Dual-energy computed tomography (DECT) with virtual monoenergetic images has been shown to reduce metallic artifact associated with spinal hardware.1–7 Standard computed tomography (CT) is limited by metallic artifacts secondary to a combination of photon starvation, beam hardening, partial volume averaging, and scatter effects. Photon starvation is a process whereby insufficient photons reach the CT detector because of the high density of the metal. This noise is subsequently amplified during the reconstruction process, leading to characteristic streaks in the image.8 Beam hardening is a separate phenomenon that occurs when the polyenergetic x-ray beam is dramatically attenuated as it passes through the metal and leaves behind only a hardened high-energy beam of photons that also contributes to metallic artifact by causing characteristic dark streaks adjacent to the implant.8 A similar problem is frequently encountered in CT imaging of the cervical spinal canal, as shoulder artifact causes photon starvation in the lower cervical spine, limiting visualization of the spinal cord as well as any soft tissue process or disc herniation within the spinal canal. Consequently, evaluation of the cervical spine on CT is suboptimal in the assessment of compression of the cervical spinal cord and nerve roots, particularly when there is a concern for disc herniation causing spinal canal or foraminal compromise in the lower cervical spine. As a result, a cervical spine magnetic resonance imaging (MRI) is usually preferred in this clinical scenario.
Numerous studies demonstrate the reduction of metallic artifact associated with spinal hardware using DECT.1–7 Virtual monoenergetic images depict how the imaged object would appear if x-ray photons at only a single energy level were produced by the x-ray source. Bamberg et al1 compared DECT virtual monoenergetic images across discrete energy levels to standard CT and found reduced metal artifacts at discrete energy levels relative to standard CT. Guggenberger et al2 performed a similar study, evaluating posterior spinal fusion implants on DECT at monoenergetic energy levels ranging from 64 to 105 keV, confirming that as the energy level increased, there was significantly improved image quality and decreased metallic artifacts from metal implants relative to standard CT. Given the previously published data on the role of DECT in reducing metal artifact in the spine, our purpose is to explore the effect of using different monoenergetic levels on reducing shoulder artifact and identify the optimal monoenergetic images for visualization of the soft tissues in the spinal canal, and to compare DECT with regular CT.
MATERIALS AND METHODS
In this retrospective study, 171 consecutive patients referred for a CT of the neck between March 1, 2018, and October 31, 2018, had a DECT examination of the neck, identified from our picture archiving and communication system. The studies were requested for a specific clinical indication related to the evaluation of the soft tissues of the neck.
Images from each study underwent postprocessing to create reformats of monoenergetic images at discrete kiloelectron volt levels of 50, 70, 100, and 140 keV in the sagittal plane on a separate dedicated workstation. Standard CT studies performed for the same patient population within the subsequent 6 months of the original DECT were included when available for comparison. Diagnostic image quality was subjectively evaluated using a 5-point scale: grade 0, artifact totally obscured the spinal cord that could not be identified and was thus nondiagnostic; grade 1, marked artifact, with questionable delineation of the spinal cord and no confidence in the ability to grade stenosis, similarly of limited diagnostic value; grade 2, moderate artifact with partial anatomic delineation of the spinal cord and a low confidence in the ability to grade stenosis; grade 3, mild artifact that did not limit the anatomic delineation of the spinal cord and a corresponding high confidence in the ability to grade stenosis; and grade 4, no artifact with excellent delineation of anatomic structures and a very high confidence in the ability to grade stenosis. Two experienced subspecialty trained neuroradiologists with 15 and greater than 20 years of experience, respectively, separately assessed the images at differing keV levels for each study, assigning a subjective score (0–4) while blinded to the exact energy level under review. Standard CT examinations were similarly assessed later in a separate session. The subjective reading sessions were more than 3 months apart. Objective evaluation of image noise was performed for DECT keV level and for standard CT using the standard deviation of the CT attenuation in Hounsfield units measured via placement of a fixed-size region of interest within the spinal canal at the level of the worst artifact.
Image Acquisition and Reconstruction
All examinations were performed using a Discovery CT750 HD 64-slice CT scanner with rapid kV switching (GE Healthcare, Chicago, Illinois). A full rotational 360° helical scan was performed using GSI 37, 0.8 seconds, 260 mA with a helical pitch of 0.984:1 and a volume CT dose index of 10.98. Soft tissue density algorithm and bone windows were acquired with 2.5-mm axial and 2-mm sagittal and coronal multiplanar reformats.
