Original Research
Musculoskeletal Imaging
November 23, 2012

Comparison of Qualitative and Quantitative Evaluation of Diffusion-Weighted MRI and Chemical-Shift Imaging in the Differentiation of Benign and Malignant Vertebral Body Fractures

Abstract

OBJECTIVE. The objective of our study was to compare the diagnostic value of qualitative diffusion-weighted imaging (DWI), quantitative DWI, and chemical-shift imaging in a single prospective cohort of patients with acute osteoporotic and malignant vertebral fractures.
SUBJECTS AND METHODS. The study group was composed of patients with 26 osteoporotic vertebral fractures (18 women, eight men; mean age, 69 years; age range, 31 years 6 months to 86 years 2 months) and 20 malignant vertebral fractures (nine women, 11 men; mean age, 63.4 years; age range, 24 years 8 months to 86 years 4 months). T1-weighted, STIR, and T2-weighted sequences were acquired at 1.5 T. A DW reverse fast imaging with steady-state free precession (PSIF) sequence at different delta values was evaluated qualitatively. A DW echo-planar imaging (EPI) sequence and a DW single-shot turbo spin-echo (TSE) sequence at different b values were evaluated qualitatively and quantitatively using the apparent diffusion coefficient. Opposed-phase sequences were used to assess signal intensity qualitatively. The signal loss between in- and opposed-phase images was determined quantitatively. Two-tailed Fisher exact test, Mann-Whitney test, and receiver operating characteristic analysis were performed. Sensitivities, specificities, and accuracies were determined.
RESULTS. Qualitative DW-PSIF imaging (delta = 3 ms) showed the best performance for distinguishing between benign and malignant fractures (sensitivity, 100%; specificity, 88.5%; accuracy, 93.5%). Qualitative DW-EPI (b = 50 s/mm2 [p = 1.00]; b = 250 s/mm2 [p = 0.50]) and DW single-shot TSE imaging (b = 100 s/mm2 [p = 1.00]; b = 250 s/mm2 [p = 0.18]; b = 400 s/mm2 [p = 0.18]; b = 600 s/mm2 [p = 0.39]) did not indicate significant differences between benign and malignant fractures. DW-EPI using a b value of 500 s/mm2 (p = 0.01) indicated significant differences between benign and malignant vertebral fractures. Quantitative DW-EPI (p = 0.09) and qualitative opposed-phase imaging (p = 0.06) did not exhibit significant differences, quantitative DW single-shot TSE imaging (p = 0.002) and quantitative chemical-shift imaging (p = 0.01) showed significant differences between benign and malignant fractures.
CONCLUSION. The DW-PSIF sequence (delta = 3 ms) had the highest accuracy in differentiating benign from malignant vertebral fractures. Quantitative chemical-shift imaging and quantitative DW single-shot TSE imaging had a lower accuracy than DW-PSIF imaging because of a large overlap. Qualitative assessment of opposed-phase, DW-EPI, and DW single-shot TSE sequences and quantitative assessment of the DW-EPI sequence were not suitable for distinguishing between benign and malignant vertebral fractures.
Benign osteoporotic compression fractures are common in elderly patients and also occur in one third of cancer patients [1]. A T1- or T2-weighted sequence and a STIR sequence are very sensitive but lack specificity in the differentiation of benign from malignant vertebral compression fractures because bone marrow edema in acute osteoporotic fractures may mimic the signal alterations observed in osseous metastases [2-4]. In previous studies, investigators evaluated the qualitative assessment of signal intensities (SIs) on diffusion-weighted (DW) steady-state free-precession (SSFP), echo-planar imaging (EPI), single-shot turbo spin-echo (TSE), and opposed-phase sequences as well as the quantitative measurement of apparent diffusion coefficients (ADCs) and the value of quantitative chemical-shift imaging to improve specificity [5-24]. Because there have been partially contradicting statements about the value of these techniques and there is still uncertainty whether quantification is really necessary and how it improves specificity, there was a need to compare these methods in one single patient cohort.
Thus, the purpose of this study was to compare the diagnostic value of qualitative diffusion-weighted imaging (DWI), quantitative DWI, and chemical-shift imaging in the differentiation of benign and malignant vertebral fractures in a prospective collective of patients.

