Section 1. Introduction
The spine is the third most common site of metastatic disease after the lungs and liver [1–2]. Several imaging modalities exist for the evaluation of epidural disease, namely computed tomography (CT) and magnetic resonance imaging (MRI). CT myelography (CTM) involves the injection of contrast material into the cerebrospinal fluid prior to imaging and requires ionizing radiation. Though previously widely used, CTM has steadily fallen out of favor to MRI at many institutions and in some cases, is used alone in radiation therapy planning. MRI is notable for its non-invasive nature, excellent visualization of soft tissues, and lack of exposure to ionizing radiation. CTM, however, can provide improved visualization of bony structures and is more resistant than MRI to geometric distortion and patient motion due to the rapid acquisition of images [3]. Additionally, CTM can be used in patients with contraindications to MRI and can also provide a direct correlation of electron density to doses used for radiation therapy planning [4]. With recent advances in radiation therapy, including intensity-modulated and image-guided radiation therapy, which aim to improve tumor targeting thereby sparing normal tissue, precise tumor localization for treatment planning often requires the use of CT integrated with MRI [5–7].
A previous study comparing CTM and MRI in the evaluation of degenerative disease in the cervical spine found that MRI was superior in its reliability for assessing the degree of nerve root compression while CTM provided greater reliability in the assessment of bony lesions and foraminal stenosis [8]. Other studies focusing on the lumbar spine also found that CTM provided greater reliability and reproducibility in the assessment of the extent of lumbar stenosis than did MRI [9] and that MRI underestimated the degree of spinal root compression in the lumbar spine [10].
While both imaging modalities have complementary advantages, studies to date have not investigated a composite evaluation of spinal epidural disease. Furthermore, given the limitations of MRI, it has not been studied whether there is merit in continuing to use CTM to complement MRI for radiation therapy planning. This study is the first to investigate whether axial and sagittal projections of the spinal epidural space using both T1 and T2-weighted MRI compared to CTM produce compatible levels of inter- and intra-rater reliability in measuring the extent of epidural disease by neuroradiologists trained in both imaging modalities. The results of this study will be useful in determining whether CTM should continue to be used in conjunction with MRI when evaluating spinal metastases and planning radiation therapy.
Section 2. Methods
2.1. Subjects
This retrospective study was approved by an institutional review board with waiver of informed consent and conducted in compliance with Health Insurance Portability and Accountability Act regulations. We reviewed medical records for patients who received radiotherapy for spine metastases registered from 2009 to 2016 at a single institution. Only patients whose MRI reports confirmed the presence of ventral spinal epidural disease at any spinal level and who also underwent CTM within seven days of spinal MRI were included in the study. Patients who underwent a procedure which could potentially affect tumor volume, such as kyphoplasty, between the acquisition of MRI and CTM images were excluded. For each patient, we compiled six images for review: sagittal and axial CTM, T1-weighted and T2-weighted MR images.
2.2. MRI Acquisition
MRI sequences were acquired using a 1.5-T GE scanner (Optima 450W, Milwaukee, WI), equipped with an 8-channel cervical-thoracic-lumbar surface coil. All patients underwent routine MRI, including sagittal T1 (field of view [FOV], 32–36 cm; slice thickness, 3 mm; repetition time [TR], 400–650 ms; flip angle [FA], 90°, Matrix 256×192), sagittal T2 (FOV, 32–36 cm; slice thickness, 3 mm; TR, 3500–4000 ms; FA, 90°; Matrix 256×192), axial T1 (field of view [FOV], 32–36 cm; slice thickness, 3 mm; repetition time [TR], 400–650 ms; flip angle [FA], 90°, Matrix 256×192), axial T2 (FOV, 32–36 cm; slice thickness, 3 mm; TR, 3500–4000 ms; FA, 90°; Matrix 256×192).
