Viscosity is one of the most important properties of the vehicle, because it influences contrast agent behavior and its applications for micro-CT (see Additional file 1 and table S1 for more details). The contrast agent must completely fill the vascular network to allow a precise study of the microcirculation: the actual resolution of vessels imaged with micro-CT is in the range of 5 to 20 μm (i.e., the effective voxel size). Depending on the contrast properties (especially viscosity), penetration of the vehicle through the vessel tree could be limited. For example, in an arterial injection, increased viscosity limits vehicle penetration at the arteriolar level only (capillaries are not filled). This aspect could be of great interest in arterial/venous delineation in micro-CT analysis, as the vehicle does not flow freely through the capillary network and filled the vein system. Other critical technical specifications for an appropriate contrast agent include:

• Absence of diffusion out of blood vessel after animal death.

• Limited medium shrinkage that could lead to vessel deformation or impaired quantification.

• Addition of inert dye, to provide delineation within the circulatory tree for microscopic vessel examination and analysis.

• Ease of manipulation and injection in small animals.

• Homogeneity of contrast agent solution

Silicone rubber has been extensively used as a filling agent to study micro-vasculature, because of its inert properties. A lead-containing radiopaque silicone rubber called Microfil (Microfil, Flow Tech, Carver, MA) has been widely reported in micro-CT studies3. It has a low viscosity that allows it to fill completely the vascular compartments with little resistance45. Coming from different reports, this compound completely fills the arterial vasculature when perfused at physiological pressure and flows freely from the veins67. The hydrophobic properties of silicone rubber keep it contained within the vascular compartment, and extravasation has only been reported in situations such as inflammation where physical leaks are present6.

Neoprene latex 671 (Dupont de Nemours) is a liquid synthetic rubber composed of polychloroprene homopolymer with high tensile strength, high elongation, excellent film formation without curing, and considerable resistance to degradation from chemical or environmental exposure. Because of its low viscosity, this compound fills the arterial vasculature down to 20 μm size vessels when it is perfused at physiological pressure. In our experiments, we never observed any extravasation of latex in tissue or any vessel damage due to neoprene polymerization8.

Gelatin at 5 to 10% concentration has also been used as a vehicle compound with bismuth or barium sulfate910. This enabled complete filling of entire microvasculature under physiologically relevant perfusion pressure11. Other groups have used Batson's No. 17 polymethylmethacrylate with an added lead pigment9.

It is not uncommon for in vivo studies to take 30-120 minutes. Such long scan times necessitate either a long-circulating contrast agent or continuous infusion of contrast material. Contrast agents used for in vivo imaging can be administered repeatedly to the animal, eliminating the need to sacrifice and assay multiple animals at various time points in longitudinal studies.

Different groups report the use of conventional iodinated vascular contrast agent encapsulated within polyethylene glycol-stabilized liposomes, showing stable enhancement of the intravascular space for more than 3 hr12. The Fenestra product line is based on iodinated triglycerides contained in the lipophilic cores of oil-in-water lipid emulsions similar to the naturally-occurring chylomicron remnants. It provides long-lasting visualization of vascular systems with micro-CT. Others have used pluronic F127 and a naturally iodinated compound, Lipiodol to form radiopaque nanoreservoir structures with good success13.

Collectively, the vasculature can be efficiently explored with a combination of contrast agent and vehicle. Notably, it should be pointed out that the ease of handling is another critical parameter. Investigators could modulate the viscosity of the injected compound depending on the target organ and the vascular bed that require filling.

Image registration computes a transformation or a deformation field that can be applied to one volume to align with another 1415161718. Specifically, functional imaging modalities such as microPET or SPECT reveal valuable insights into biochemical, physiological, and pharmacological processes in vivo. Their major limitation, however, is the lack of high spatial resolution and detailed anatomical or vascular information. This problem that can be easily alleviated by using complementary information provided by micro-CT and image registration. As illustrated in Figure 3, functional activity can be overlaid with structural information from micro-CT for more detailed analysis and validation of results.

Segmentation is an image processing technique that partitions an image into meaningful regions or structures of interest. The goal of segmentation is to separate the image points representing a particular object from the rest of the image. Similar to registration, it is another fundamental machine vision task that prepares an image for subsequent quantification and analysis18. Among the most common applications of segmentation for micro-CT are studies aimed at anatomical structures, vasculature, and pathologic diseases.

