of the 23-somite embryo (arrow), even though all other intersomitic vessels have been aligned to a high degree. Similarly, when we examine the intersomitic regions of the same 23-somite C57Bl/6 embryo (Fig. 6B, left) and a different 21-somite C57Bl/6 embryo (Fig. 6B, middle), once again the presence of an additional intersomitic vessel can be detected in the vascular tree of the 23-somite embryo that is not accounted for in the 21-somite embryo (arrow). Similarly, small developmental differences can be detected in the CD-1 strain of mice. Figure 6C is a representation of a 23-somite CD-1 embryo (left), a 22-somite embryo (middle), and the overlap of the two vascular maps (right; all images are depicted as an average of five slices through plane, creating a 27.5-μm slab). Once again, there is an obvious mis-match in the intersomitic region, where two intersomitic blood vessels from the 22-somite embryo cannot be aligned with the three intersomitic blood vessels from the 23-somite embryo (bracket in overlap image), even though the rest of the intersomitic vessels have been registered to a high degree.
At this same stage of development, the mouse heart is undergoing rightward looping (Srivastava and Olson, 2000), and trabeculations from the endocardium begin to form in the ventricle. When we examined the endocardial region of the C57Bl/6 embryos, we noted some small developmental differences between the registered datasets. Specifically, when the registered vascular maps of a 23-somite embryo (Fig. 7A and B, left) and a 20-somite embryo (Fig. 7A and B, middle) are examined side-by-side, a higher abundance of ventricular trabeculations in the heart of the 23-somite embryo (arrows in overlap image, right side) are evident, even though much of the rest of the heart (such as the endocardial cushion region) is highly aligned (Fig. 7A and B represent two different sagittal slices through the hearts of the embryos, and all images are an average of five slices through plane, creating a 27.5-μm slab). This suggests that rapid heart development/maturation is occurring in C57Bl/6 embryos at this point, and that such developmental differences can be captured by OPT. No such differences were detectable in the CD-1 population of mice at this stage of development; all embryos imaged had abundant and almost complete ventricular trabeculation patterns.
The presence of additional intersomitic blood vessels and an increase in ventricular trabeculation patterning illustrates the fact that small but detectable developmental maturation steps can be imaged and detected at early developmental time points. These small developmental landmarks would otherwise be difficult to detect without directly comparable images from 3D registration.
Presented here is a novel method in which early developmental processes can be tracked and visualized in their native three-dimensional space so that questions of scale and spatial orientation are no longer a confounding issue. The development of the cardiovascular system is an extremely complex process that has been studied extensively, in many animal model systems. Until now, however, it has not been possible to determine how similar the vascular structures of genetically identical mouse embryos are, and, conversely, how dissimilar these respective vasculatures can be when separated by only a few hours of developmental gestation. By systematically generating vascular maps of early mouse embryos at fixed developmental time points and registering these maps to littermate controls, it is now possible to observe the amount of change and strict patterning that occurs in the cardiovascular system over short time intervals. Here we chose to focus on embryos that were somite-stage matched at 20 to 23 somites, but previous work from our lab has shown that embryos up to 30 somites in development are amenable to PECAM-1 staining and OPT visualization (Walls et al., 2008). Unpublished results from our lab suggest that embryos up to 35 somites of development can be imaged using OPT, after which light scattering becomes a confounding issue as the tissue becomes too dense to be cleared effectively.
In the embryo proper, the formation of the major blood vessels and the developing heart is a tightly controlled and hierarchal process, which is why 3D cardiovascular maps generated by OPT are so amenable to registration. However, the blood vessels that arise in the extraembryonic tissues such as the developing yolk sac, though developing an arborized hierarchy of structures, are not stereotypically patterned. As such, any attempt to register the yolk sac vasculatures of multiple mouse embryos fails. Though mutations that affect the cardiovascular system in the embryo will often also have a negative effect on the yolk sac vasculature (Tanaka et al., 1999; Wakimoto et al., 2000; Huang et al., 2003; May et al., 2004), these developmental processes cannot be investigated using registration-based techniques.
