Overview
Source: Peiman Shahbeigi-Roodposhti and Sina Shahbazmohamadi, Biomedical Engineering Department, University of Connecticut, Storrs, Connecticut
It's a little-known fact that the discovery and (inadvertent) use of X-rays garnered the first ever Nobel Prize in Physics. The famous X-ray image of Dr. Röntgen's wife's hand from 1895 that sent shock waves through the scientific community looks like most modern day 2D medical X-ray images. Though it is not the newest technology, X-ray absorption imaging is an indispensable tool and can be found in the world's top R&D and university labs, hospitals, airports, among other places. Arguably the most advanced uses of X-ray absorption imaging involve attaining information like the kind found in a 2D medical X-ray but realized in 3D through a computed tomography (CT or micro-CT). By taking a series of 2D X-ray projections, advanced software is capable of reconstructing data to form a 3D volume. The 3D information can, and most likely will include information from the inside of the probed object without having to be cut open. Here, a micro-CT scan will be obtained, and the major factors impacting image quality will be discussed.
Principles
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).
Subscription Required. Please recommend JoVE to your librarian.
Procedure
1. Mounting a Sample (Bone)
- For examining a network of bones, like a spine, suspend the structure in an agarose gel and allow to cure in a very thin-walled plastic tube (Figure 2). The thinness of the tube is very important, greatly affecting the signal throughput and overall image quality. This in turn affects your ability to resolve features. The transmission value of the tube should be as close to 100% as possible.
- Mount the tube on sample stage with tape or by making a custom stand, ultimately ensuring that the sample is stationary and stable when the stage rotates.
Figure 2: Mouse spine suspended in agarose gel inside thin-walled plastic tube sitting on sample stage of micro-CT system.
2. Image Acquisition
- Turn on the X-ray source to an energy around the range of 90 keV (90 kV and 10 W).
- After the source warms-up and settles on the energy, acquire an image through the system's software.
- Check the transmission value by normalizing the image against an image of air. For automatic reference acquisition and application, ensure the sample can move in a given direction without crashing.
- If image has too high of a transmission, lower the energy incrementally until the transmission value is sufficient. Be sure to increase the exposure time accordingly so that the image doesn't appear noisy and grainy. If the image has too low transmission, increase the energy incrementally until the transmission value is sufficient.
- Begin to move the X-ray source closer to the sample, while being very careful not to crash them. Bringing the source as close as possible to the sample is a step in the direction of maximizing throughput and securing the best possible resolution.
- As this is done, refine the field of view of the sample by shifting the sample stage with its linear actuators.
- The software will show a parameter known as pixel size which is treated similar to the current resolution (though it is indeed different).
- If the number is still too large and the source is very close, the detector can begin to move away from the sample.
- If the number is too small, cautiously move the detector closer to the sample.
- Try different optical objectives and detector positions as well but beware of the complication this introduces when trying to optimize the scan parameters.
- Check every orientation of the sample to ensure there are no crashes to find the best exposure time. This is performed because of the change of working distance, pixel size, and sample position.
- Slowly rotate the sample by 2-degree increments while monitoring its position relative to the source and detector via the in-cabinet camera. Be sure to move the source and detector away if a collision might occur.
- Find the longest X-ray path length that results in the lowest number of counts/transmission values and find the exposure time needed to secure about 5000 counts everywhere.
3. Tomography Submission and Reconstruction
The reconstruction process from a user perspective is not more complicated than any of the other parameter selections made in the earlier steps. However, the programming and computational expense for this process is actually quite substantial. Users must aim to best understand what is happening underneath the smooth software UI and how decisions impact the final product. Many CT systems utilize iterative algebraic reconstruction algorithms in which the 2D projections are converted into a series of linear equations describing pixel values. Some other systems utilize filtered back projection algorithms where Radon transforms convert the projections to a sinogram and are then passed through a series of line integration operations. Of course, some use other approaches and even hybrid methods. At the lowest level of involvement with these algorithms, it is known that number of projections and the total rotated displacement have an impact on the final reconstructed volume.
- First, decide whether to scan over 180° or 360° based on the aspect ratio of the sample. If the sample has a high aspect ratio such that the X-ray path length is about 4 or more times longer at the 90°-orientation than the 0°-orientation than a 180°-scan would be a smart choice. (The argument being that the information gathered on the short X-ray path length is not that different from one side to the other. If more projections can be dedicated to the long X-ray path length orientations than the data set will benefit. The angular displacement between projections will be lower and there will more information from those tricky orientations being fed into the reconstruction algorithms.) If the aspect ratio is not that high, use 360°-scans.