Subjects and Methods

Patients

After internal review board approval (ethics commission) and informed consent had been obtained, the protocol was applied to 46 patients (27 women and 19 men; mean age, 66.6 years; age range, 24 years 8 months to 86 years 4 months) who presented with acute vertebral collapse and were prospectively examined at our hospital between April 2008 and November 2010.
Participation in this study was voluntary. The consecutive patients were admitted from either the acute day ward or the orthopedics outpatient department. The additional sequences required for this study were added to the routine MRI protocol for patients with acute vertebral collapse.
The inclusion criteria were age of more than 18 years, an acute or subacute clinical presentation with back pain at the spinal level of fracture (< 3 months), and bone marrow edema at the fracture sites.
The exclusion criteria were pregnancy, contraindications to MRI (e.g., cardiac pacemaker) or gadolinium-containing contrast agents (e.g., glomerular filtration rate < 30 mL/min), diffuse hematologic disorders, and mental incapacity.
One representative lesion, defined as the lesion with the most marked signal change on STIR, was chosen in patients with more than one vertebral fracture.
Patients were divided into two groups according to the cause of the vertebral fracture, which was determined by the gold standard histology (n = 24), follow-up MRI (n = 11), PET/CT (n = 3), or clinical follow-up including CT more than 6 months after presentation (n = 8).
Group 1 consisted of 26 patients with osteoporotic vertebral collapse (18 women and eight men; mean age, 69 years; age range, 31 years 6 months to 86 years 2 months). The presence of a malignant fracture in these patients was ruled out by histology (n = 7); follow-up MRI with disappearance of edema and exclusion of any morphologic signs of malignancy in the fractured vertebral body in combination with clinical follow-up (disappearance of pain and no suspicion of osseous metastases) (n = 11); and CT more than 6 months after presentation, showing no morphologic signs of malignancy in combination with additional clinical follow-up (disappearance of pain and no suspicion of osseous metastases) (n = 8).
Group 2 consisted of 20 patients with malignant vertebral collapse (nine women and 11 men; mean age, 63.4 years; age range, 24 years 8 months to 86 years 4 months). Primary neoplasms included lung cancer (n = 1), adenocarcinoma (n = 3), bladder cancer (n = 1), nonseminoma (n = 1), hypopharyngeal cancer (n = 1), breast cancer (n = 5), multiple myeloma (n = 6), thyroid carcinoma (n = 1), and renal cell carcinoma (n = 1). The diagnoses were confirmed by histopathologic examination of specimens obtained during surgery (n = 4), CT-guided biopsy (n = 13), or PET/CT finding of a definite pathologic standardized uptake value 7-20 months after presentation (n = 3).

MRI

Measurements were performed on a 32-channel 1.5-T whole-body scanner (Magnetom Avanto, Siemens Healthcare). For signal reception, a quadrature spine surface coil was used. Before diffusion measurements, T1-weighted (TR/TE, 531/12), STIR (TR/TE, 3790/61; inversion time, 180 ms), and T2-weighted (TR/TE, 4420/118) TSE images of 21 sagittal slices with a slice thickness of 3 mm were acquired using a 44 × 44 cm2 FOV and a matrix size of 384 × 384. The total acquisition time for these morphologic images was 16 minutes 30 seconds.
Diffusion-weighted steady-state free-precession imaging—A DWI sequence based on reverse fast imaging with steady-state free precession (PSIF) was performed with a matrix size of 256 × 192 and a bandwidth of 100 Hz/pixel (TR/TE, 25/7.17; flip angle, 40°). The diffusion gradient strength was kept constant at 23 mT/m, whereas the duration of the diffusion gradient was varied (δ = 3.0, 4.5, and 6.0 ms). The slice thickness was 5 mm; 10 averages were used to calculate one image. The total acquisition time was 2 minutes 35 seconds.
In- and opposed-phase imaging—A sagittal gradient-echo in-phase sequence (TR/TE, 120/4.76; flip angle, 70°) and an opposed-phase sequence (TR/TE, 120/2.38; flip angle, 70°) with a slice thickness of 5 mm, a matrix size of 320 × 240, and a bandwidth of 280 Hz/pixel were performed. Two averages were used to calculate one image. The total acquisition time was 0:40 seconds.
Diffusion-weighted echo-planar imaging—A fat-saturated DW-EPI sequence (TR/TE, 3000/87; b = 50, 250, 500 s/mm2) was performed with a matrix size of 192 × 144 and a bandwidth of 965 Hz/pixel. The diffusion signals were averaged over five repeated acquisitions and three orthogonal gradient directions. In both measurements, the b value was varied by changing the amplitude of the diffusion gradient to keep the gradient duration constant. The total acquisition time was 2 minutes 23 seconds.
Diffusion-weighted single-shot turbo spin-echo imaging—A fat-saturated DW single-shot TSE sequence with four b values (b = 100, 250, 400, 600 s/mm2) was applied. The imaging parameters were a 128 × 92 matrix, TR/TE of 3000/72, flip angle of 180° for the refocusing pulses, and bandwidth of 735 Hz/pixel. The diffusion weightings were applied in a diagonal direction (gradients in all three dimensions were applied simultaneously). Because of the low signal of bone marrow on DWI, 10 averages were taken to improve the signal-to-noise ratio. The process of averaging was performed on magnitude data to avoid image artifacts due to motion-induced phase variations. The total acquisition time was 2 minutes 13 seconds.

