Basics and capabilities of X-ray CT
X-ray CT is a nondestructive method for observing the interiors of objects in
three dimensions. Most of us are familiar with X-ray CT and MRI images of the human body taken for
diagnostic purposes. Stacks of closely-spaced images can be visualized in three dimensions to create
perspective views of objects of interests. With proper calibration, such data sets can be used to
measure interior dimensions, surface areas or volumes of specific features. These data can then be
analyzed statistically to quantify the morphology of the structure. The technology is also common
in industrial applications where it is used to reverse engineer manufactured parts like engine blocks or
Quantitative analysis of X-ray CT data can go far beyond simple visual images.
A CT image is really just an array of numbers which have been assigned a color, either in gray scale or
full color scale. The numbers represent the X-ray absorption of individual area or volume elements
(called pixels and voxels respectively) of the scanned object. The magic comes from understanding
what controls the X-ray absorption of each voxel.
X-rays are attenuated as they pass through an object by two phenomena. One of
these (Compton scattering) depends almost solely on the bulk density of the object The other
depends on the average (effective) atomic number of the atoms that make up the object. To
learn more about these mechanisms click here
For example, if you scan an object that is composed uniformly of one chemical compound (say polyethylene),
a map of the X-ray attenuation is really just a map of its density structure. You can tell if there
are porous spots even if the pores are too small to resolve individually. If you calibrate against
polyethylene standards of known density, you can even plot two- and three-dimensional maps of the actual
density values. If you know the density of pore-free poly ethylene, you can present the data in terms
of quantitative porosity maps.
Porosity and Saturation
Sounds wonderful, but what if your sample is not of uniform composition and you still want a porosity map? The procedure requires a few more steps, but you can still get there. In our own work, we often work with natural rock samples whose mineral composition varies widely. Here we scan first with only air in the pores of the rock. We then vacuum saturate with a liquid and rescan. By subtracting the two, we get an image that represents the effect of the fluid in the pores. The average value of this three-dimensional image can be calibrated against the known bulk porosity of the sample to provide a scale of local porosity. Several related techniques are also available, including dual-energy scanning. This technique uses duplicate scans taken at different source voltages to separate the effects of density variation and compositional variation
We can extend the technique to studies using two or three different fluids (say oil, gas, and water)
simultaneously. Again with careful calibration, it is possible to obtain maps of fluid saturation
levels of each fluid phase. Repeated scans under different conditions allow us to perform real-time
experiments in the CT scanner, observing changes in fluid saturation and or porosity as conditions are
Similarly, if you have materials that are of uniform density but with compositional variations, a map
of X-ray attenuation represents the compositional variation in the sample. For example a Silica
(SiO2) glass might contain patches rich in GeO2 without significantly changing the bulk density.
X-ray attenuation will be greater in those regions because of the higher average atomic number of the
GeO2. Again, using properly calibrated standard materials, in this case glasses with known GeO2
content, it is possible to present maps and graphs of the distribution of GeO2.
Part of tuning the scanner for optimum results is selecting the best voltage and current at which
to run the X-ray source. As with many other parameters, there are trade offs that may require some trial and error to optimize.
Higher voltage results in more energetic, more penetrating X-rays. This allows lower power settings, smaller focal spot sizes
and higher resolution. However the difference in attenuation through various materials is less, leading to lower contrast images.
A classic example of this is shown by these images of an iron containing silicate rock. We used 300 kV
and 5.5 mA (about 1600 watts) to scan it with the industrial scanner resulting in the first image below. The voxel dimensions
here are about 100 microns on each side. The image is remarkably featureless. The second image is of the same rock taken with
the medical scanner. Here the voltage is only 130 kV. In this image a weathering “rind” consisting of different chemical
composition material is clearly visible where it was not in the industrial scanner. What makes this possible is the high power of
the medical scanner. For this image we used 250 mA current or 32000 watts. The lower voltage could not be used in the industrial
scanner because of its lower maximum 1600-watt power.
A Word about Resolution
Resolution is a touchy subject, frequently confused with voxel size. The difference
between the two is important to understanding what is possible and what is not. Voxel size refers to the
physical dimension of the voxels (three-dimensional pixels) created by the scanner. It is a purely
mathematical parameter based on the diameter of the image reconstruction, the number of pixels in the
computational matrix, and the thickness of the slice determined by the size of the detectors and/or the
aperture of the collimator slit. Medical scanners are usually standardized to use a pixel grid that is
512 x 512. The size of the reconstruction can be varied on our scanner from less than 100 mm up to more
than 500 mm. The corresponding voxel size ranges from about 0.2 mm (200 microns) up to about 1 mm.
The industrial scanner is more flexible and allows magnification of an object by
moving it closer to the X-ray source. The size of the pixel array is also arbitrary and can be set
mathematically as high as 4095 x 4095. Using a scan field of 4 mm for example could result in voxel
dimensions as low as 0.1 microns.
However, this is not the actual resolution of the resulting image. First, the
resolving power is no better than the size of the X-ray source. If the size of the source is large,
radiation from different sides of the source spot will project a shadow of a particular part of the scanned
object on a different part of the detector as shown in the illustration. This results in a blurring of the
of the object, reducing its resolution. The smaller the source spot, the better the resolution.
The number of individual detectors on the detector screen is also important. The
more detectors, the higher the resolution. In our case, the medical scanner has a maximum resolution
of about 0.25 mm or 250 microns. The industrial scanner has much higher resolution due to its smaller
focal spot size and can achieve about 5-micron resolution at lower power settings. Increased power,
using either higher voltage or higher tube current, requires a larger focal spot to avoid damage to
the target anode. Thus, resolution is decreased as X-ray power is increased.
The noise in an image also affects resolution in the sense of the ability to see separate
features of an image. There is always a random signal superposed on the actual image data that is caused by fluctuations
in beam intensity and in the response of individual detectors. The ratio of this noise to the actual image attenuation
is called the signal-to-noise ratio. Low signal to noise ratio will result in a “snowy” image, making it
difficult to resolve small objects. For example, a single dark voxel might represent a pore or it might be random noise.
Several adjacent dark voxels however are unlikely as a random event and are much more likely to be a real pore. I.e., large
pores are easier to resolve that small ones. Signal to noise ratio can be increased by increasing source power (unfortunately
necessitating increased focal spot size) or by taking longer to scan the object, thus averaging out random noise.
Optimizing for your Objectives
A fundamental part of any scanning procedure is tuning the system to optimize results and minimize cost.
This depends on what you want to know about your object or process. Source power and focus, magnification,
beam filtration, beam-hardening corrections, and number of views to reconstruct are among the factors we
can adjust. Basically, higher resolution and better image quality lengthen scan times and increase cost,
so it is important to understand the resolution needed to make the desired observations. Source settings,
as noted above, depend on the contrast needed. This depends on the range of densities and compositions that
make up the object to be scanned. Again, the settings will depend on what aspects of the object need to be
observed. We work closely with researchers to optimize the way in which data are collected and reconstructed.