- Next, choose the number of projections and total angular displacement, which will dictate the angle between projections. The smaller this angle, the less interpolation will be conducted and less fine feature information will be truncated. (A balance must be established because more projections mean a longer scan time, which correlates to a larger window where samples can move, less time to scan other things, and a shorter source lifetime. A rule of thumb is to have at least 800 projections over 360° and to not exceed 3200 projections.
- Submit the scan.
- After the tomography finished (usually between 4 - 16 hours), bring 2D series of images into the system's (or some open source) reconstruction software.
- Select the optimal center shift correction. Center shift is a parameter that aligns the projects to line up (think of a roughly shuffled deck of cards that need to be collected and aligned to sit like a neat stack). This value is usually somewhere between -10 to 10 pixels.
- Select the optimal beam hardening correction coefficient. Beam hardening correction is an artificial removal of contrast produced by sample filtering. If a sample is thick enough or contains a range of light and heavy materials, it will have false contrast based on the attenuation of low-energy (soft) X-rays. This should be applied conservatively. An average value is somewhere between 0 - 0.5.
- Submit reconstruction.
A micro-CT scan is a three-dimensional image that is created from a series of X-ray images at different orientations. The first Nobel prize in physics was awarded to Dr. Rontgen in 1901 for the discovery of X-rays and their use as demonstrated by imaging his wife's hand.
X-ray absorption imaging continues to be an indispensable tool, especially in university labs and hospitals. One of the most advanced uses involves taking a series of two-dimensional X-ray projections to reconstruct a three-dimensional volume. This is known as computed tomography or CT. Micro-CT uses the same basic method, but produces much higher resolution images of smaller volumes.
This video will demonstrate how to obtain X-ray images and use them to produce a micro-CT scan, illustrate the principles of the technology, and finally, discuss some of its applications.
Now let's look at how X-ray images are formed and examine the principles behind assembling them into a micro-CT scan.
A typical micro-CT system features three primary components, an X-ray source, a rotational stage for the sample, and a detector. In the X-ray source, negatively charged electrons are shot in a vacuum where they strike and interact with a target. The electrons decelerate through the target material and emit X-rays. This phenomenon of X-ray generation is known as bremsstrahlung, or braking radiation. The X-rays then leave the source and are either absorbed, scattered, or transmitted by the sample before arriving at the detector. Absorption is the predominant interaction measured in micro-CT, which is due to the large variation in X-ray absorption by different materials in the sample.
Bones contain a lot of atomic calcium and absorb X-rays more so than soft tissue. The absorbed X-rays do not reach the detector and the bones appear white in an X-ray. The output of the tomography is a series of 2D projections at different orientations that can be reconstructed into a 3D volume. The X-ray energy needs to be balanced so there is sufficient attenuation in the sample and signal at the detector.
The intensity, or number of X-rays measured, I, depends on the intensity before attenuation, I-naught, the mass absorption coefficient of the material, mu, the density of the material, rho, and the X-ray path length, X. Ideally, the transmission value I over I-naught, should be between five and 95% for all orientations of the sample, with the best results at the middle range. This value is checked by taking an image of the sample and then dividing the image's pixel values by those in an image of air.
Now that you understand the principles behind micro-CT scans, let's now demonstrate how to produce one.
In this demonstration, a micro-CT scan of the spinal column of a mouse will be obtained.
First, obtain a sample that is suspended in agarose gel. The sample should be cured in a thin walled plastic tube to prevent sample movement and dehydration. The tube walls should be as thin as possible to reduce signal throughput and improve overall image quality.
Then, mount the tube on the sample stage using tape or by making a custom stand. Ensure that the sample is stationary and stable when the stage rotates. Now, turn on the X-ray source and set it to an energy of 90 kiloelectron volts or a voltage of 90 kilovolts, and set the power to eight watts. Once the source warms up, acquire an image through the system software. For automatic acquisition and application, ensure the sample can move in a given direction without crashing. Check the transmission value by normalizing the image against an image of air.
If an image has too high of a transmission, lower the energy incrementally until the transmission value is sufficient. If the image has too low of a transmission, increase the energy incrementally until the transmission value is sufficient. If the sample appears noisy or grainy, increase the exposure time as needed.