2.3. CTM Acquisition
Fluoroscopic-guided lumbar puncture was performed on each patient followed by intrathecal contrast administration. CT images were acquired with a 64-slice clinical CT system (Discovery CT750 HD, GE Healthcare, Waukesha, WI) equipped with the Hi-Res and Hi-Res/HD modes. Eight reconstruction kernels were used for the conventional scan mode. Seven additional HD kernels were used for the Hi-Res scan. Acquisition parameters are 120 kV, 800 mAs, small focal spot (1.0 × 0.7 mm), large bowtie filter, and 20 mm detector collimation width. The acquired data were reconstructed using a slice thickness of 5 mm, 512 × 512 image matrix, and 50 mm FOV.
2.4. Measurement
Forty-six unique sites of spinal epidural disease were identified and six images were obtained for each site: CTM, T1- and T2-weighted MR in both sagittal and axial projections. Each subject’s full set of imaging was previously reviewed by an independent radiologist not participating in this reliability assessment. A single image with the largest width of epidural disease was selected as the representative slice. Four neuroradiologists trained in both modalities independently assessed each static image. Each image was shown to raters twice in randomized fashion; thus, each rater evaluated a total of 552 images. Each rater was asked to evaluate whether ventral epidural disease was present, absent, or could not be determined by the image shown. If ventral epidural disease was present, raters measured the maximum width of tumor infiltration into the spinal canal in an anterior to posterior orientation on both axial and sagittal projections. All measurements were made using ImageJ, an open source Java platform which imports series of images and calculates the width of the tumors as indicated by each neuroradiologist using pixel- based scales [11]. Krippendorff’s alpha, a coefficient that measures reliability among different raters by calculating disagreement, rather than agreement, was used to calculate the degree of both inter-rater and intra-rater agreement. All analyses were conducted using R (R version 3.3.3).
Section 3. Results
Forty-six sites of ventral epidural disease were evaluated from 31 unique patients (Figure 1) and the average time between acquisition of MRI and CTM was 4 days (range 0–7) (Table 1). Sixty- two images were excluded from the final analyses due to disagreement from at least one rater with regards to whether ventral epidural disease could be measured from the image presented; the number of discordances in which at least one rater did not believe that epidural disease could be measured from the images did not differ significantly between CTM, T1-weighted and T2- weighted MRI (range 9–11). An example of rater measurements for each image projection for a single tumor is shown in Table 2.
Figure 1: Representative images from a single patient:
CTM axial (a) and sagittal (b), T1- weighted MRI axial (c) and sagittal (d), and T2-weighted MRI axial (e) and sagittal (f).
Table 1.
Patient Demographics
| M:F | 18:13 |
| Mean age | 56 [21–84] |
| Mean days between acquisition of MRI and CTM | 4 [0–7] |
| Primary diagnosis | |
| Renal cell carcinoma | 8 (25.8%) |
| Sarcoma | 5 (16.1%) |
| Melanoma | 4 (12.9%) |
| Breast cancer | 2 (6.5%) |
| Colon cancer | 2 (6.5%) |
| Lung cancer | 2 (6.5%) |
| Prostate cancer | 2 (6.5%) |
| Other | 6 (19.4%) |
Table 2. Rater measurements for a single tumor.
Example of rater measurements of a single tumor in sagittal and axial projection in each imaging modality. All measurements are in millimeters (mm).
| Rater 1 | Rater 2 | Rater 3 | Rater 4 | |
|---|---|---|---|---|
| MR_S_T2 | 6.225 | 6.187 | 5.845 | 5.792 |
| MR_S_T2 | 7.332 | 5.984 | 5.132 | 6.822 |
| MR_S_T1 | 6.627 | 3.938 | 4.215 | 3.402 |
| MR_S_T1 | 7.523 | 3.940 | 5.246 | 5.136 |
| CT_S | 3.641 | 6.248 | 4.320 | 5.692 |
| CT_S | 5.561 | 4.563 | 4.694 | 5.274 |
| CT_A | 4.668 | 4.075 | 6.249 | 4.747 |
| CT_A | 5.935 | 5.353 | 3.540 | 4.236 |
| MR_A_T1 | 4.009 | 5.554 | 3.571 | 4.007 |
| MR_A_T1 | 5.497 | 4.945 | 4.014 | 5.378 |
| MR_A_T2 | 4.444 | 7.174 | 3.685 | 3.901 |
| MR_A_T2 | 6.828 | 4.747 | 5.498 | 5.565 |
3.1. Inter-rater reliability
On sagittal projections, inter-rater agreement was similar using CTM and T1-weighted MRI (Krippendorff’s α=0.63 and 0.66, respectively) and slightly greater than T2-weighted MRI (Krippendorff’s α=0.59). On axial projections, inter-rater reliability was greater in CTM images (Krippendorff’s α =0.67) than in T1-weighted or T2-weighted MRI (Krippendorff’s α =0.56 and 0.57, respectively). (Table 3)
Table 3. Inter- rater reliability.