Postprocessing

The data were postprocessed offline on a PC using in-house-built software (PMI 0.4-Platform for Research in Medical Imaging [25]) written in IDL (version 6.4, Exelis Visual Information Solutions). All images of each patient were interpreted in consensus by two radiologists who had more than 10 and 6 years of experience in musculoskeletal imaging, respectively, at the time of the study.
The SIs of the fractured vertebral bodies were qualitatively evaluated on opposed-phase images and each of the DW-PSIF (delta = 3, 4.5, and 6 ms), DW-EPI (b = 50, 250, and 500 s/mm2), and DW single-shot TSE (b = 100, 250, 400, and 600 s/mm2) images and were described as hypointense, isointense, or hyperintense in relation to the surrounding normal-appearing vertebral bone marrow. Standard diagnostic displays calibrated to the DICOM standard were used for image interpretation.
Regions of interest (ROIs) for in- and opposed-phase quantification were selected manually in the lesions using the STIR and T1-weighted sequences. The size of each ROI was exactly adapted to the area of hyperintense signal on the STIR images and to the area of hypointense signal on T1-weighted images. The ROIs were exactly copied to the in- and opposed-phase and DW images and were corrected for image distortions, if necessary (Fig. 1D). SIs for varying b values were fitted to a monoexponential decay model using a least-squares algorithm to determine the ADCs [26].

Statistical Evaluation

Hyperintense signal seen on DW images and opposed-phase images was assumed to indicate malignancy. Isointense or hypointense signal was assumed to indicate benign edema, as described in the literature [5, 15, 17, 27].
Assuming that the majority of all malignant lesions would show a diffusion restriction on PSIF imaging (probability 1: 0.9) and that the degree of edema in osteoporotic fractures is associated with a lower proportion of diffusion restriction (probability 2: 0.4 [28], a sample size of 17 subjects per group would provide sufficient power (β = 0.9) to show a difference between the two groups with an alpha value of 0.05. All other comparisons were of exploratory nature and were used to generate a hypothesis.
Fig. 1A 59-year-old woman with osteoporotic fracture of L1 vertebra.
A, T1-weighted (A) and STIR (B) images.
Fig. 1B 59-year-old woman with osteoporotic fracture of L1 vertebra.
B, T1-weighted (A) and STIR (B) images.
Fig. 1C 59-year-old woman with osteoporotic fracture of L1 vertebra.
C, Diffusion-weighted (DW) steady-state free-precession image shows isointense signal at delta value of 3 ms.
Fig. 1D 59-year-old woman with osteoporotic fracture of L1 vertebra.
D, DW echo-planar image obtained using b value of 50 s/mm2 shows hyperintensity. Region of interest for quantitative measurement is highlighted in red.
Fig. 1E 59-year-old woman with osteoporotic fracture of L1 vertebra.
E, In-phase image shows hypointense signal.
Fig. 1F 59-year-old woman with osteoporotic fracture of L1 vertebra.
F, Opposed-phase image shows hyperintense signal.
Fig. 1G 59-year-old woman with osteoporotic fracture of L1 vertebra.
G, DW single-shot turbo spin-echo (TSE) image shows hyperintense signal at b value of 100 s/mm2.
Fig. 1H 59-year-old woman with osteoporotic fracture of L1 vertebra.
H, Apparent diffusion coefficient (ADC) is 1.25 × 10-3 mm2/s on DW echo-planar image (H) and 1.73 × 10-3 mm2/s on DW single-shot TSE image (I).
Fig. 1I 59-year-old woman with osteoporotic fracture of L1 vertebra.
I, Apparent diffusion coefficient (ADC) is 1.25 × 10-3 mm2/s on DW echo-planar image (H) and 1.73 × 10-3 mm2/s on DW single-shot TSE image (I).
Cross-classified tables and a two-tailed Fisher exact test were used to show significant differences in the relative SIs between the two independent groups of patients with benign and those with malignant vertebral fractures. Sensitivities, specificities, and diagnostic accuracies with 95% CIs were calculated accordingly.
Mean values, median values, SDs, minimum and maximum values and 95% CIs of the quantitative parameters in the two groups with benign and malignant vertebral fractures were determined to compare the quantitative parameters in lesions of both independent patient groups, and an unpaired two-tailed Student t test was performed. If distribution was not normal, a Mann-Whitney U test was performed. The sensitivities and specificities of the quantitative parameters for the detection of malignant fractures to be differentiated from osteoporotic fractures were calculated with receiver operating characteristic (ROC) analysis. As a cutoff value, the parameter yielding the highest accuracy was chosen. All analyses were performed using SPSS software (version 12, SPSS), and a p value of < 0.05 was defined to indicate statistical significance.