The result of image segmentation is a set of image points representing one or multiple objects. The image points are clustered together based on some common property such as intensity or variation of its rate of change over an image region. Depending on the approach, segmentation could be performed directly on the raw image data or during its visualization and volume rendering. Depending on the user interaction and implementation, it could be either manual or automatic. In manual segmentation, the user interactively traces contours in 2D micro-CT slices to delineate a region of interest and stack the 2D slices in 3D to form the segmented volume. Although this is a very robust approach that works well even with low image quality and little contrast between image structures, it is a tedious and time consuming task better suited for convex or round shaped objects. An automated and robust approach to segmentation is required for more complex objects, such as blood vessels or vascular vessel trees. Thresholding is an image processing technique that tries to separate an object based on its intensity values by setting all image points below and/or above a threshold of certain intensity18. It is simple and works well if there is a sharp change of intensity on the boundary of the object and the surrounding image points. It is not very reliable in cases where the object's intensity values merge with the background. One way to increase the contrast of blood vessels in micro-CT is to adjust the concentration of contrast agent. Denser contrast agents result in higher intensity, which differentiates vessels from background, and vice versa. More complex techniques based on partial differential equations modeling physical processes and deformable shapes have also been developed18. The best approach depends mainly on the quality of the micro-CT volume. Factors that should be considered are: how well the object of interest stands out from the background, complexity of the structure of interest, reliability and repeatability of the approach.

Since its introduction, software post processing for preclinical micro-CT has posed the widespread acceptance of the need for objective, accurate, and reproducible quantification of results rather than manual measurement.

General techniques for quantification of micro-CT volumes include combinations of morphometric and fractal analysis. Morphometry is the quantitative measurement of structures. Morphometric parameters such as volume, area, connectivity, thickness, and degree of anisotropy are standard for analysis of trabecular bone microstructure1920, and were recently applied for quantification of blood vessels10212223. The morphometric parameters are primarily computed after dimensional reduction of the 3D micro-CT volumes to 2D sections102123. Basic morphometric parameters, such as vessel volume, have also been computed directly from the 3D micro-CT volumes22. Depending on how the 2D sections are obtained, the resulting morphometric parameters can be used to reconstruct the morphometric parameters in 3D, which provides a 3D interpretation of the planar sections 23 or direct quantitative measures from 2D 1021. The measurement of morphometric parameters is relatively simple and is generally carried out as an arithmetical operation over a thresholded binary image that yields information on the properties of the structure of interest.

Fractal analysis has also been used to quantify complex geometric structures in micro-CT volumes2425. A fractal is a quantity that provides a measure of the "self-similarity" of an object at different scales. The more complex the object, the more space it fills. An excellent review of the technique is provided by Baish et al 26. The calculation of the fractal dimension in micro-CT is straightforward and is performed by thresholding the micro-CT volume to segment out the object of interest and counting the number of voxels representing the object at different scales. A multi-scale fractal analysis of microvascular networks in the brain showed that normal cortical vascular networks have scale-invariant fractal properties on a small local scale as described by Risser et al27.

Due to large volume size (10⁸-10¹⁰ voxels), processing, transmitting, and archiving of micro-CT data as well as its intelligent analysis within a reasonable time frame is a real challenge. An efficient 2D serial section analysis method was initially proposed by Li W, et al21. A global threshold to visualize 2D vasculature and eliminate bone using a modified Image Pro-Plus 5.0 algorithm (Media Cybernetics, Silver Spring, MD) was applied to 500 serial slices from the upper limb and 250 serial slices from the lower limb to segment and expressed the data as vascular segment number and volume. The algorithm was built within the AutoPro programming language (Visual Basic SAX engine) inside Image-Pro Plus. They summed up serial 2D imaging information to the 3D volume.

As illustrated in Figure 5, Zhuang and colleagues have successfully established the dynamic 3D geometry of the entire peripheral arterial tree in a mouse hindlimb28. The major focus now is to develop an automated algorithm to extract detailed morphometric data such as the diameters, area, number of vessels, and distributions of different size of vessels. Ideally, such methods for vascular analysis of micro-CT would involve five principle steps: 1) segmentation of arteries from bone and contaminated venous system; 2) re-creation of the 3D vascular tree according to a re-orientated central line along the long axis of the major vessels (femoral or anterior tibial artery); 3) evaluation of the mean diameter and cross-sectional area of the segmented branches in each slice; 4) estimation of the distribution of the diameter of blood vessels into different groups; and 5) computation of the total area and volume in the serial slices of both upper and lower limbs.

Quantitative methods for cancellous bone architecture have been used to evaluate the histomorphometric values of vascular networks22. These parameters include vessel volume, connectivity, number, thickness, thickness distribution, separation, and degree of antisotropy. Using this method, vessel volume can be computed based on the voxel size and the number of segmented voxels in the 3D image after application of the binarization threshold. It is really challenging to define the universal binarization threshold for all evaluations within a study, which is highly dependent on the scanning resolution, imaging quality, and decalcification effect. If a value of the binarization threshold is set too high, smaller vessels will be erased from the image. In that case, the volume will be down-estimated. If a threshold is too low, vessels will appear artifactually large and the volume will be over-estimated.

