Typical Artifacts in Computed Tomography

Simon Schmid, summer semester 2018

Computed tomography (CT) represents a further development of radiographic testing using X-rays. In contrast to radiographic testing, computer tomography uses a computer to generate sectional images calculated from the absorption or attenuation values of X-ray signals passing through the object. Computer-based evaluation of a large number of X-ray images of an object, taken from different directions, allows digital reconstruction of a volume representation of the object. In CT various artifacts from a range of different origins can occur. ^[1] An artifact is defined as any discrepancy between the reconstructed values and the true attenuation coefficient of an object. ^[2] Artifact can be grouped in physics-based artifacts, hardware-based artifacts, patient-based artifacts and setup-related artifacts. ^{[3] [4]}

Physics-based Artifacts

Beam Hardening

X-ray beams are composed of photons within a specific range of energy (see X-Ray spectrum) and therefore they are not monochromatic. This is explained by the Planck-Einstein relation which states that the energy E of a photon is proportional to its frequency $//<![CDATA[ \begin{array}{l}v\end{array} //]]>$. $//<![CDATA[ \begin{array}{l}h\end{array} //]]>$ is the constant of proportionality with the name Planck’s constant. ^[5]

$//<![CDATA[ \begin{array}{l}E=hv\end{array} //]]>$

During the propagation of the X-Rays through the material, lower energy photons are absorbed more frequently and the beam becomes “harder”. This causes the frequency spectrum to shift in the direction of higher frequencies, but this does not mean that the maximum energy is increased. The intensity of beam hardening depends on the material and geometry of the object. ^[6] In Figure 1 the beam hardening effect is displayed on two objects of different thickness. Through the beam hardening object b) in Figure 1 appears thicker. ^[6] When radiating through a cylinder X-Rays passing through the middle are hardened more because they are passing through more material. ^[3] This is called cupping artifact and can be seen in Figure 2.

The beam hardening can result in dark streaks along the lines of greatest attenuation. ^[4] The beam hardening artifacts can be reduced by using a higher voltage for producing the X-ray radiation, which leads to a harder beam and thus less beam hardening artifacts. ^[2] Alternatively a flat piece of material, typically made of copper, can be used to “pre-harden” the beam and filter out the low energy radiation. ^[7] Another correction method for beam hardening is performed with the calibration on the average attenuation of the material. However high attenuation material can thus not be fully corrected because of the large difference to the mean value. This problem can be addressed through an iterative approach. Furthermore, a dual energy CT can be used to reduce beam hardening. ^[2]

Scatter and Noise

During the propagation of the radiation through the material a part of the photons are scattered and as a result diffracted from the original path. ^[1] The diffraction angle is random and the distribution of the scattered photons on the detector is dependent of the object. The scattering is caused by interactions with the material, for instance by Compton scattering. The biggest error is created if the scattered photons end up in a detector area which would count otherwise only few photons. The scattering leads to a reduced contrast in the image from the high attenuating structure, which is displayed darker in Figure 3, and the other part of the object. Through the reconstruction the scattered radiation results in streak artifacts. ^[2] Scatter can be reduced similar to the beam hardening artifacts with a reference measurement. ^[3] Furthermore smaller detectors can be used to limit the examined region in which the scattered photons could be detected. ^[1] Moreover, the detector can simply be moved further away from the object. This reduces the amount of scatter radiation reaching the detector. However, the distance-square law leads to a significantly higher necessary exposure. Furthermore, the structural limitations restrict the moving distance of the detector. With beam stop arrays the scattered radiation can be measured and used for a correction. Beam stop arrays use beam-stoppers to block primary radiation, thereby secondary signals can be measured in their shadows. ^[8] Moreover, collimators can be placed in front of the detector and ensure that the photons traveled in a straight path. As a result less scattered photons reach the detector. ^[2] There are many more methods for scatter correction which cannot be explained here for reasons of space. Instead, reference is made to ^[8]. Scattering also leads to noise. Noise represent inconsistent attenuation values in the image and can be caused by statistical error of low photon counts. ^[4] Therefore the attenuation values contain a large standard deviation, where constant values should be present. Noise can be reduced by applying higher voltages and so increasing the contrast and increasing the signal-to-noise ratio. ^[1]

Partial Volume Artifacts

Partial volume artifacts occur when high attenuation structures do not cover the entire acquisition slice of the volume, i.e. only a part of a pixel of the detector. Since common detectors detect the deposited energy in each pixel element, an inherent spatial averaging is performed which leads to a discrepancy between the measured and the true intensities and consequently also of the attenuation values. ^[9] Furthermore, partial volume artifacts appear when within the same reconstructed voxel, materials with different absorption coefficients are represented. ^[10] Partial volume artifacts can be reduced by using a thin acquisition section width. ^[3] However, this can lead to an increase in the noise level. To diminish the noise several slices could be summed. ^[9] Also, computer algorithm can correct partial volume artifacts. ^[2]

Photon Starvation

The photon starvation artifact describes an increased image noise in certain image sections, which is caused by an increased attenuation of the X-ray radiation due to differences in the morphology, i.e. a diminished signal-to-noise ratio. In the reconstructed volume the noise leads to stripes in CT-images. ^[9] Photon starvation artifacts can be reduced by increasing the current of the tube. This can lead in medical application to an unnecessary high dose of radiation. Alternatively, an automatic tube current modulation, based on how much material the radiation has to pass through, or an adaptive filtration can be used to avoid these artifacts. With an adaptive filtration the attenuation profile is smoothed with a software. ^[3]