Results

An overview of the qualitative evaluation of the SIs on the DW images and opposed-phase images is given in Table 1. The sensitivities, specificities, and accuracies and the comparison of the qualitative parameter with the gold standard are summarized in Table 2. Examples of an osteoporotic fracture and a neoplastic fracture are shown in Figures 1A, 1B, 1C, 1D, 1E, 1F, 1G, 1H, 1I, 2A, 2B, 2C, 2D, 2E, 2F, 2G, 2H, and 2I, respectively.
The qualitative assessment of the opposed-phase images showed a high sensitivity of 95% but a very low specificity of 30.8% (accuracy, 58.7%) with most lesions being hyperintense. Eighteen osteoporotic fractures were associated with a hyperintense SI (false-positives). The Fisher exact test did not show significant differences between benign and malignant vertebral fractures (p = 0.06). Thus, the qualitative assessment of opposed-phase images was not useful in the differentiation of benign and malignant vertebral fractures.
The qualitative assessment of the relative SI in the PSIF sequence with a delta value of 3 ms showed the highest accuracy (93.5%) of all qualitative and quantitative parameters in our study with a sensitivity of 100% and specificity of 88.5%. At a delta value of 4.5 ms, the second highest accuracy (91.3%) of all examined parameters was found (sensitivity, 85%; specificity, 96.2%). At a delta value of 6 ms, the diagnostic accuracy was lower (71.7%; sensitivity, 35%; specificity, 100%). PSIF imaging exhibited significant differences between benign and malignant vertebral fractures at all used delta values (Fisher exact test: p < 0.0001 at delta values of 3 and 4.5 ms; p = 0.01 at a delta value of 6 ms).
TABLE 1: Signal Intensities (SIs) of Lesions on Chemical-Shift, Reverse Fast Imaging With Steady-State Free Precession (PSIF), Echo-Planar Imaging (EPI), and Single-Shot Turbo Spin-Echo (TSE) Diffusion-Weighted (DW) Imaging by Fracture Type
TABLE 2: Sensitivities, Specificities, Accuracies, and Statistical Differences of the Qualitative Parameters (Signal Intensities) on Chemical-Shift, Reverse Fast Imaging With Steady-State Free Precession (PSIF), Echo-Planar Imaging (EPI), and Single-Shot Turbo Spin-Echo (TSE) Diffusion-Weighted (DW) Imaging With Respect to the Diagnosis of a Malignant Lesion
The qualitative assessment of SIs on DW-EPI sequences at b values of 50, 250, and 500 s/mm2 was sensitive (85-100%) but not specific (0-53.8%) in the differentiation of benign from malignant vertebral fractures. The diagnostic accuracy was 43.5-67.4%. Edema and tumor showed hyperintense signal in most cases. The distribution of the SIs on the EPI sequences at b values of 50 s/mm2 (p = 1.00) and 250 s/mm2 (p = 0.50) was not statistically significant between malignant fractures and osteoporotic fractures. At a b value of 500 s/mm2, relative SIs revealed significant differences between both fracture types (Fisher exact test, p = 0.01).
The qualitative assessment of SIs on DW single-shot TSE sequences at b values of 100 s/mm2 (p = 1.00), 250 s/mm2 (p = 0.18), 400 s/mm2 (p = 0.18), and 600 s/mm2 (p = 0.39) could not distinguish between benign and malignant fractures with random associations of the relative SIs between benign and malignant vertebral fractures in the Fisher exact test.
The quantitative results are summarized in the boxplots shown in Figures 3A, 3B, and 3C; the corresponding mean values, median values, and SDs as well as statistically significant differences between osteoporotic and malignant vertebral fractures are shown in Table 3. Table 4 summarizes the sensitivities and specificities at the point of highest accuracy determined with ROC analysis in the differentiation of osteoporotic and malignant vertebral fractures.
The signal loss in opposed- versus in-phase images was significant (Mann-Whitney U test, p = 0.01) between malignant (mean, -8.5%; median, -3.0%; SD, 16.4%) and osteoporotic (mean, -24.8%; median, -17.5%; SD, 25.7%) fractures. The area under the ROC curve (AUC) was 0.725; the highest accuracy (71.7%) was determined at a cutoff of ≥ -1.44% for malignancy with a sensitivity of 50% and a specificity of 88.5%. However, the overlap (as seen in Figs. 3A, 3B, and 3C) is large and there is no definitive cutoff value that would exclude false-negative results. Osteoporotic fractures also can show a minor signal loss on opposed-phase images. At a cutoff of ≥ -47% signal loss, there were no more false-negative results. There were always false-positives because of a large overlap.
There was no significant difference (unpaired Student t test, p = 0.09) of the ADC in the DW-EPI sequence between malignant (mean, 1.05 × 10-3 mm2/s; median, 1.03 × 10-3 mm2/s; SD, 0.26 × 10-3 mm2/s) and benign (mean, 1.23 × 10-3 mm2/s; median, 1.29 × 10-3 mm2/s; SD, 0.37 × 10-3 mm2/s) fractures. The AUC was 0.690; the highest accuracy (69.6%) was determined at a cutoff of ≤ 1.18 × 10-3 mm2/s for malignancy with a sensitivity of 75% and a specificity of 65.4%.
The ADC values in the DW single-shot TSE sequence differed significantly (p = 0.002) between malignant (mean, 1.31 × 10-3 mm2/s; median, 1.32 × 10-3 mm2/s; SD, 0.36 × 10-3 mm2/s) and benign (mean, 1.64 × 10-3 mm2/s; median, 1.68 × 10-3 mm2/s; SD, 0.31 × 10-3 mm2/s) fractures. The AUC was 0.781; the highest accuracy (78.3%) was determined at a cutoff of ≤ 1.48 × 10-3 mm2/s for malignancy with a sensitivity of 75% and a specificity of 80.8%.