The ability to register comparative multiple mouse embryos that have been imaged by OPT is a major advance in quantitative 3D imaging: to date only brains and embryos imaged by MRI or micro-CT have been amenable to this process (Spring et al., 2007; Lerch et al., 2008; Zamyadi et al., 2010; Cleary et al., 2011; Wong et al., 2012b). Since OPT has a much higher resolution and a superior contrast to both MRI and micro-CT, it allows for the imaging of much younger and smaller embryos, and it allows for the capture of molecular-specific markers. As shown here, it is possible to image minor structural differences in the developing hearts and growing vasculatures of embryos that differ by only a couple of somites.
Genetic mutations that affect the developing cardiovascular system have manifestations that may present earlier than previously appreciated, and may go unnoticed by conventional molecular biology techniques. Thus, there are many additional applications of this method. By using this newly developed registration process and comparing mutant models of vascular disease to wild-type controls, it will be possible in the future to pinpoint the critical time of expression of mutations by visualizing very specific differences in the comparative vascular maps that are generated. This will allow for a more specific, accurate, and quantifiable description of mouse models of cardiovascular disease. As well, this method need not be confined to the study of the cardiovascular system, nor to mouse models themselves. Presented here is a method that is very robust, and could be applied to diverse areas of scientific research, from mammalian limb development to zebrafish vascular development.
Both the C57Bl/6 and CD-1 mice were purchased from the isolator-raised, elite colony at Charles River (Wilmington, MA). Upon arrival at our mouse facility, following an extensive three-day health check in a full barrier, limited access conditioning room, the mice are housed in Tecniplast Greenline sterile ventilated caging with access to automated watering (acidified/RO/UV treated) and are fed irradiated Teklad 2918x extruded diet (Harlan Laboratories, Indianapolis, IN). In order to ensure genetic homogeneity in the mouse populations, the breeding colonies are only bred for three generations and all breeding cages are replenished with new stock from Charles River at the end of the three-generation cycle.
Embryo Collection and Staining
All embryos were collected and stained as described previously (Walls et al., 2008). Briefly, wild-type C57Bl/6 or CD-1 embryos were collected at embryonic day (E) 9.25–9.5 (20–23 somites). Noon of the day of vaginal plugging was considered to be E0.5. Embryos were dissected and fixed in 4% paraformaldehyde for 2 hr. Endogenous peroxidase activity was quenched by immersing the embryos in 3% H2O2, and non-specific antibody staining was blocked by pre-incubating embryos in 1% heat-inactivated fetal calf serum (FCS) and 1% normal goat serum. Embryos were then stained overnight with 5 μg/mL anti-PECAM-1 antibody (Mec13.3) (BD Pharmingen, San Jose, CA). The primary antibody was then detected by incubating the embryos overnight with an anti-rat horseradish peroxidase (HRP) secondary antibody (BioSource, Grand Island, NY), followed by incubation with a tyramide-Cy3 reagent (1:50; Perkin-Elmer, Waltham, MA) for 1 hr. The secondary antibody was then washed away with Tris-NaCl-Triton X-100 (TNT) buffer overnight. All animal experiments were approved by the Animal Care Committee at Mount Sinai Hospital (Toronto, ON), and were conducted in accordance with guidelines developed by the Canadian Council on Animal Care.
Optical Projection Tomography (OPT) of Embryos
Optical projection tomography was performed as described previously (Walls et al., 2007). The OPT microscope was home built (Wong et al., 2012a) to provide a good optical spatial resolution of ∼6 μm, a large solid angle, robust components for stability, and a large sample holder for specimens. Briefly, specimens were embedded in 1% low melting point agarose and subsequently cleared using a 1:2 mixture of benzyl alcohol and benzyl benzoate (BABB). The embryos were then suspended from a stepper motor and immersed in an optically flat cuvette containing BABB. Images of the specimen were formed using a Qioptiq Telecentric Zoom 100 microscope equipped with a 0.5× Optem objective lens. The zoom setting used for image formation resulted in a numerical aperture of 0.05. Images were acquired with a Qimaging Retiga 4000DC CCD camera with pixel size equal to 7.4 μm/pixel. Light from a mercury lamp was directed onto the specimen and filter sets were used to created fluorescent images of the specimen. An autofluorescence view was captured with a GFP excitation filter set in the illumination and detection light path, and a view of the PECAM-1 fluorescence from the specimen was captured using a Cy3 excitation filter set in the illumination and detection light path. The sample was rotated stepwise with a 0.3-degree step size through a complete revolution and views were acquired at each step. The acquisition time for a single filter set is approximately 20 min (Wong et al., 2012a).
