| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Musculoskeletal Imaging |
1 From the Boston Medical Center, Boston, Mass (D.C.Z.); Thomas Jefferson University Hospital, Philadelphia, Pa (W.B.M., J.A.P.); New York University, Hospital for Joint Diseases, New York, NY (M.E.S.); and Harvard Medical School, Brigham and Women's Hospital, 75 Francis St, Boston, MA 02115 (J.A.C.). From the 2002 RSNA Annual Meeting. Received June 3, 2004; revision requested August 11; revision received November 22; accepted December 27. Address correspondence to J.A.C. (e-mail: JCarrino{at}partners.org).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Institutional Review Board approval for this retrospective, HIPAA-compliant study was obtained; informed consent was waived. A total of 569 normal vertebrae in 75 patients (42 women, 33 men; mean age, 57.5 years; age range, 26–84 years) (control group) and 221 lesions in 92 patients (50 women, 42 men; mean age, 59.0 years; age range, 27–85 years) (study group) who had focal vertebral marrow abnormalities were studied by using 1.5-T chemical shift MR imaging. Imaging time was less than 1 minute. The proportional change in signal intensity on in-phase compared with out-of-phase images was calculated by using 1 x 1-cm regions of interest (ROIs) in the control group and ROIs as large as possible for focal lesions in the study group. This change in signal intensity (expressed as a percentage) was compared with that of normal levels and benign and malignant lesions. For statistical analysis, a random effect model was used that was adjusted for multiple comparisons.
RESULTS: A substantial decrease in signal intensity was noted for all normal vertebrae (mean, 58.5%) and for benign lesions, including endplate degeneration (mean, 52.2%), Schmorl nodes with edema (mean, 58.0%), hemangiomas (mean, 49.4%), and benign fractures (mean, 49.3%). Metastases exhibited either a minimal decrease or an increase in signal intensity (mean, 2.8%). Although there was some overlap in the range of signal intensity values among malignant lesions, benign lesions, and normal marrow, the differences in signal intensity loss for normal marrow and benign and malignant lesions were significant (P < .01 for all pairwise comparisons after adjusting for multiplicity).
CONCLUSION: Bone marrow in the vertebral bodies displays somewhat variable behavior at chemical shift MR imaging. Results suggest that a decrease in signal intensity greater than 20% on out-of-phase images compared with in-phase images should be used as a cutoff threshold for normalcy to allow distinction between benign and malignant causes of vertebral marrow abnormalities.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MR Imaging Techniques
All imaging was performed by using a 1.5-T imager (Signa; GE Medical Systems, Milwaukee, Wis) and a phased-array spine coil. The following pulse sequences were used for all patients: sagittal T1-weighted spin-echo (400–700/8–16 [repetition time msec/echo time msec]) MR imaging, sagittal fat-suppressed T2-weighted fast spin-echo (2000–5000/80–100, with an echo train length of 16) MR imaging, sagittal short inversion time inversion-recovery (STIR) fast spin-echo (2000–4000/20–40/150 [repetition time msec/echo time msec/inversion time msec]) MR imaging, and sagittal in-phase (100–165/4.2; flip angle, 30°) and out-of-phase (100–165/2.1; flip angle, 30°) fast multiplanar spoiled gradient-echo MR imaging. Sequences were performed with two to four signals acquired. For chemical shift MR imaging, the total imaging time was 40–50 seconds for the entire pulse sequence. Transverse plane pulse sequences were variable and were not used in this study. All other parameters were held constant for acquisition of in-phase and out-of-phase images. For all sagittal sequences, the field of view was 20 cm for cervical vertebrae, 34 cm for thoracic vertebrae, and 24 cm for lumbosacral vertebrae. The matrix was 256 x 192, and the section thickness and gap were 4.0 mm, with a skip of 1.0 mm.
Chemical Shift MR Imaging Measurements
By using a picture archiving and communication system workstation (Canon PACS; Canon, Lake Success, NY), two authors (D.C.Z., a radiology resident, and W.B.M., a radiologist with 11 years experience in musculoskeletal MR imaging) placed a rectangular region of interest (ROI) on each vertebral body or lesion. The signal intensity values for in-phase and out-of-phase images were then measured for each vertebral body or lesion at the same location on both images (vertebral body selection is described later). The proportional change in signal intensity was expressed as a percentage. Positive values were assigned to cases that demonstrated a decrease in signal intensity on out-of-phase images. Theoretically, the replacement of fat by means of a neoplastic process should result in a decreased loss of signal intensity on out-of-phase images compared with in-phase images, which would therefore result in a lower proportional score when classified according to the scoring system. Our internal reference standard was cerebrospinal fluid. In each patient, we performed an ROI analysis for cerebrospinal fluid, but the results did not vary substantially from in-phase to out-of-phase images. Because the difference in cerebrospinal fluid values was approximately zero in all cases, no correction factor was applied to our calculations. Use of a dual gradient technique, however, obviates internal standards.