Connectivity is defined as the maximal number of branches that may be cut without separating the structure. Connectivity can easily be calculated from a 3D dataset by using the Euler number30. As pointed out, the edge problem needs special attention in all quantifications of connection and adjustment of the Euler number might be necessary30.

Vessel number, thickness (lumen diameter), thickness distribution, and separation can be calculated using a model-independent method for assessing thickness in 3D volumes82231. This technique defines a local thickness at every voxel in the volume of interest (VOI) as the diameter of the largest sphere that both contains the voxel and is completely within the structure of interest. The average of the local voxel thicknesses yields the vessel thickness parameter and a similar calculation on the background voxels determines the vessel separation. To calculate vessel number, the segmented volume is skeletonized, leaving just the voxels at the mid-axes of vessels in the structure. Vessel number is defined as the inverse of the mean spacing between the mid-axes of the structures in the segmented volume.

Anisotropy describes the degree to which the direction of the segmented 3D vascular bed is oriented. It can be determined by the mean intercept length, the volume orientation, star volume and length distribution30. Duvall CL, et al 22 used eigen analysis to find the principal material directions from a local image neighborhood and determined the eigenvalues (ratio of the maximum and minimum radii of the ellipsoid) as described by Gundersen et al. 32. It can be done at the scale of a group of vessels. A degree of anisotropy of 1.0 denotes that the vascular network is perfectly isotropic, and higher values of degree of anisotropy indicate that a structure contains a preferential material direction. In a recent contribution by Risser et al33, anisotropy has been used to merge discontinuities in 3D images of vascular structures representing undesirable gaps in vascular networks.

Although 3D data has a dimensional advantage over its 2D counterpart, there is very little, if any, commercially available software for real 3D quantification of blood vessels from micro-CT. Novel quantitative methods and clear definitions of what needs to be quantified are yet to be developed. What we do know is that most of the currently available software tools are for quantification of bone, and before we use them for vascular exploration, we have to determine the extent to which the neovasculature is similar to the trabecular network of rod-like and plate-like structures.

Micro-CT scanners that provide a resolution between 1 and 100 μm can be used for vascular research. If the imaging targets are small capillaries, scanners close to the 1 μm limit provide a significant advantage. Regardless of the resolution, blood vessels in Micro-CT volumes cannot be differentiated from soft tissue without a contrast agent. Consequently, the use of the imaging modality is limited by the properties of the contrast agent (e.g. concentration, viscosity, opacity, etc.). A poorly chosen contrast agent will result in a low image quality that can be very difficult to analyze or quantify.

The main weakness of micro-CT for vascular research, however, is the lack of appropriate software tools for image analysis and quantification. This is partially due to the tortuous nature of blood vessels. For example, peripheral vasculature consists of a combination of tree and arcade patterns. After ischemia, angiogenesis and arteriogenesis form other patterns such as anastomoses and collaterals, which make the vascular tree even more complicated. In order to quantify collateral growth, scientists must be able to differentiate existing from newly grown vessels in 3D. This is not a trivial image processing task and new algorithms must be developed. Commercially available morphometric analysis tools for bone applications have found wide acceptance in the vascular community due to the lack of better alternatives. Those tools perform 2-D serial slice analysis of micro-CT along a given axis1019202123. When applied to vascular micro-CT volumes, such tools usually output the number of blood vessels with a given diameter and volume. These results are conceptually inaccurate, because they depend upon the orientation of the sample and carry an unknown amount of error caused by the dimensional reduction. Depending on the orientation of the sample, estimation of vessel diameter from 2-D may not be possible, because vessels do not always appear as circles on planar sections. Such limitations raise serious questions about the accuracy of serial 2-D slice analysis when applied to vascular micro-CT volumes.

Despite the recent development of morphometric data in the peripheral vasculature and hence the sophistication of the resulting hindlimb models, a rigorous hemodynamic analysis of spatial perfusion of the skeleton muscle is still unattainable. The reason for the inaccessibility of such a model is the lack of quantitative data on the spatial relation between the skeleton muscle and the supplying vessels. Therefore, methods for reconstruction of the 3D branching pattern of the peripheral vessels for a spatial analysis of hindlimb blood flow are also very important.

Correlation of micro-CT data with optical microscopy may help in interpretation of biological results by using complementary information from the two imaging modalities. However, it is not always possible to perform microscopy on sections obtained from ex vivo samples previously imaged with micro-CT and correlate results. In addition, more careful inspection of micro-CT sections imaged with a confocal microscope also indicates that not all blood vessels get filled with contrast (Figure 6). Depending on the type of contrast, the size of its particles, viscosity, and properties, different level of detail in micro-CT are obtained. A quantitative measure that would indicate the amount of difference in blood vessel volume from micro CT and microscopy is needed.