Discussion

Vertebral bone marrow generally consists of yellow (mainly fat) and red (similar amounts of fat and water) marrow, which dominate the signal characteristics in MRI. The bone marrow of adults consists of approximately 50% fat and 50% water. However, the amount of fat varies among different age groups: There is an increasing amount of fat with increasing age [29]. In osteoporotic patients, the loss in the trabecular osseous network is also usually replaced by fat cells. In malignant vertebral fractures, the bone marrow is usually replaced by tumor cells. In osteoporotic vertebral fractures, on the other hand, bone marrow is infiltrated by a variable amount of edema [7, 20, 30-32]. Despite the morphologic signs that aid in differentiating benign vertebral compression fractures (retropulsion of osseous fragments, preserved marrow on T1-weighted, and isointense signal on T2-weighted and on contrast-enhanced T1-weighted) from malignant vertebral compression fractures (posterior bulging of cortex, epidural mass, diffusely low T1-weighted signal in the whole vertebral body, heterogeneous or high T2-weighted and contrast-enhanced T1-weighted signal, destruction of pedicles) [30, 33, 34], there are no reliable signal differences allowing discrimination of benign osteoporotic and neoplastic vertebral compression fractures when conventional pulse sequences are used. Therefore, it is important to find additional qualitative or quantitative criteria for more reliable distinction between these two fracture types.
Fig. 2A 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
A, T1-weighted (A) and STIR (B) images.
Fig. 2B 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
B, T1-weighted (A) and STIR (B) images.
Fig. 2C 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
C, Diffusion-weighted (DW) steady-state free-precession image shows isointense signal at delta value of 3 ms.
Fig. 2D 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
D, DW echo-planar image obtained using b value of 50 s/mm2 shows hyperintensity.
Fig. 2E 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
E, In-phase image shows hypointense signal.
Fig. 2F 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
F, Opposed-phase image shows hyperintense signal.
Fig. 2G 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
G, DW single-shot turbo spin-echo (TSE) image shows hyperintense signal at b value of 100 s/mm2.
Fig. 2H 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
H, Apparent diffusion coefficient (ADC) is 0.9 × 10-3 mm2/s on DW echo-planar image (H) and 1.38 × 10-3 mm2/s on DW single-shot TSE image (I).
Fig. 2I 64-year-old woman with malignant fracture of L3 vertebra due to metastasis of adenocarcinoma.
I, Apparent diffusion coefficient (ADC) is 0.9 × 10-3 mm2/s on DW echo-planar image (H) and 1.38 × 10-3 mm2/s on DW single-shot TSE image (I).
Fig. 3A Comparison of osteoporotic and malignant vertebral fractures. ⋄ = outlier.
A, Boxplots summarize values of signal intensity (SI) differences between in- and opposed-phase measurements (A) and apparent diffusion coefficients (ADCs) determined with diffusion-weighted (DW) echo-planar imaging (B) and DW single-shot turbo spin-echo imaging (C). EPI = echo-planar imaging; HASTE = half-Fourier acquisition single-shot turbo spin-echo.
Fig. 3B Comparison of osteoporotic and malignant vertebral fractures. ⋄ = outlier.
B, Boxplots summarize values of signal intensity (SI) differences between in- and opposed-phase measurements (A) and apparent diffusion coefficients (ADCs) determined with diffusion-weighted (DW) echo-planar imaging (B) and DW single-shot turbo spin-echo imaging (C). EPI = echo-planar imaging; HASTE = half-Fourier acquisition single-shot turbo spin-echo.
Fig. 3C Comparison of osteoporotic and malignant vertebral fractures. ⋄ = outlier.
C, Boxplots summarize values of signal intensity (SI) differences between in- and opposed-phase measurements (A) and apparent diffusion coefficients (ADCs) determined with diffusion-weighted (DW) echo-planar imaging (B) and DW single-shot turbo spin-echo imaging (C). EPI = echo-planar imaging; HASTE = half-Fourier acquisition single-shot turbo spin-echo.
TABLE 3: Differences of Signal Intensity (SI) on In- and Opposed-Phase Images and Apparent Diffusion Coefficient (ADC) Values in Patients With Osteoporotic and Malignant Vertebral Fractures
In the past, DWI and chemical-shift imaging have been evaluated with different results [11, 15, 18-20, 22, 24-27, 29, 35, 36]. In our study, we present for the first time, to our knowledge, an approach in which these different methods, as described in the literature, are compared in one patient cohort.