A 1-way analysis of variance was used to assess differences in overall subjective image quality by the 2 readers for each study at standard CT and differing DECT keV levels. Objective assessment of image noise was similarly compared using analysis of variance. A post hoc Tukey test was performed to evaluate for differences between individual energy levels for both subjective and objective measurements. Data analysis was performed using R statistical software V.4.0.2 (R Project for Statistical Computing). Significance was set at P < 0.05 for all analyses.
This study was approved by our institutional research ethics board (REB# 20170509).
Within the study population of 171 patients (Table 1), 36 subsequently had a standard CT neck within 6 months of the original DECT study. Eleven studies were excluded because of nondiagnostic image quality, as determined by review of the images by the 2 readers and corresponding reports stating significant limitations in quality secondary to motion artifact (Fig. 1). This resulted in a total study cohort of 160 patients, 95 male and 65 female, with a mean age of 60.9 years ranging from 18 to 88 years.
TABLE 1 -
Basic Patient Data
||Mean (Range) or % (n/N)
Basic patient data for study participants are shown, with mean age and range, and sex.
Analysis of subjective image quality compared across the various energy levels varied slightly between the 2 readers (Table 2). Results for reader 1 demonstrated higher DECT energy levels to have improved image quality when compared with lower energy levels (Table 2). The exception was that 100 keV was superior to 140 keV, which was statistically significant (P = 0.01). Importantly, DECT at 100 keV also showed superior image quality than standard CT (P < 0.01; Table 2). Standard CT demonstrated improved image quality relative to the lower DECT energies of 50 and 70 keV.
TABLE 2 -
One-Way Analysis of Variance: Differences in Subjective Image Quality
|Subjective Score Difference
||Subjective Score Difference
|100 KeV and 50 KeV
|70 KeV and 50 KeV
|140 KeV and 50 KeV
|Standard CT and 50 KeV
|70 KeV and 100 KeV
|140 KeV and 100 KeV
|Standard CT and 100 KeV
|140 KeV and 70 KeV
|Standard CT and 70 KeV
|Standard CT and 140 KeV
“Subjective Score Difference” indicates difference in score between different keV levels and standard CT as a measure of subjective image quality. The score is expressed as the mean of subjective scores at the level of the worst artifact.
For reader 2, the higher the keV energy level, the more subjective image quality improved, with the exception of 100 keV, which demonstrated no statistically significant difference relative to 140 keV. Both DECT at 100- and 140-keV levels showed improved image quality relative to standard CT, results that were statistically significant. Standard CT demonstrated improved image quality than the lower energy level of 50 keV. No significant subjective difference was found between 100 and 140 keV or between standard CT and 70 keV.
Analysis of objective measurements demonstrated that images taken at 100 keV showed the overall lowest amount of noise among the various energy levels. Increasing DECT energy levels resulted in decreased noise (Table 3) with associated statistical significance, except for the 100-keV energy level, which demonstrated the lowest noise among the various energy levels. Images at 70 keV demonstrated no statistically significant difference from 100 keV, nor was a difference seen between 140 keV and 70 keV. In the paired comparisons between 100 and 140 keV, however, images at 140 keV showed more noise relative to images at 100 keV. Importantly, there was no significant objective difference between DECT at 100 keV and standard CT (Table 3). Standard CT was found to have a statistically significant advantage over DECT at the other monoenergetic levels of 50, 70, and 140 keV (Table 2) with a lower noise measurement or overall improved image quality (P values of <0.001, <0.001, and 0.02, respectively).
TABLE 3 -
One-Way Analysis of Variance: Differences in Objective Image Quality
|140HU and 100HU
|50HU and 100HU
|70HU and 100HU
|Standard CT and 100HU
|50HU and 140HU
|70HU and 140HU
|Standard CT and 140HU
|70HU and 50HU
|Standard CT and 50HU
|Standard CT and 70HU
* Variation in noise between different keV levels and standard CT as a measure of objective image quality. The noise is expressed as the standard deviation of hounsfield units (HU) measured at the level of the worst artifact.
Our results show that subjective image quality of the cervical spine is superior in DECT at 100 keV relative to all other DECT energy levels and, importantly, also to standard CT (Fig. 2). Objectively, 100 keV had the lowest noise measurements among the various monoenergetic energy levels, and there was no significant difference in objective noise measurements between DECT 100 keV and standard CT.
