|
|
Orthopedic medical devices and cross-sectional imaging:
protocols and artifacts
by David Melville, MD
Introduction
Post-surgical evaluation of bone and joints is critical to guiding clinical care following trauma and joint replacement, as it allows for assessment of fracture healing, infection and prosthesis failure or complication. While radiographs are the mainstay technique for post-operative imaging, CT and MRI provide critical problem solving tools for the assessment of recurrent injury or complication.
Magnetic resonance imaging (MRI) and computed tomography (CT) of bones and joints following orthopedic intervention presents both technical and diagnostic challenges due to imaging artifacts related to implanted hardware and metallic shavings, as well as post-surgical scarring and anatomic alteration. While the presence of orthopedic hardware may reduce the image quality and anatomic visibility, it should not be regarded as a contraindication to performing CT or MRI. Additional artifacts may arise due to additional technical and physical factors resulting in degraded image quality and, in some situations, simulated pathology. Musculoskeletal radiologists should be able to recognize the appearances of imaging artifacts, particularly those related to orthopedic hardware and be familiar with CT and MRI artifact reduction techniques.
Computed Tomography
Conventional polychromatic computed tomography (CT) employs a fan-shaped beam of radiation traveling through the patient as he or she is moved through the imaging gantry on an automated table. Images may be obtained at various slice thicknesses. A projection of this beam as it passing through the subject is recorded using a detector array, and multiple projections are acquired from the same table position to create a single axial image. Resulting images are reconstructed using a filtered back projection algorithm, and both spatial and contrast resolution are affected by the selected filter algorithm, such as bone or soft tissue.
The axial CT images may be reformatted in various planes with most musculoskeletal CT studies including axial, sagittal, and coronal reformations oriented along the anatomic axes of the imaged structure with additional reformations of varying obliquities to evaluate specific anatomic features, such as the glenoid fossa, acetabulum or femoral head-neck junction. The density of the imaged structures affects the CT appearance with more dense structures resulting in greater beam attenuation. They will consequently appear “brighter” or white while less dense structures, such as fat or gas, will appear “darker” or black.
The presence of very high density material, such as metal, will markedly attenuate the x-ray beam prohibiting some radiation from reaching the detector altogether. The highly attenuating metal results in beam hardening, excessive quantum noise, photon starvation, scattered radiation and edge effects with consequent beam alteration (Li, 2009). The loss and alteration of x-rays due to metal results in distortion of the projection data and will appear on reconstructed CT images as artifact, such as streaking and shadowing (Buckwalter, 2011; Watzke, 2004) (Figure 1). Further, multidetector CT scanners have resulted in new image artifacts as a result of incomplete sampling in the z-axis (table motion) direction, which is termed splay or windmill artifact, and appears as rotating lines surrounding regions with high-contrast boundaries, particularly about bone and soft tissue or lung (Flohr, 2005). Given the high attenuation of metal, splay artifact is substantially exaggerated at the margins of implanted hardware.
Conventional Post-operative CT Imaging
Given the effects of metallic hardware on the quality of CT imaging, concerted effort is required to optimize studies by minimizing artifact. Various factors affect the severity of metallic artifact, including hardware features, such as alloy type and geometry, and technical factors including kVp, mAs, pitch setting and image reconstruction algorithm.
A variety of metal alloys are employed for surgical hardware, and the degree of beam attenuation varies depending on the hardware composition. While it is not possible to ascertain the type of specific alloy used for the implanted hardware, the expected degree of attenuation can be assessed on the initial scout image. Highly attenuating (or dense) hardware will appear bright white on initial scout images or prior radiographs, while less attenuating metallic hardware will appear gray (Figure 2). The most radiopaque and attenuating hardware consists of stainless steel and cobalt chrome alloy, which was commonly used in older total hip prostheses (Buckwalter, 2007). Titanium appears much less radiopaque and is currently used in new femoral prostheses. Zirconium implants will appear less radiopaque than cobalt, but more radiopaque than titanium (Buckwalter, 2011). If the area of interest for the CT study does not include the implanted metallic hardware, the field of view and imaging plane can be adjusted. In addition, if the hardware is in the contralateral extremity, such as with knee or ankle implants, the imaged extremity can be extended while the contralateral extremity is flexed reducing the effect of the hardware on the quality of the CT images.
