A Review of Factors That Affect Artifact From Metallic Hardware on Multi-Row Detector Computed Tomography

Example—metal artifact due to bilateral knee prostheses. kVp, 120; mA, 400; ms, 500; mAs, 200; thickness, 1 mm.

Physics of metal artifacts: (A) CT uses a polychromatic X-ray beam. As the polychromatic beam passes through an object, the effective energy is shifted toward higher values; thus, the beam progressively becomes “harder” as it traverses the object. For a given material, the mass attenuation coefficient varies according to the hardness of the incident beam. Materials, such as metal, with a higher mass attenuation coefficient result in significantly more pronounced beam hardening. The calculated CT number is therefore technically “in error” and results in beam hardening artifact (B). Beam hardening causes an error in the data received at the detector as shown. The projected data value of the monochromatic beam is proportional to its pass length. By contrast, the polychromatic beam results in a smaller value compared to the result expected for a monochromatic beam. This differential—an error in the projected data value—causes artifacts in the final image. The degree of error—and the resultant artifact—depends on the mass attenuation coefficient and the thickness (the pass length) of the object. (Color version of figure is available online.)

Artifact is worse with stainless steel (A) than with titanium (B). (Color version of figure is available online.)

The number of interfaces between an X-ray beam and a piece of metal can also influence the artifact produced. More artifact can be expected with more complex shapes or greater numbers of hardware parts (e.g., plate and screws). (A) A single screw is simple in shape, so there is a relatively simple interface with the beam and more uniform attenuation and the artifacts are more pronounced in one direction. (B) Multiple screws in one cross section create more interfaces and the artifacts are dispersed across the entire image.

Increased kVp. CT images of a phantom scanned with 120 kVp (A) and 135 kVp (B), with all other factors remaining the same. These images demonstrate a very subtle improvement in the artifact using higher kVp. (Our scanner calibration does not allow for testing of other kVp values.) In a clinical setting, kVp must be balanced against dose considerations. (A) 120 kVp, (B) 135 kVp.

Increased mA. CT images of a phantom scanned with 100 mA (A) and 500 mA (B), with all other factors remaining the same. Note diminished artifact obtained with higher mA technique. (Color version of figure is available online.)

Scan rotation (exposure) time. CT images of a phantom scanned with exposure times of 500 mseconds (A) and 1500 mseconds (B), with all other factors remaining the same. Note diminished artifact obtained with longer scan rotation time. (Color version of figure is available online.)

Positioning. The same phantom was scanned at different angles in relation to the scanner gantry, and coronal reformats were performed (approximating same coronal “slice”). Phantom aligned parallel to long axis of scanner (A), phantom rotated 45° to long axis (B), and phantom at 90° (C). Note that the orientation of the artifact with respect to the bone varies as the phantom position within the scanner is rotated. (Color version of figure is available online.)

Reconstruction algorithm. Noncontrast CT scan of the ankle in a patient with Charcot arthropathy and osteoarthritis status post triple arthrodesis and medial column fusion. kVp, 135; mA, 350; ms, 500; Thk, 0.5 mm. Images were reconstructed in both bone (A) and soft-tissue (B) algorithms. Note that although bone detail is degraded, the hardware artifact is improved using the soft-tissue algorithm reconstruction.

Acquisition thickness. The difference in the projection data between thick and thin slice acquisition causes partial volume artifact (A). In the back projection image reconstruction, the output from the detector is amplified by being processed through a “log-amplifier.” For thick slice acquisition, the projected data value “−ln({1 + exp(−μL)}/2)” is not correctly projected as in the thin-slice acquisition; the projected data value is “μL” (the best possible situation is shown here) (B). Here, CT images of the phantom were acquired with (C) 2.0 mm × 16 detectors, (D) 1.0 mm × 32 detectors, and (E) 0.5 mm × 64 detectors, with all other factors remaining the same. Note the decrease in artifact in (E) vs (C) due to thinner section acquisition. (Color version of figure is available online.)

Reconstruction thickness. Left tibial fracture status post ORIF with intramedullary rod and interlocking screw. Noncontrast axial MIP reformations in 0.5 mm (A) and 5 mm (B) thickness are shown. Artifact is diminished with increasing thickness of the MIP reconstruction. kVp, 135; mA, 350; ms, 500; Thk, 0.5 mm. (Color version of figure is available online.)

Multiplanar reformat displays. Acquired axial image (A) shows artifacts extending 360° around the ankle. Sagittal (B) and coronal (C) reformats show artifacts restricted primarily to 1 direction. The sagittal plane corresponds to the dotted line; coronal corresponds to the solid line in (A).

The cone angle (θ) refers to the divergence of the radiation beam along the z-axis. The detector fan angle is represented by φ. Objects that are imaged with the central rays of the beam are correctly registered, while objects that are traversed and imaged with divergent rays, especially distant from the central imaging plane, are misregistered away from their true position, i.e. the cone beam effect, resulting in a form of misregistration artifact. The cone beam effect is more apparent with data acquired at the outer rows. Thus, the artifact is more pronounced for structures that are distant from the central rows of the detector. The effect is also more pronounced when structures are off-axis in the body because, unlike a central object, an off-axis object is detected by different detector rows, at different degrees of beam angulation. (Color version of figure is available online.)

Image reconstruction method. CT images of phantom reconstructed without (A) and with (B) correction for cone beam artifacts (TCOT; Toshiba Medical Systems, Tustin, CA). Note the decreased distortion of the cortex once the reconstruction technique is applied.

Postprocessing filter. CT images of patient with humeral hardware without (A) and with (B) postprocessing filter (BOOST; Toshiba Medical Systems). Note decreased conspicuity of artifact once the filter is applied. (Color version of figure is available online.)

Shoulder pain status post ORIF of fracture. Coronal reconstruction (A) and axial acquisitions CT images (B). Acquisition parameters: kVp, 135; mA, 440; ms, 1000; mAs, 440; thickness, 0.5 mm. Despite the presence of a large amount of metal and considerable artifact, the position of metal in relation to the bones can still be evaluated, and the quality of cortical and medullary bone, as well as surrounding soft tissue anatomy, can be assessed.

Radiographs of the ankle in a patient with Charcot arthropathy and osteoarthritis status post triple arthrodesis and medial column fusion (A). Absence of fusion is well-depicted despite the presence of hardware in the CT images (B). Acquisition parameters: kVp, 135; mA, 350; ms, 500; mAs, 175; thickness, 0.5 mm.

Ninety-year-old female with claudication and suspected ischemia. Despite the presence of a dynamic compression screw in her femur, the CT angiography with 3D volume rendering reconstruction was diagnostic and successfully demonstrated atherosclerosis without occlusion. Care must still be taken to avoid mistaking artifacts for pathology. Acquisition parameters: kVp, 120; mA, 400; ms, 500; thickness, 1 mm. (Color version of figure is available online.)