Virtual monoenergetic imaging simulates images acquired with a monoenergetic x-ray beam and thus can decrease beam-hardening artifacts. Previous studies have proven that high-energy virtual monoenergetic images (>70 keV) from different DECT technologies improve visualization of prosthesis and periprosthetic tissues, with no corresponding increase in radiation dose.1,2 With the advent of DECT techniques, methods to maintain a constant radiation dose relative to conventional CT have also been developed. Previous studies modifying the tube current and using advanced dose modulation algorithms led to improved visualization of surrounding tissues at specific energy levels without a corresponding increase in radiation dose, to the point where pathology was only seen at the optimized energy level.1,2 A study by Schenzle et al9 assessed dose and image noise for 2 different DECT settings with reference to standard scans, further establishing that DECT at various monoenergetic energy levels is feasible without an additional dose.
In several cases, periprosthetic lesions were in fact only discernible at virtual monoenergetic imaging performed at the optimal energy level and not with conventional CT at all.3 It stands to reason, then, that beam-hardening artifacts related to shoulder attenuation, like metal, can similarly be reduced for improved diagnostic efficacy of pathologies affecting the spinal canal. It has been shown that the ideal energy level for decreasing metallic artifact in the spine ranges from 108 to 149 keV, depending on the scanner, prosthesis, pathologic abnormality, and reader preference.1–3 This kiloelectron volt range is in close agreement with our ideal energy level of 100 keV for reducing beam-hardening artifacts related to shoulder attenuation. Pomerantz et al10 similarly evaluated different monoenergetic imaging levels to determine the optimum kiloelectron volt for the assessment of brain parenchyma on noncontrast head CT scans. They concluded that maximum contrast-to-noise ratio and signal-to-noise ratio for supratentorial gray and white matter was 65 keV, although slightly higher at 75 keV for the posterior fossa, as the elevation of energy led to a reduction in beam-hardening artifacts so commonly seen in the posterior fossa,10 further strengthening our conclusion that shoulder attenuation is optimally reduced at 100 keV, which can be used to improve diagnostic accuracy in the interpretation of pathologies of the spinal canal. Using virtual monoenergetic images at specific energy levels has also been used in the musculoskeletal system, with 1 study evaluating monoenergetic images at energy levels from 50 to 130 keV in differentiating osteoblastic metastases from benign bone islands in 122 patients with lung cancer. A much-improved relative diagnostic efficacy in differentiating these entities was found at 100 keV, with a sensitivity of 93.0% and a specificity of 93.3%.11
Although improved image quality can be quantified by measuring the image noise, the degree to which this impacts an interpreting radiologist's ability to assess underlying cord pathology such as disc protrusions or extrusions and subsequent spinal and foraminal compromise is indirectly implied rather than directly assessed. It is felt, however, that a significant difference in subjective image quality does sufficiently translate to an increased ability to identify and characterize such abnormalities with confidence, which will subsequently impact patient care and further management decisions. A recent study assessed the ability to characterize lumbar disc pathologies using DECT collagen sensitive maps, showing better correlation between DECT and MRI in the assessment of the degree of anteroposterior disc displacement and superior intramodality agreement with MRI relative to standard CT.12 A similar study used color-coded DECT virtual noncalcium (VNCa) reconstructions for the detection of lumbar disc herniation and spinal nerve root impingement. They showed dramatic diagnostic improvement over standard CT, with DECT VNCa images having equivalent diagnostic confidence, image quality, and noise scores to that of MRI.13 Although these studies used advanced DECT imaging techniques and examined findings in the lumbar rather than cervical spine, they highlight the trend of DECT's march toward greater delineation of the spinal canal and its associated pathologies. Furthermore, additional studies have recently highlighted the strength of DECT in the characterization of vertebral compression fractures, with one recent study confirming an increased detection rate of recent fractures when using VNCa DECT images relative to standard CT images alone, improving diagnostic accuracy of both experienced and less experienced readers and approaching the accuracy of MRI.14 Further studies comparing MRI and DECT in the cervical spine will help to elucidate the diagnostic differences and serve as a framework for future technological advances to bridge the gap between the two. In the meantime, these studies and our results indicate that DECT certainly improves on standard CT and can serve as a foundation to build toward greater diagnostic accuracy.