Control Group: Normative Data
The control group consisted of 75 consecutive patients (42 women, 33 men; mean age, 57.5 years; age range, 26–84 years; mean weight, 171.1 lb [77.6 kg]; weight range, 121–280 lb [55–127 kg]) who met the inclusion criteria. MR imaging of the spine included examination of the cervical (n = 20), thoracic (n = 7), or lumbosacral (n = 48) vertebrae. Clinical indications for the examinations included back pain, radicular symptoms, and sciatica; examinations were also performed to rule out neoplasm. Additional exclusion criteria for the control group were intervertebral body fusion at a specific level, exclusion of a level from the field of view with all necessary pulse sequences, history of malignancy, or presence of signal intensity abnormalities in vertebral marrow on T1-weighted, T2-weighted, or STIR MR images. All patients who demonstrated a focal abnormality were assigned to the study group. A total of 569 vertebrae were studied, including cervical levels C2 through C7 (n = 113), thoracic levels T1 through T12 (n = 170), and lumbosacral levels L1 through S1 (n = 286). It was assumed that all patients had seven cervical vertebrae, 12 thoracic vertebrae, and five lumbosacral vertebrae to simplify counting and analyses. Signal intensity values for all vertebral bodies that demonstrated normal findings on T1-weighted, T2-weighted, and STIR images were calculated within a 1 x 1-cm ROI that was matched on in-phase and out-of-phase images; for each ROI, signal intensity values were obtained in the central portion of each vertebral body by using a workstation (Canon PACS; Canon).
Study Group: Pathologic Analysis
The study group consisted of 92 consecutive patients (50 women, 42 men; mean age, 59.0 years; age range, 27–85 years) who demonstrated focal findings at MR imaging of the spine. Lesions consisted of endplate degeneration (40 patients), hemangioma (21 patients), Schmorl node (16 patients), benign (non–neoplasm-related) fracture (five patients), and metastatic neoplasm (10 patients). Criterion standards for each lesion were established according to characteristic MR imaging appearance, pathologic correlation, or imaging follow-up. Imaging criterion standards were determined by the consensus decision of two or three authors, each with substantial experience in interpreting MR images of the spine (J.A.C., W.B.M., M.E.S., and J.A.P., with 11, 11, 15, and 4 years experience, respectively). The type of degenerative endplate marrow abnormality was designated by using the classification criteria established by Modic et al (4,5) for simplicity of reference. Type 1 findings refer to abnormalities with edema-like signal intensity (ie, low signal intensity on T1-weighted images and high signal intensity on T2-weighted or STIR images), type 2 findings refer to abnormalities with fatlike signal intensity (ie, high signal intensity on T1-weighted images), and type 3 findings refer to abnormalities with sclerosis-like signal intensity (ie, low signal intensity on T1-weighted, T2-weighted, and STIR images).
To be considered a reactive endplate finding, the adjacent intervertebral disk needed to show evidence of degeneration (ie, loss of height, desiccation, or contour abnormality [bulge or herniation]). Typical hemangioma was defined as a round lesion in the vertebra, with intermediate to high signal intensity on T1-weighted images and high signal intensity on T2-weighted or STIR images (6,7). A Schmorl node was defined as a focal, rounded lesion within the endplate of a vertebral body that was contiguous with an intervertebral disk demonstrating intraosseous disk herniation (8,9). Lesions that were associated with marrow edema were also classified as Schmorl nodes (10). Fractures were defined as areas of marrow edema located within a deformed vertebral body (wedge or compression) (11,12). Another characteristic of fractures was the presence of a horizontal, linear region of low signal intensity (fracture line) located within the vertebral body; this region of low signal intensity represented the fracture line; this finding, however, did not need to be present in all cases (13). Diagnosis was established either at biopsy or at imaging follow-up if the results showed resolution of marrow edema and reconstitution of normal signal intensity. All eight vertebral fractures in this study were benign; five fractures were documented as benign at MR imaging follow-up and three were documented as benign at biopsy.