Interrater reliability for each imaging modality in the axial and sagittal projections. Krippendorff’s alpha coefficient range from 0 (perfect disagreement) to 1 (perfect agreement).
| Sagittal projection | Krippendorff’s alpha |
|---|---|
| CTM | 0.63 |
| MRI T1 | 0.66 |
| MRI T2 | 0.59 |
| Axial projection | Krippendorff’s alpha |
| CT | 0.67 |
| MRI T1 | 0.56 |
| MRI T2 | 0.57 |
3.2. Intra-rater reliability
On sagittal projections, intra-rater reliability ranged from Krippendorff’s α = 0.73 to 0.86 for CTM compared to α =0.73 to 0.83 for T1-weighted MRI and α =0.67 to 0.88 for T2-weighted MRI. On axial projections, intra-rater reliability ranged from Krippendorff’s α = 0.69 to 0.84 for CTM and α =0.54 to 0.82 for T1-weighted MRI and α =0.75 to 0.80 for T2-weighted MRI. (Table 4)
Table 4. Intra- rater reliability.
Intra-rater reliability for each of the four evaluators for sagittal and axial images.
| Sagittal projection | Krippendorff’s alpha |
|---|---|
| CTM | [0.73, 0.86] |
| MRI T1 | [0.73, 0.83] |
| MRI T2 | [0.67, 0.88] |
| Axial projection | Krippendorff’s alpha |
| CT | [0.69, 0.84] |
| MRI T1 | [0.54, 0.82] |
| MRI T2 | [0.75, 0.80] |
Section 4. Discussion
Evaluating the extent of epidural disease on CTM axial images leads to slightly higher inter-rater reliability than MRI. Evaluation of sagittal images suggests that inter-rater reliability is similar between CTM and T1-weighted MRI, but that CTM is slightly superior to assessments made using T2-weighted MRI. The lower inter-rater agreement of epidural disease using sagittal T2- weighted MRI could be due in part to CSF pulsation artifact in which the pulsation effect may obscure accurate evaluation of epidural disease [12]. Intra-rater reliability showed no significant difference between imaging modalities.
The ability to accurately measure epidural disease on imaging has important clinical implications, specifically when used for radiation therapy planning. It has been shown that a 5% difference in radiation dose can produce up to 20% changes in tumor control probability at the steepest segments of a dose-response curve [13]. When used for radiation planning, the high accuracy of CTM will allow radiation oncologists to plan the most optimal dose for tumor control. Furthermore, in patients who have contraindications to obtaining MRI, our study has shown that CTM can be reliably used as an alternative to MRI.
There are several limitations to this study. The neuroradiologists participating in the study received a single static CT or MR image representing the maximum width of epidural tumor and did not have the ability to adjust the window and level settings or scroll through the entire image series. The goal of this study, however, was to determine reliability between raters; therefore, imaging comparisons were limited to a single static image to examine precision. This does not reflect clinical practice in which radiologists use a composite evaluation of a series of images. Another limitation is that measurement differences were within millimeter and submillimeter range. With such a small range, there may be a training effect in place as raters progressed through the series of images and became more acquainted with the exercise. To minimize this effect, the axial and sagittal projections from the different modalities were randomized and presented to the rater. Finally, the cost and availability of CTM and MRI are important factors to consider when incorporating imaging into clinical practice; these factors are, however, beyond the scope of our current analysis.