The size and geometric configuration of the metallic implant substantially impacts the appearance of the associated streak artifact. Implants with round transverse cross-section, such as intramedullary nails, will result in a uniform projection independent of beam orientation with minimal artifact on reconstructed images. On the other hand, implanted hardware with asymmetric cross-sectional geometry, such as a thin, but wide metallic plate, will result in propagation of streak artifact along the longest cross-sectional axis. The presence of perpendicular transfixing screws will further increase artifact.
Complex cross-sectional geometry implants will result in streak artifact along the axes of greatest cross-sectional thickness (Buckwalter, 2011). In the presence of multiple implants the artifact will be most severe along the axis containing both implants. For example, with bilateral hip arthroplasties, the artifact will be most pronounced along the right-left axis as the beam travels through both implants (Figure 3). By orienting the hardware with shortest dimension of the implant perpendicular to the table, the amount of associated streak artifact will be reduced.
Basic CT technical factors can be adjusted to reduce the effects of beam attenuation due to metallic hardware with greater benefits in the presence of smaller hardware, such as suture anchors or scaphoid screws (Buckwalter, 2011; White, 2002). First, beam energy may be increased to 140 kVp, as opposed to standard 120 kVp. This modification may increase the patient radiation dose by 20-30%, and caution should be exercised when imaging radiation-sensitive regions, such as the pelvis, neck and chest; however, distal extremity imaging may be performed without concern for significant increase in total body dose (Nickoloff, 2001; Biswas, 2009). Following this when employing multislice CT scanners, mAs can be increased when scanning through regions containing hardware while utilizing a low pitch setting. This combination of increased mAs and kVp will further increase radiation exposure. Close attention should be paid to dose modulation techniques, and technical adjustments should be made to minimize radiation exposure while obtaining diagnostic quality images.
Splay artifact may also be addressed using a lower pitch setting on multislice CT imaging. By increasing the thickness of reconstructed slices, the appearance of splay artifact decreases and will be essentially eliminated when the slice thickness is two or more times greater than the detector element thickness (Buckwalter, 2011). Therefore, it is typically desirable to reconstruct axial slices thicker than the minimum, particularly in routine pelvic imaging.
It is important to consider that thinner slices will reduce the effect of metallic artifact, and a reduced pitch will be useful to handle splay artifact in this setting. In addition, z-axis focal spot dithering (or flying z spot) technique can also address splay artifacts by acquiring additional samples along the z-axis to fill in undersampled data without requiring decreased pitch or resulting in compromised imaging quality (Flohr, 2005).
Bone (or “detail”) filters use a very sharp image filter, which accentuate fine spatial details, including image noise and artifact. Since musculoskeletal CT images are typically reviewed using “bone” windows (window 2000 HU, level 500 HU) these fine artifacts may be obscured and overlooked (Figure 4). In the presence of metallic hardware, these artifacts occur to a much greater extent and can result in degraded image quality. In the setting of metallic implants, it is helpful to use a soft image reconstruction, such as the standard filter used for general abdominal imaging, and even softer filters can be considered in the setting of highly attenuating implants, such as cobalt or steel, or if there is concern for soft tissue abnormality adjacent to hardware (Buckwalter, 2011). Evaluation of fine hardware in a small region of interest, such as a scaphoid fracture, offers an exception to this rule where a sharp bone algorithm may be beneficial (Ohashi, 2009).
When performing multiplanar reformations, it is important to consider that metallic streak artifact will degrade the quality of reformatted images, particularly when created using minimum pixel thickness, which is the default setting. By increasing the reformat thickness, the severity of hardware artifacts is reduced, and a slice thickness between 1-2 mm will provide adequate artifact reduction without exaggerated partial volume averaging (Buckwalter, 2011; Buckwalter, 2006).
Three-dimensional volume rendering techniques may also be useful to reduce hardware-associated streak artifact (Calhoun, 1999; Fayad, 2009). However, even so, the presence of severe streak artifact may result in rendering of unacceptable streaking limiting the quality and utility of the three-dimensional reconstruction. Given this, three-dimensional volume rendering may be most useful for evaluation of metal hardware rather than adjacent bone (Figure 5).