Diffusion-Weighted Imaging

DWI reflects the free mobility of water molecules in interstitial tissue, which is different in water and fat. The self-diffusion of water molecules can be described qualitatively as signal alteration in comparison to healthy surrounding vertebral bone marrow or quantitatively as the ADC. Malignant fractures show restricted diffusion—that is, low ADCs and high SI on DW images—because dense tumor cell packing should lead to a smaller and more restricted extracellular space and to decreased diffusion capability [7]. Acute osteoporotic fractures, on the other hand, mostly show increased diffusion with high ADCs and low SIs on DW images because water proton mobility is increased as a result of the bone marrow edema [20, 21, 23].
The PSIF sequence is a DW-SSFP sequence in which only a single (monopolar) diffusion gradient is inserted into each TR. The exact quantification of ADC and b value is not possible because this pulse sequence is influenced by many sequence parameters [35]. In our study, PSIF imaging with a delta value of 3 ms had the highest accuracy (93.5%) of all qualitative and quantitative parameters. At a delta value of 4.5 ms, the accuracy (91.3%) was slightly lower. At a delta value of 3 ms, there was no false-negative result (hypo- or isointense signal in malignant fractures) but three false-positive benign fractures showed hyperintense signal. At a diffusion pulse length of 4.5 ms, there were three false-negative fractures that showed isointense signal in neoplastic fractures and one false-positive benign fracture. At a delta value of 6 ms, sensitivity markedly decreased (Tables 1 and 2). To avoid missing a pathologic fracture, we consider a diffusion pulse length of 3 ms as the most appropriate one. Divergent results in other studies with several hypointense vertebral metastases on DW-SSFP [11] might be explained by previous treatment with radiotherapy or sclerotic metastases with lack of water protons in these other studies [35]. Vertebral metastases—from prostate cancer, in particular—show less signal than metastases from other tumors because of the high amount of sclerosis usually present in that cancer entity [36].
TABLE 4: Cutoff Values Determined With Receiver Operating Characteristic Analysis, Area Under the Curve (AUC), and Resulting Performance Values
The EPI technique allows a decrease in the acquisition time, but general problems are a limited spatial resolution, sensitivity to eddy currents and local susceptibility gradients, and chemical-shift artifacts [37]. With increasing b values, the (relative) background noise in our study was higher, as has also been reported in previous studies [14]. At b values of 50 and 250 s/mm2, most malignant lesions as well as benign lesions were hyperintense probably because of the T2 shine-through effect [22].
We did not find significant differences in the relative SIs of benign and malignant fractures using a DW single-shot TSE sequence at b values from 100 to 600 s/mm2 because all lesions showed high SI, possibly due to T2 shine-through effects. In our study, when using DW-EPI with a b value of 500 s/mm2, significant differences in the relative SIs of malignant and benign fractures were found; however, specificity (53.8%) was very low. SI depends not only on the b value, but also on other sequence parameters [35].