Reconstruction of OPT Data
Each OPT image approximates a parallel ray projection through the embryo. As subsequent images are acquired from the CCD, a sinogram is formed, which is then used to reconstruct the corresponding slice through the embryo using the standard convolution filtered back-projection algorithm (Slaney and Kak, 1988). The reconstruction of all slices produces a 3D volumetric representation of the embryo with isotropic pixel size of 4.45 μm. The resulting 3D reconstruction of autofluorescence images and its corresponding 3D reconstruction of Cy3 images are co-registered. The datasets are typically on the order of 2.0 GB in size.
3D Registration of Embryos
Autofluorescence scans of all embryos used for the registration were first subjected to a rigid body alignment [3 rotations, 3 translations (1 in each of the x, y, z directions)] to orient the embryos into a preliminary atlas. This step removes postural and angular discrepancies. Next, all possible pair-wise 12-parameter affine transformations (3 scales, 3 shears, 3 rotations, and 3 translations) were computed, and a transform to an unbiased aggregate was created for each individual embryo. All scans were then averaged to create the first population average, which represents the average anatomy of the embryos after accounting for overall differences in body orientation and size. Next, an iterative 6-generation multi-scale, non-linear alignment procedure was computed, initially registering each embryo towards the 12-parameter registration atlas, and subsequently towards the atlas of the previous non-linear generation. Each step involves detailed matching of anatomical features using a coarse grid that becomes progressively finer with each non-linear step, finally ending at the resolution of the imaging voxels. All registrations were performed using the MNI autoreg tools (Collins et al., 1994). The end result places all autofluorescence scans into precise alignment with each other in an unbiased manner. Evaluation by cross-correlation analysis revealed that for the C57Bl/6 embryos, the cross-correlation coefficient for the autofluorescence data was 0.79 ± 0.01 (indicating a 79% overlap of all voxels, ± the standard error of the mean), while for the CD-1 embryos, the cross-correlation coefficient for the autofluorescence data was 0.86 ± 0.02. The deformation field for each individual embryo, which contains all information pertaining to the functions performed on it, was then applied to each embryo's respective vasculature scan, thus approximately aligning all vasculature scans of the embryos with their autofluorescence scan counterparts. At this point the cross-correlation coefficient for the C57Bl/6 embryos was 0.55 ± 0.01, while for the CD-1 embryos the cross-correlation coefficient was 0.65 ± 0.02. In order to account for any residual inconsistencies in alignment at this stage (since the vasculature has a very fine resolution), and to more precisely align the vascular trees, the vasculature scans were put through another iterative 6-generation multi-scale, non-linear alignment procedure as before (matching of vascular features using a coarse grid that becomes progressively finer with each non-linear step, finally ending at the resolution of the imaging voxels). Thus, all vascular scans are placed into precise alignment with each other in an unbiased fashion. Evaluation by cross-correlation analysis revealed that for the C57Bl/6 embryos the cross-correlation coefficient for the vascular data was 0.69 ± 0.02 following registration, while for the CD-1 embryos, the cross-correlation coefficient following registration was 0.80 ± 0.01.
G.A.A. is a recipient of an Ontario Graduate Scholarship. R.M.H. is a Canada Research Chair in Imaging Technologies in Human Disease and Preclinical Models. The authors gratefully acknowledge funding for the Ontario Preclinical Imaging Consortium from the Ontario Research Fund.