Undersampling

The undersampling artifacts occur if a too small number of projections is used to reconstruct the CT volume. ^[11] A too large interval of projection, which means that too high angle steps are used, leads to so called aliasing effects. In the CT-image stripes appear to be radiation of the edges of high attenuating structures because of this effect. This can be seen in Figure 4, where CT-images with were reconstructed with different numbers of projections. Aliasing effects have only a small impact on the interpretability of the CT-images because they do normally not mimic any structure in the object. However, if high resolution is required aliasing artifacts need to be avoided. Artifacts through undersampling can be minimized by using a higher number of projections per rotation. ^[3]

Hardware-based Artifacts

Ring Artifacts

Ring artifacts are concentrically arranged, exactly circular rings that alternately appear darker or brighter in an CT-image. They result either from a malfunction or an insufficient calibration of the detector. Because of the consistent occurrence of mismeasurements over each angle position circular rings appear in the CT-images (i.e. for instance, a defective pixel is displayed in each projection at the same position within the image). A correct calibration of the detector prevents the occurrence of ring artifacts. ^[9] In Figure 5 right are ring artifacts displayed on an eraser. The rings are centered around the location of the axis of rotation. ^[1]

Tube Arcing

Impurities in the X-ray tube can result in a temporary short-circuit and a loss of X-ray output. This is called tube arching. Tube arching often leads to streaking artifacts due to the reconstruction. For preventing the occurrence of this behavior, a monitoring device is implemented into the power supply of the CT system. If a tube arching event is detected the tube in turned of and on again in a short period of time. ^[2]

Patient-based Artifacts

Metallic Materials

Metal objects in the scan field can cause severe streaking artifacts. These artifacts result from the high attenuation of the metal in comparison to the attenuation of a human body. ^[3] This can be seen as an example in Figure 6 on a CT-image of a plastic cable with copper wires inside. Additionally, the metal can lead to beam hardening, noise, scatter and partial volume artifacts. ^[4] Metal artifacts can be avoided if removable metal objects such as jewelry is taken of before CT examination. If the metal objects are nonremovable the metal can be excluded from the scans with gantry angulation. Furthermore, software corrections with various interpolation techniques enable the removal of streaking artifacts introduced by metal. ^[3]

Patient Motion

In medical CT, artifacts occur if a patient moves during the scan. Artifacts through motion are displayed in Figure 7 with the CT-images of a hand phantom. The patient motion can be involuntary or voluntary. Involuntary motion is for example the movement of the heart or the chest through breathing. Artifacts caused by voluntary motion can be avoided by positioning aids or using a shorter scan time. A shorter scan time reduces also involuntary motion artifacts. Moreover, artifact caused by heart movement can be diminished by an EKG-triggered CT protocol and artifacts through breathing can be reduced by instructing the patient to hold his breath. ^[10] Naturally, this artifact is not relevant for NDT-applications.

Incomplete Projections

An incomplete projection appears when a part of the projection is not available for reconstruction. In medical applications centering the patient in the scanning field of view is not always possible. As a result, a portion of the projections are truncated. The truncated projections produce through the reconstruction bright shading artifacts. Artifacts by incomplete projections can be avoided by centering the patient or the object in the scanning field of view. However, sometimes projection truncation is unavoidable. A lot of research has been conducted to solve this problem. A solution could be a software correction for example. ^[2]

Setup-related Artifacts

Helical Artifacts

Helical artifacts result from the interpolation of the measurement data, which is a requirement for the reconstruction of the spiral CT projections. These artifacts have a windmill-like appearance because several rows of detectors intersect the plane of reconstruction during the path of each rotation. Helical artifacts can be reduced by using thin acquisition sections. ^[3]

Cone Beam Artifacts

Cone beam effect artifacts result from the geometry of the X-ray beam at multichannel detectors. The X-Rays diverge on their path from the tube to the detector. In the reconstruction algorithm is a however a parallel beam geometry assumed. This leads to a conically shaped distortion of the voxel. For compensation are cone beam reconstruction algorithms available. ^[10]

Literature

1. Schulze, R.; Heil, U.; Grob, D.; Bruellmann, D.; Dranischnikow, E.; Schwanecke, U. ; Schoemer, E.: Artefact in cbCT: a review in Dentomaxillofacial Radiology (2011) 40:5 p. 265–273
2. Jiang H.: Computed Tomography. Principles, Design, Artifacts and Recent Advances. Society of Photo-Optical Instrumentations Engineers (2015)
3. Barrett, J. F.; Keat, N.: Artifacts in CT: Recognition and Avoidance in Radiographics (2004) 6 p. 1679–1691
4. Boas, F. E.; Fleischmann, D.: CT artifacts: Causes and reduction technique in Imaging in Medicine (2012) 4 p. 229-240
5. Griffiths, D. J.: Introduction to Quantum Mechanics. Prentice Hall international (2004)
6. Christoph, R.; Neumann, H. J.: Röntgentomographie in der industriellen Messtechnik. Moderne Industrie (2012)
7. Cullity, B. D.; Stock, S.R.: Elements of X-Ray Diffraction. Prentice Hall international (2001)
8. Schörner K.: Development of Methods for Scatter Artifact Correction in Industrial X-ray Cone-beam Computed Tomography. Dissertation at TUM Department of Physics (2012)
9. Kalender W. A.: Computed Tomography. Fundamentals, System Technology, Image Quality, Applications. Publicis Publishing Erlangen (2011)
10. Alkadhi, H.; Leschka, S.; Stolzmann, P.: Wie funktioniert CT?: Eine Einführung in Physik, Funktionsweise und klinische Anwendungen der Computertomographie. Springer publ., Heidelberg (2011)
11. Buzug, T. M.: Einführung in die Computertomographie: Mathematisch-physikalische Grundlagen der Bildrekonstruktion. Springer publ., Heidelberg (2004)