How to Reduce Beam Hardening Artifacts in CT

If you are reading this article, chances are you have beam hardening artifacts in your X-ray CT (computed tomography) images and want to get rid of them.

I run micro-CT scans on various samples from different industries and research areas daily, so I know how troublesome beam hardening artifacts can be.

The bad news is that beam hardening is an unavoidable physical phenomenon in X-ray CT when using the broad X-ray spectrum most laboratory CT scanners use.

        Read: What Is Beam Hardening in CT

The good news is that there are ways to reduce the beam hardening artifacts. In this article, I will explain the following common approaches, using examples:

Increase the X-ray energy by increasing the excitation voltage

Assuming all other conditions are the same, using higher X-ray excitation voltage lessens the hardening of X-rays and reduces related artifacts. This approach has only one downside: You lose contrast in the low-density area if the sample is a mixture of high- and low-density materials.

For example, the image below shows a CT cross-section of a servo motor collected at 130 kV with a 1 mm aluminum filter. This voltage setting produces high-energy X-rays, which helps them get through the metal parts well and reduces the beam hardening artifacts. However, the high-energy X-rays are not absorbed much by the plastic part, so the low-density plastic parts show low contrast.

meal and plastic in the same CT scan

 

If your sample is mostly made of high-density materials, or you are only interested in seeing the high-density parts, increasing the excitation voltage to generate high-energy X-rays effectively reduces beam hardening artifacts.

Below is a comparison of 40 kV and 130 kV scans of an aluminum diecast part. The 130 kV scan shows less significant beam hardening artifacts, such as shading and streaking. As a result, you can see the voids more clearly and analyze the dimensions more accurately.

Al diecast excitation voltage comparison

 

Increase the X-ray energy by using a thicker and denser filter

What if even the highest excitation voltage available on your X-ray source doesn’t reduce the artifacts enough? The effect of this second method is more subtle than drastically increasing the excitation voltage, but you can shift the peak X-ray energy by adding thicker or denser filters.

You see a comparison of four different filters below. All scans were collected at 130 kV, yet the contrast in the projections (top) improves as the filter becomes more absorbing, indicating the peak X-ray energy is increasing. You can see a significant reduction of artifacts and improvement in image quality when a more absorbing filter is used.

 

Al diecast filter comparison

Note that this approach has the same side effect as increasing the excitation voltage: reduced contrast in the low-density areas. In addition, a denser and thicker filter reduces the overall X-ray intensity, thereby increasing the data collection time.

Make the sample thinner by cutting or changing its orientation

Significant X-ray absorption causes severe beam hardening artifacts. High-density samples can cause it, but thick samples can also cause it, even when the density is relatively low.

Below is a comparison of large and small rock samples. Both scans were collected at 130 kV with a 1 mm aluminum filter.

sample size comparison

The 6-inch rock shows significant shading and a high noise level, making it difficult to see the density contrast inside the rock. Meanwhile, the 1-inch rock has no significant shading artifacts and shows low noise, making it easier to see multiple phases with different densities.

This approach’s side effect is that you might have to cut the sample and work with a smaller field of view.

Correct for beam hardening in the reconstruction process

There are several beam hardening correction techniques that can reduce artifacts. The empirical cupping correction and the empirical dual energy calibration are relatively robust because they do not require knowledge about the sample shape and materials or the X-ray spectra in use. They are also computationally less taxing than the iterative reconstruction technique. However, these techniques require refinement of the template images to produce optimum results.

For a sample with a relatively uniform and known approximate density, you can also use approximate X-ray spectra to apply a correction. This approach approximates the beam hardening effects to apply the correction and reduce artifacts (patent pending).

Below is an example of this correction applied to a scan of a coral reef. By comparing the 17-min scans without (left) and with the correction (center), you can see the correction reduces the streaks around the sample.

In the close-up of the corrected 6 hours and 20 min scan (right), you can also see that the gray level of the solid part is uniform and there are no visible shading (cupping) artifacts.

coral reef beam hardening correction

This approach does not have side effects, but it doesn’t necessarily work for all cases, depending on how the correction is applied. It works well when the sample has a relatively uniform and mid-level density of around 2 to 5 g/cc.

Ready to try these techniques?

I hope you found a technique you want to try to reduce beam hardening artifacts in your CT images.

When your goal is to segment images, try segmentation after reducing the artifacts to see if the correction had a sufficient effect. Also, remember you can fall back to using deep learning segmentation when you can’t reduce the artifacts enough to make thresholding segmentation work.

If you are not sure which technique is suitable for your samples, our team of CT experts can look at your CT images and help you figure out the best way to reduce beam hardening artifacts. You can talk to one of our CT experts by clicking the “Talk to an expert” button at the top right of the page or send us a message at imaging@rigaku.com.

 

Aya Takase
Head of Global Marketing at Rigaku. Aya holds a PhD in engineering from Osaka University and an MA in physics from Tokyo University of Science and has been with Rigaku for 25 years. She started in the X-ray Diffraction Application Lab and transitioned to X-ray Imaging in 2017. She now focuses on providing educational and helpful content for X-ray users. Her goal: Help non-expert X-ray users achieve expert results with less time and effort. She has worked on many projects designing automated and user-friendly X-ray instruments and analysis software. She is very passionate about helping people learn more about X-rays and working with X-ray users to solve their specific problems.