Micro-CT Imaging of a Mouse Spinal Cord

X-rays can be viewed as photons with an energy range of 0.1 - 100 keV or electromagnetic waves with wavelengths ranging from 0.01 - 10 nm. X-rays can be created in several different ways, but here the discussion is limited to continuous spectra cone-beam X-ray imaging. These X-rays are created by a phenomenon known as "Bremsstrahlung," which means "braking radiation" in German. This occurs when a charged particle undergoes an acceleration [1]. In an X-ray source, a negatively charged electron is shot in a vacuum tube and impacts a target material (usually tungsten, molybdenum, copper, or another metal) and through its deceleration, emits photons on the scale of an X-ray. A continuous spectrum of X-rays is generated because the deceleration is not uniform nor instantaneous, though there are spikes in the distribution centered around the characteristic energies of the target material as seen in [2]. There are different curves for different energies and target materials. This is a very important thing to consider when performing a micro-CT scan and will be discussed in a later section.

A typical micro-CT system features three primary components, as shown in [3]. The basic components of this system include: a) X-ray source, b) rotational stage for sample mounting, and c) flat panel or optical objective with CCD detectors. X-rays leave the source and are either absorbed, transmitted, or scattered by the sample before arriving at the detector. Absorption is the predominant interaction measured in microCT, as different materials in the body absorb X-rays differently. For example, bones contain a lot of atomic calcium, which absorbs X-rays well. Thus, bones block X-rays from reaching the detector, and end up showing up in the image as a shadow. The sample is then rotated incrementally and the process is repeated until the sample has been imaged for 360° or in some cases, 180°. The output of the tomography is a series of 2D projections at different orientations that can be reconstructed into a 3D volume.

Micro-CT is a form of microscopy, earning its name from its ability to resolve micro-scale features. The limitation of resolution for this specific category of micro-CT is governed by the source spot size and energy spread, and the detector's type and efficacy; not by the X-ray's wavelength. The best resolution possible for this category of micro-CT is around 500-700 nm in three dimensions. Though, it is possible to detect features a tenth or hundredth of that size.

Magnification is performed in most CT systems by means of geometric magnification. The image in [3] illustrates the idea of geometric magnification. It is easily imagined as a shadow effect. The closer a light source is to an object; the larger the object's shadow will appear on a wall or screen. Similarly, if that wall or screen were to move away from the stationary light source and sample, the object's shadow would increase in size but grow fainter. Optimizing the X-ray source and detector working distances is a very challenging and mentally stimulating task when sufficient signal, high resolution, and a short scan time are desired. As with anything, there are limitations on geometric magnification, where the idealization of a point source is destroyed and aberrations become overbearing.

One of the many challenges in micro-CT is that the visual quality of the 3D volume will be somewhat unknown to the investigator until the scan is finished and reconstructed. Though, with enough experience, a careful examination of a few 2D projections can provide enough information to feel confident about a CT scan. In the following sections, a set of exercises will uncover the effects of different imaging parameters on a data set and will equip a scientist with the understanding necessary to get a clean 3D volume. For the purposes of this investigation, biological samples will be probed though the procedure is not limited to a sample type.

Mounting a sample sounds like a trivial step but it is one of the most important and overlooked. Regardless of the application, the sample should be mounted in the most compact way possible. Considering the image in [3], imagine if the range of source and detector positions were limited because the sample stuck out in one direction and how that would affect geometric magnification. In addition to working distance, mounting greatly affects throughput. If the sample was an intact knee joint from a rat, it would make sense to mount the sample so that the tibia and femur stood upright. This way, the X-rays would pass through a short distance and there would be sufficient signal at the detector. The last thing to consider when mounting a sample is stability. The greatest enemy to micro-CT is movement. If the magnitude of sample movement approaches the resolution of the scan, it will probably be useless data. Movement should be limited by secure adhesion to a mount and with control to sample compositional change. For biological samples, that means ensuring that it will not change morphology through evaporation over the length of the scan. Suspension in agarose gel or wrapping in a thin layer of paraffin film are both possible approaches to avoid dehydration and movement.

X-ray energy will also have a great effect on the quality of the final 3D volume. The goal is to acquire a sufficient signal at the detector while having sufficient attenuation from the sample. X-ray attenuation follows Equation (1), where I is the final number of counts (intensity), I₀ is the initial number of counts, µ is the mass absorption coefficient which is unique to a given material and X-ray energy (widely published), ρ is the material's density, and x is the X-ray path length.

 (1)

Ideally, I/I₀ (a.k.a. transmission value) should be between 5 - 95% for all orientation of the sample, with the best results coming around the middle range. To check this value, take an image of the sample and then divide the image's pixel values by an image of air (i.e. with the sample outside the field of view). This normalization is commonly found in system software workflows. It is not often that biological samples demand the use of X-ray filtering at the source, so that will not be covered here. In addition to having an ideal transmission value, the ideal number of counts for any part of the sample is 5000 counts. To secure this, the exposure time per projection may need to be increased. This will increase the overall scan time. Figure 1 displays clean 2D projections.

Figure 1: 2D images of mouse spinal cord normalized against air at 0° (left) and 90° (right).

The following images give an overview of results that can be obtained from using micro-CT with the above stated procedure. Qualitative measurements on varying absorption can be directly noted based on these images. Quantitative data such as material porosity, feature size and distribution, etc. would require additional image processing in a different software.

Figure 3: 3D volume of mouse spinal cord (left) and two digital cross-sectional slices (right)

This experiment examined the many factors that should be considered when using micro-CT, particularly for a biological sample. This project was designed to help the investigator understand how their decisions will impact the data that micro-CT can provide. As demonstrated there are many dependent and sensitive parameters that should be considered including: mounting, X-ray energy, exposure time, source and detector positioning, number of projections, and total scan angular displacement. This exercise is meant as an introduction and only scratches the surface of control over a CT data set.

This experiment focused on giving an introduction to micro-CT with respect to imaging a biological sample but the application of 3D X-ray tomography extends to the worlds of microelectronics, geology, additive manufacturing, coatings, fuel cells, and so much more. These microscopes are used for inspection, failure analysis, characterization, quality control, and even non-destructive testing. Because real, 3D information is now accessible non-destructively, the geometries extracted from CT can be imported to simulations where objects can be tested virtually.