Metastastic neoplasm was defined as a lesion with low signal intensity on T1-weighted images and high signal intensity on T2-weighted or STIR images that did not meet the criteria of any of the previously described lesions (14). Diagnosis was established by means of histologic documentation of the tumor at biopsy or, in cases of multiple similar lesions, by means of histologic proof of tumor within a second lesion. In our study, each patient underwent a single biopsy, and no biopsy results showed a lack of tumor. We interpreted multiple levels of similar-appearing abnormalities in a single patient on the basis of biopsy results at a single level. For each malignant lesion, available radiographs and computed tomographic (CT) scans were reviewed, and lesions were further characterized as lytic, blastic, or mixed (lytic and blastic). Imaging review was performed by two or three authors in consensus (J.A.C., W.B.M., M.E.S., and J.A.P, with 11, 11, 15, and 4 years experience, respectively). Investigators were blinded to the final diagnosis. For patients with malignancy, a retrospective chart review was performed (D.C.Z.) to examine history and biopsy results, which were used to document the presence and type of underlying malignancy. Signal intensity values for rectangular, matched ROIs, which were as large as possible for both in-phase and out-of-phase images, were obtained within the central portion of the lesion by using a workstation (Canon PACS; Canon).
Statistical Analysis
The proportional change (percent decrease) of marrow signal intensity on out-of-phase images compared with in-phase images was calculated for each vertebra (control group) or lesion (study group). The range and mean change in signal intensity were calculated and graphed according to vertebrae or lesion type. Exploratory graphic analyses did not reveal substantial deviations from normal distribution. A comparison of signal intensity change between malignant lesions and normal vertebrae, pairwise comparisons of signal intensity change between lesion types, and metastasis subgroup analysis were performed by using an analysis of variance model to assess for significant differences among levels. This model included patient random effects to account for the fact that there were multiple observations per patient (15). Because several pairwise comparisons were employed (eg, normal vs benign, normal vs malignant, and benign vs malignant), we used the Tukey adjustment to control the family-wise type I error rate at 5% (16). Signal intensity analysis was performed by two authors (D.C.Z., W.B.M.), and statistical analysis was performed by two statisticians, one of whom was an author (J.A.C.). For statistical analysis, two computer software programs were used (StatCrunch, version 3.0; www.statcrunch.com and SAS PROC MIXED; SAS Institute, Cary, NC). A P value of less than .05 was considered to indicate a statistically significant difference.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
Summary values for signal intensity changes in all lesions are graphically represented in Figure 2. For degenerative endplate findings, the mean decrease in signal intensity was 52.2% ± 16.7 for all types. Subgroup analysis for degenerative endplate findings (Fig 3) showed a mean decrease in signal intensity of 53.1% for type 1 findings, 48.5% for type 2 findings, and 60.4% for type 3 findings. For type 2 changes, a bimodal distribution was observed with separate peaks at about 37% and 63%. Marrow edema that was adjacent to a Schmorl node showed a mean decrease in signal intensity of 58.0% ± 17.2. Hemangiomas showed a mean decrease in signal intensity of 49.4% ± 20.0. Benign fractures with associated marrow edema showed a mean decrease in signal intensity of 49.3% ± 18.4. For metastatic lesions, the mean decrease in signal intensity was 2.8% ± 21.3. Subgroup analysis (Fig 4) showed a mean decrease in signal intensity of 0.5% for lytic lesions, 2.4% for blastic lesions, –3.0% for mixed lesions, and 8.7% for unknown lesions. MR images of metastatic lesions can be seen in Figure 5. Pairwise comparison showed significant differences between malignant lesions and normal vertebrae (P < .01) and between malignant lesions and benign lesions (P < .01 for each comparison) but not between benign lesions and other benign lesions—that is, between endplate degeneration and Schmorl nodes (P = .20), endplate degeneration and hemangiomas (P = .47), endplate degeneration and fractures (P = .48), Schmorl nodes and hemangiomas (P = .11), Schmorl nodes and fractures (P = .17), and hemangiomas and fractures (P = .78). There was no significant variability among the four metastases subtypes (P = .52). Pairwise comparisons that were performed by using the Tukey method also showed no significant differences between any combinations of metastases subtypes, including unknown versus lytic lesions (P = .22), unknown versus blastic lesions (P = .34), unknown versus mixed lesions (P = .19), lytic versus blastic lesions (P = .81), lytic versus mixed lesions (P = .96), and blastic versus mixed lesions (P = .76).