Back to Top
Back to top
Dual-Energy CT and Metal Artifact Reduction
Recent advances in CT imaging have permitted the application of dual-energy CT (DECT) theories, which were first considered in the 1970s (Chiro, 1979). Simultaneously acquiring two CT datasets using two beams at separate kilovolt peaks, most commonly 80 and 140 kVp, from the same anatomic location, allows computation of differences in attenuation values within a structure and determination of the dual energy index (DEI) (Coupal, 2014). This dual energy processing obtains material specific information and has been applied in many imaging applications, including diagnosis of gout (Nicolaou, 2010), automatic bone removal (Johnson, 2007), virtual unenhanced imaging (Takahashi, 2008) and renal stone composition characterization (Boll, 2009). At present, this technology is limited to specific DECT scanners and cannot be performed using conventional CT technology, which continues to be the greatest limitation of this technique. Table 1 shows a sample DECT protocol.
Table 1: Sample Dual-Energy CT Protocol |
Scan Direction: |
Cranial-caudal |
Tube voltage: |
100 kV/140kV |
Tube current: |
95-230 eff. mAs/85-165 mAs |
Pitch: |
0.6 |
Slice collimation : |
40 x 0.6 mm |
Slice width: |
1 mm |
Reconstruction Increment: |
1 mm |
Reconstruction kernel: |
D40f |
In addition to these useful applications, virtual monochromatic images may be synthesized using the data from the dual-energy scans with the benefit of reduced beam-hardening artifact (Alvarez, 1976). Moreover, the high photon energies in DECT reduce the effects of photon starvation with improvement in streak artifact (Coupal, 2014). These virtual monochromatic images appear as if created at a polychromatic, or wide, spectrum of energy levels. Using post-processing tools, the virtual monochromatic energy level of the image can be varied allowing a windowing effect. This permits the radiologist or technologist to find the ideal setting where the energy is high enough to overcome metal artifact, but not obscure adjacent soft tissues, and is typically in the range of 105-140 keV (Coupal, 2014; Meinel, 2012; Lee, 2012; Yoo, 2018). As with conventional methods, the metal artifact is reduced, but not completely eliminated (Figure 6).
Numerous studies have reported that DECT is an effective method of reducing metal artifact and improving diagnostic quality when directly compared to conventional polychromatic CT. An additional benefit of DECT is the ability to obtain images with improved diagnostic quality and reduced artifact without an increase in total radiation dose (Johnson, 2007; Pessis, 2013; Bamberg, 2011). This stands in comparison to the previously discussed methods of increasing kVp and mAs for conventional CT imaging.
Along with limited availability and cost of DECT technology, there are additional shortcomings that will be continued to be addressed as the technique is further developed. First, virtual monochromatic reconstruction requires omission of lower energy photon data to extrapolate high energy levels, resulting in noisier images (Coupal, 2014). Further, processing of dual-energy CT data can only be performed using a soft tissue algorithm with associated loss of high spatial frequency detail found in bone. At present, dual-energy images cannot be reconstructed from the CT scanner’s raw data set in coronal and sagittal planes. Consequently, multiplanar reformations can only be produced from image data on the workstation or with post-processing software, resulting in further loss of tissue detail. Finally, despite the advantages of DECT, large or extensive dense metallic implants continue to generate significant metallic artifact resulting in degraded imaging quality (Figure 7) (Yu, 2012). Virtual monoenergetic imaging (VME), another metal artifact reduction (MAR) application for dual-source CT, should be considered as an alternative to evaluate orthopedic hardware when other methods fail (Jagoda, 2018).
Post-Operative CT Imaging Applications
With the continued refinement of CT metal artifact reduction techniques, the modality has seen increased usage for hardware evaluation, particularly for associated complications, and progression of fracture healing. The interpreting radiologist should be familiar with the common indications for post-operative CT imaging to add further diagnostic value to image interpretation and ensure performance of the optimal examination to address the clinical question.
Immediate and Subacute Post-operative Complications
Radiographs remain the modality of choice to evaluate immediate post-operative appearances of fracture fixation and joint arthroplasty and confirm adequate fracture reduction and appropriate hardware placement. CT is uncommonly employed for acute post-operative extremity evaluation, but it may be considered following reduction of an intra-articular fracture or hardware placement, such as an external fixator, when planning another intervention or considering revision (Ohashi, 2009).
In contrast, due to the complexity of spine anatomy, pedicle screw placement in the spine is more effectively imaged with CT, which is 10 times more sensitive in the detection of medial pedicle cortex violation (Farber, 1995). Accurate screw placement with CT can help predict post-operative outcomes, as pedicle screw canal perforation less than 4 mm on images reformatted perpendicular to the screw is associated with no neurological complication (Laine, 1997). In addition to spinal cord or nerve root compromise, additional paraspinal injuries may be diagnosed more readily, including pneumothorax and hematoma.
Back to top
|