This is the first study to our knowledge that directly compares both the intra- and inter-observer reliability for measuring spinal epidural disease on different imaging modalities evaluated by neuroradiologists. The results of this study suggest that there is merit in continuing to use CTM in addition to MRI for evaluating the extent of spinal epidural disease and CTM can reliably be used as an alternative to MRI in patients with contraindications to MRI.
Highligts.
CT myelography should be used in addition to MRI to evaluate epidural disease and can be used in patients with contraindications to MRI
Inter-rater reliability for CT myelography is higher than that of MR for axial images
Inter-rater reliability for CT myelography and T1-MRI are similar for sagittal images
Funding sources:
This work was supported by the following grants from the National Institutes of Health (NIH): NIH-NCI P30 CA008748 (Thompson, PI), NIH-NCI R25CA020449 (Wolchok (PI)
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflicts of interest: None
References
- [1].Witham T, Khavkin YA, Gallia GL, Wolinsky JP, and Gokaslan ZL. Surgery insight: current management of epidural spinal cord compression from metastatic spine disease. Nat Clin Pract Neurol. 2006;2(2):87–94. doi: 10.1038/ncpneuro0116. [DOI] [PubMed] [Google Scholar]
- [2].O’Sullivan G, Carty F, and Cronin C. Imaging of bone metastasis: An update. World J Radiol. 2015;7(8):202–11. doi: 10.4329/wjr.v7.i8.202. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Shah LM and Salzman KL. Imaging of Spinal Metastatic Disease. Int J Surg Oncol. 2011;769753. doi: 10.1155/2011/769753. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Ozdoba C, Gralla J, Rieke A, Binggeli R, and Schroth G. Myelography in the age of MRI: Why we do it and how we do it. Radiology Research and Practice. 2011;329017. doi: 10.1155/2011/329017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Garibaldi C, Jereczek-Fossa BA, Marvaso G, et al. Recent advances in radiation oncology. Ecancermedicalscience. 2017;11:785. doi: 10.3332/ecancer.2017.785. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Miller GM and Krauss WE. Myelography: Still the gold standard. Am J Neuroradiol. 2003;24(3):298. [PMC free article] [PubMed] [Google Scholar]
- [7].Schmidt MA and Payne GS. Radiotherapy planning using MRI. Phys Med Biol. 2016; 60(22):R323–361. doi: 10.1088/0031-9155/60/22/R323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Song KJ, Choi BW, Kim GH, et al. Clinical usefulness of CT-myelogram comparing with the MRI in degenerative cervical spinal disorders: is CTM still useful for primary diagnostic tool? J Spinal Disord Tech. 2009;22(5):353–357. doi: 10.1097/BSD.0b013e31817df78e. [DOI] [PubMed] [Google Scholar]
- [9].Morita M, Miyauchi A, Okuda S, and Iwasaki M. Comparison between MRI and myelography in lumbar spinal canal stenosis for the decision of levels of decompression surgery. J Spinal Disord Tech. 2011;24(1):31–36. doi: 10.1097/BSD.0b013e3181d4c993. [DOI] [PubMed] [Google Scholar]
- [10].Bartynski WS and Lin L. Lumbar root compression in the lateral recess: MR imaging, conventional myelography, and CT myelography comparison with surgical confirmation. Am J Neuroradiol. 2003;24(3):348–60. [PMC free article] [PubMed] [Google Scholar]
- [11].Schneider CA, Rasband WS, and Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nature Methods. 2012;9:671–675. doi: 10.1038/nmeth.2089 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Lisanti C, Carlin C, Banks KP, and Wang D. Normal MRI appearance and motion-related phenomena of CSF. Am J of Roentgenology. 2007;188:716–25. doi: 10.2214/AJR.05.0003 [DOI] [PubMed] [Google Scholar]
- [13].Thwaites D Accuracy required and achieve in radiotherapy dosimetry: Have modern technology and techniques changed our views? Journal of Physics: Conference Series. 2013;444:12006. doi: 10.1088/1742-6596/444/1/012006 [DOI] [Google Scholar]