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The use of chemical shift MR imaging in bone marrow to distinguish benign from malignant processes has been reported in the axial skeleton at midfield (ie, at 0.5 T). Zampa et al (19) evaluated 86 lesions at 0.5 T by using an MR protocol that consisted of a T1-weighted spin-echo and an out-of-phase gradient-recalled echo MR imaging sequence, both of which used the same parameters except for echo time (10 msec for spin-echo sequence and 7 msec for gradient-recalled echo sequence) to obtain out-of-phase images. Quantitative analysis consisted of a signal intensity ratio that was expressed by comparing the signal intensity of the lesion on out-of-phase images with the signal intensity of the lesion on conventional T1-weighted images (ie, signal intensity on out-of-phase gradient-recalled echo MR images divided by signal intensity on T1-weighted spin-echo MR images). A cutoff value of 1.2 resulted in 88.8% sensitivity, 80.5% specificity, 84.9% accuracy, 86.4% negative predictive value, and 83.3% positive predictive value. Lesions with values that were higher than the cutoff value were considered neoplastic, whereas lesions with values that were lower than the cutoff value were considered benign. The parameters for chemical shift MR imaging are dependent on field strength (ie, a longer echo time is needed at lower field strengths), and the results may not be directly extrapolated from one field strength to another.
Nonetheless, the cutoff value obtained by Zampa et al (19) was similar to the cutoff value obtained in our study (ie, 20% decrease in signal intensity). Our value, however, was calculated as a percentage, which is a different method to express the difference in signal intensity between in-phase and out-of-phase images. Qualitative assessment has been performed at high field strength (1.5 T) in the context of vertebral compression fractures, but no quantitative recommendations were made (12). In general, images that demonstrate fat in acute benign compression fractures show only partial replacement of normal fatty marrow by areas of low signal intensity, which is in contrast to the complete absence of marrow signal intensity that is typical of pathologic, neoplastic fractures.
There was some variability in the loss of signal intensity according to location (spinal segment), but there were no associations between age or weight. Age-related change might be expected because hematopoietic marrow generally decreases with age (20). This might be explained by a selection bias because we did not control for other factors, such as anemia, smoking history, osteoporosis, or body mass index, which are associated with a decreased amount of hematopoietic marrow. Chemical shift MR imaging can demonstrate the relationship between the amount of fat and water that coexist in the same voxel. Osseous elements, however, will also affect this relationship; thus, the degree of signal intensity change may not be proportional to the quantity of hematopoietic marrow alone.
Hemangiomas are slow growing, benign neoplasms that are commonly found in the vertebral bodies. Histopathologically, they consist of thin-walled, blood-filled vessels and sinuses that are lined by endothelium, are interspersed among the bone trabeculae, and have a variable amount of fat. Some, such as those with predominant fat content, do not demonstrate a decrease in signal intensity because there are few or no nonlipid elements. These hemangiomas, however, are easily characterized at standard T1-weighted MR imaging and do not pose a diagnostic dilemma. Atypical hemangiomas, which contain only small or microscopic quantities of fat, may demonstrate the utility of chemical shift MR imaging because they will lose signal intensity on out-of-phase images; such lesions may have otherwise been difficult to distinguish from malignant neoplastic lesions.
Changes in signal intensity for endplate degeneration are also typically and sufficiently characterized at routine MR imaging. In type 1 (edema-like) degenerative endplate lesions, some fat is preserved, which explains the loss of signal intensity on out-of-phase images. Type 2 (fatlike) degenerative endplate lesions showed a bimodal peak in signal intensity loss, which was likely because a decrease in signal intensity is sometimes characterized by complete fat replacement of the marrow (no myeloid elements). Lesions that are composed of virtually 100% fat would not be expected to decrease in signal intensity, which is the most likely reason for the bimodal peak. Type 3 (sclerosis-like) degenerative endplate lesions consistently showed decreases in signal intensity that were greater than 20%, but this is usually not problematic for diagnosis at MR imaging. The behavior of type 3 lesions may be related to susceptibility-induced signal intensity loss, which is greater on out-of-phase images, or proton density. Edema-like signal intensity is a generic term that refers to fluidlike hyperintensity on T2-weighted images and can result from many causes (eg, degenerative, infectious, inflammatory, traumatic, and neoplastic changes). Acute and subacute benign fractures have marrow edema, but the underlying marrow should have at least microscopic fat that appears obscured on standard T1-weighted and fluid-sensitive images. Our results are consistent with this presumption.