Next, move the X-ray source as close to the sample as possible to maximize throughput and obtain the best possible resolution. Be careful not to crash them together. Refine the field of view of the sample by shifting the sample stage using its linear actuators. Then, locate the pixel size of the image. If the CT system supports optical magnification by converting the X-ray signal to visible light signal, try different optical objectives and detector positions. However, be aware that this will affect the scan parameters.
After making necessary adjustments, find the optimal exposure time. Slowly rotate the sample in two-degree increments while monitoring its position relative to the source and detector via the in cabinet camera. Move the source and detector further apart if a collision might occur.
Finally, find the longest X-ray path length that results in the lowest number of counts and determine the exposure time needed for approximately 5,000 counts everywhere.
Now, let's see how a series of images can be acquired. First, select a scan over either 180 degrees or 360 degrees based on the aspect ratio of the sample. For high aspect ratios, select a 180 degree scan, and for low aspect ratios, select a 360 degree scan. If the X-ray path length is four or more times greater in one direction than the other, choose a 180 degree scan.
Next, choose the number of projections and total angular displacement that will dictate the angle between projections. A smaller angle decreases the amount of interpolation of fine feature information, but increases the scan time. A rule of thumb is to have at least 800 projections, but usually no more than 3,200 projections over a 360 degree scan.
Now, submit the scan. The full series of X-ray images will take on the order of a few to tens of hours to acquire. Once the scan is complete, load the series of 2D images into the reconstruction software. Now, select the optimal center shift corrections so that the images line up around a shared axis. This value is usually somewhere between negative ten and ten pixels.
Next, select the optimal beam hardening correction coefficient. This removes false contrast deriving from low energy X-ray attenuation. An average value is somewhere between zeri and 0.5. Then submit the reconstruction. Once the micro-CT scan has been reconstructed, the results are ready for analysis.
Here is a representative micro-CT scan that was obtained using this procedure. Here, we see the 3D volume of a mouse spinal cord. Further image processing two digital cross-sectional slices allows quantitative data such as material porosity, and feature size can be obtained using software tools. The spacing between the sections of a vertebrae and intervertebrae passageways was measured to be on the order of hundreds of microns.
Here is another micro-CT scan that was obtained of a rat's knee. We can see the porosity of the cortical bone and can measure the spacing within the cortical bone of a rat's knee and the thickness of the articular cartilage.
You have just seen a micro-CT scan of a mineralized biological sample, but the applications of 3D X-ray tomography extend to the worlds of microelectronics, geology, additive manufacturing, fuel cells, and more. We will examine a few other instances.
High resolution X-ray images of animal soft tissues can be obtained despite their natural low X-ray absorption. This is accomplished by use of simple contrast staining. In this example, a mouse hind brain is stained using Lugol's iodine solution prior to imaging. The sample is then prepared, loaded, and X-ray images are taken. Finally, a micro-CT scan is created clearly showing lesions in the hind brain.
Micro-CT can be used to characterize the micro structure of electronic devices. In this example, an LED is scanned. Micro-CT scans enable engineers to analyze device failure or reverse engineer a device.
Three-dimensional structures can be created from micro-CT data. In this example, a rat is anesthetized and scanned. The data can then be analyzed to distinguish bone structure from surrounding tissue. Finally, a physical model of the result can be created using a 3D printer.
You've just watched JoVE's introduction to creating 3D micro-CT scans from 2D X-ray images. You should now understand the principles behind X-ray imaging, the relationship between X-ray images and CT scans, how to produce a micro-CT scan of a sample, and some applications. Thanks for watching!
Subscription Required. Please recommend JoVE to your librarian.
Results
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)
Subscription Required. Please recommend JoVE to your librarian.
Applications and Summary
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.
Subscription Required. Please recommend JoVE to your librarian.
References
- http://www.spectroscopyonline.com/tutorial-attenuation-X-rays-matter [cited 1 November 2017]
- http://hyperphysics.phy-astr.gsu.edu/hbase/quantum/xrayc.html [cited 1 November 2017]
- A.G. Rao, V.P. Deshmukh, L. L. Lavery, H. Bale, "3D investigation of the microstructural modification in hypereutetic aluminum silicon (Al-30Si) alloy." Microscopy and Analysis 2017 [cited 1 November 2017].