Some limitations must be discussed. Our analysis is limited by the small sample size of 160 patients, relatively small standard CT comparison group, and retrospective data collection. For this reason, statistical analysis is likely underpowered, manifested by the tendency of 100 keV to have less objective noise than 140 keV, or the lack of statistical significance in a difference between reader 2's subjective assessment of 100 and 140 keV. It remains to be seen whether standard CT or DECT at 100 keV shows less objective noise, however, and these results would benefit from validation in a larger prospective setting to delineate the full extent to which DECT at 100 keV differs in the evaluation of the spinal cord relative to the other energy levels and standard CT. Although our results demonstrate the differences between these imaging parameters and their impact on image quality, it would be interesting to identify the potential impact that patient weight has in reducing or increasing artifact within the various imaging methods because the differences in image quality can be potentially attenuated in patients with elevated body mass index because of the increased difficulty inherent in penetrating a greater degree of subcutaneous tissue. Finally, these results cannot be transferred to the thoracic or lumbar spine because this was beyond the scope of our study.
In conclusion, the optimal monoenergetic energy level for visualization of the soft tissues in the cervical spinal canal is 100 KeV. Increasing the monoenergetic energy level is associated with a reduction in shoulder artifact in the cervical spine. There is no difference between the monoenergetic images at 100 keV and standard CT, although the 100-keV image may subjectively improve upon the assessment of soft tissues and may play a role in patients with contraindications to MRI. Future studies are recommended to study the diagnostic accuracy of DECT in visualization of pathologies in the spinal canal.
1. Bamberg F, Dierks A, Nikolaou K, et al. Metal artifact
reduction by dual energy computed tomography using monoenergetic extrapolation. Eur Radiol
2. Guggenberger R, Winklhofer S, Osterhoff G, et al. Metallic artefact reduction with monoenergetic dual-energy CT: systematic ex vivo evaluation of posterior spinal fusion implants from various vendors and different spine levels. Eur Radiol
3. Meinel FG, Bischoff B, Zhang Q, et al. Metal artifact
reduction by dual-energy computed tomography using energetic extrapolation: a systematically optimized protocol. Invest Radiol
4. Pessis E, Campagna R, Sverzut JM, et al. Virtual monochromatic spectral imaging with fast kilovoltage switching: reduction of metal artifacts at CT. Radiographics
5. Lambert JW, Edic PM, FitzGerald PF, et al. Complementary contrast media for metal artifact
reduction in dual-energy computed tomography. J Med Imaging (Bellingham)
6. Dangelmaier J, Schwaiger BJ, Gersing AS, et al. Dual layer computed tomography: reduction of metal artefacts from posterior spinal fusion using virtual monoenergetic imaging. Eur J Radiol
7. Große Hokamp N, Neuhaus V, Abdullayev N, et al. Reduction of artifacts caused by orthopedic hardware in the spine in spectral detector CT examinations using virtual monoenergetic image reconstructions and metal-artifact
-reduction algorithms. Skeletal Radiol
8. Barrett JF, Keat N. Artifacts in CT: recognition and avoidance. Radiographics
9. Schenzle JC, Sommer WH, Neumaier K, et al. Dual energy CT of the chest: how about the dose?Invest Radiol
10. Pomerantz SR, Kamalian S, Zhang D, et al. Virtual monochromatic reconstruction of dual-energy unenhanced head CT at 65-75 keV maximizes image quality compared with conventional polychromatic CT. Radiology
11. Dong Y, Zheng S, Machida H, et al. Differential diagnosis of osteoblastic metastases from bone islands in patients with lung cancer by single-source dual-energy CT: advantages of spectral CT imaging. Eur J Radiol
12. Schömig F, Pumberger M, Palmowski Y, et al. Vertebral disk morphology of the lumbar spine: a retrospective analysis of collagen-sensitive mapping using dual-energy computed tomography. Skeletal Radiol
13. Booz C, Nöske J, Martin SS, et al. Virtual noncalcium dual-energy CT: detection of lumbar disk herniation in comparison with standard gray-scale CT. Radiology
14. Kaup M, Wichmann JL, Scholtz JE, et al. Dual-energy CT-based display of bone marrow edema in osteoporotic vertebral compression fractures: impact on diagnostic accuracy of radiologists with varying levels of experience in correlation to MR imaging. Radiology