As a group, neoplastic malignant lesions showed a significantly lower percentage of signal intensity loss compared with all other groups. Nevertheless, eight (16%) of 51 lesions demonstrated a decrease in signal intensity that was greater than 20%. Of the lesions that demonstrated false-negative findings at chemical shift MR imaging, four had a decrease in signal intensity that was greater than 40%, with the greatest decrease in signal intensity being 75.9%. The reason for this overlap is in part related to the lesion subtype. Lytic lesions lost less signal intensity than blastic lesions. The substantial number of metastases that could not be characterized (unknown lesions) also contributed to the variable behavior of malignant lesions at chemical shift MR imaging. More work is needed to determine the characteristics of malignant lesions that may demonstrate a loss in signal intensity that is similar to that of benign processes. In the context of suspected metastasis with indeterminate behavior at chemical shift MR imaging, however, CT is a complementary technique that can reveal a sclerotic focus.
There are several limitations to this investigation. The sample size for the normal thoracic spine examinations was small, and more values are needed to establish a robust normal range. The sample size for some lesions was small and did not include malignant vertebral fractures. We did not compare chemical shift MR imaging to traditional methods or define the incremental value as an adjunct for diagnosis, but rather, we evaluated the merit of using chemical shift MR imaging as a stand-alone technique. We also did not compare our findings with those of other novel techniques, such as diffusion-weighted MR imaging. Traditionally, the diffusion-weighted MR imaging sequence takes a substantial amount of imaging time (6–7 minutes), but now there are more rapid acquisition techniques, such as line-scan diffusion-weighted imaging (21), which makes this a competitive pulse sequence.
Another potential study design weakness is that not all malignant lesions had direct pathologic correlation, and only one representative lesion was used per patient. Also, the sample size of 10 patients with malignant lesions, despite analysis of 51 lesions within this cohort, is somewhat small.
In conclusion, the results of our study show that normal marrow and benign lesions have characteristic behavior at chemical shift MR imaging (ie, a consistent loss of signal intensity), while metastatic lesions tend to have different behavior at chemical shift MR imaging (ie, slight [if any] loss of signal intensity). One inference that can be made from the results of this study is that, on chemical shift MR images, there is a threshold cutoff value for the decrease in signal intensity. We put forth that any lesion that is not composed of fat on T1-weighted images and that loses less than 20% of its signal intensity on out-of-phase images compared with in-phase images should be considered suspicious for malignancy. This value needs to be validated in a larger cohort with receiver operating characteristic curve analyses, but we believe this result is reasonable because virtually no benign lesion or normal marrow showed this behavior. Of note, extreme outliers were observed, particularly in the control group. The importance of this finding is that the distribution of the study population has more extreme cases than a normal distribution. Typically, it is believed that outliers represent a random error that one would like to be able to control. Our strategy was not to exclude outliers but to propose a range of normalcy on the basis of 2 standard deviations from all normal levels (mean, 58.5% ± 15.9). Alternatively, these outliers may be indicative of the occurrence of a phenomenon that is qualitatively different from the typical pattern observed or expected in the sample, but this is yet to be determined.
Results of this investigation provide baseline data to aid in the interpretation of clinical chemical shift MR images of vertebral bone marrow. This fast and widely available technique should improve the accuracy of MR imaging in facilitating discrimination between benign and malignant vertebral marrow lesions.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: ROI = region of interest • STIR = short inversion time inversion-recovery
See Materials and Methods for pertinent disclosures.
Author contributions: Guarantors of integrity of entire study, W.B.M., J.A.C.; study concepts, W.B.M., J.A.C.; study design, W.B.M.; literature research, D.C.Z., J.A.C.; clinical studies, D.C.Z., J.A.P., M.E.S.; data acquisition, D.C.Z., M.E.S., J.A.P., W.B.M.; data analysis/interpretation, J.A.C., W.B.M.; statistical analysis, J.A.C.; manuscript preparation, D.C.Z., J.A.C.; manuscript definition of intellectual content, J.A.C.; manuscript editing, D.C.Z.; manuscript revision/review, J.A.C.; manuscript final version approval, D.C.Z.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. B. Siepmann, J. McGovern, J. H. Brittain, and S. B. Reeder High-Resolution 3D Cartilage Imaging with IDEAL SPGR at 3 T Am. J. Roentgenol., December 1, 2007; 189(6): 1510 - 1515. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. M. Gerdes, R. Kijowski, and S. B. Reeder IDEAL Imaging of the Musculoskeletal System: Robust Water Fat Separation for Uniform Fat Suppression, Marrow Evaluation, and Cartilage Imaging Am. J. Roentgenol., November 1, 2007; 189(5): W284 - W291. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. S. Steinbach "MRI in the Detection of Malignant Infiltration of Bone Marrow"--A Commentary Am. J. Roentgenol., June 1, 2007; 188(6): 1443 - 1445. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