In the literature, ADC values of benign osteoporotic and benign traumatic fractures vary from 0.32 to 2.23 × 10-3 mm2/s and range from 0.19 to 1.04 × 10-3 mm2/s in malignant fractures or metastases; this variability can be explained by the different pulse sequences and diffusion weightings used, especially the use of fat saturation [35]. Because the ADC of vertebral fat is close to zero, DWI sequences without fat saturation systematically show decreased values [35]. Although ADC values should theoretically be platform-independent, the ADCs obtained with DW-EPI in our study were generally higher than those obtained with DW single-shot TSE probably because of more artifacts and higher background noise. As in previous studies, DW-EPI showed more susceptibility heterogeneities, resulting in distortion artifacts, than DW single-shot TSE [10]. The ADCs of osteoporotic and malignant fractures using DW-EPI did not differ significantly, whereas the ADCs of osteoporotic and malignant fractures determined with DW single-shot TSE differed significantly but showed a large overlap. At an estimated cutoff value of 1.48 × 10-3 mm2/s, the accuracy was 78.3% with five false-positive and five false-negative fractures.

Chemical-Shift Imaging

Chemical-shift MRI, or in- and opposed-phase MRI, is able to quantify fat in tissue and is routinely used to classify adrenal and liver lesions [5]. However, only a few studies have assessed this technique in the spine [5, 17-19]. Water and fat protons have different precession frequencies and are in phase at a TE of 4.8 ms and are 180° opposed at a TE of 2.4 ms at 1.5 T. This difference results in strong signal suppression if the amount of fat and water is nearly equal [5]. If the bone marrow is replaced by tumor, there is a lack of signal loss on opposed-phase images. Because fatty components within osteoporotic fractures are more or less preserved, a stronger signal loss is expected [5].
Zampa et al. [24] examined 68 patients at 0.5 T and noted a different behavior of benign and malignant fractures based on visual inspection of opposed-phase images. They reported that opposed-phase imaging showed strong hyperintensity in 38 patients (34 with malignant fractures, four with benign fractures), moderate hyperintensity in 28 patients (nine with malignant fractures, 19 with benign fractures), and low SI in 20 patients (two with malignant fractures, 18 with benign fractures). We could not confirm their findings in our study. We did not find significant differences in the SIs on opposed-phase images of benign and malignant vertebral fractures; 69.2% of all osteoporotic fractures showed a hyperintense signal (false-positive). This finding could be explained by the fact that the equal amount of 50% fat and 50% water is also no longer present in benign fractures because of edema, fibrosis, and sclerosis, resulting in more or less hyperintense signal.
Similar to previous studies [5, 17-19], our results showed that the mean signal loss in osteoporotic fractures (-24.8%) is greater than that of malignant fractures (-8.5%) and that both groups differ significantly (p = 0.01). However, the significant overlap must be noted. With a highest accuracy of 71.7%, a sensitivity of 50%, and a specificity of 88.5%, the qualitative assessment of chemical-shift MRI is inferior to quantitative DW single-shot TSE and qualitative PSIF imaging at delta values of 3 and 4.5 ms. This result differs from those of other studies that found high sensitivities of 95% and specificities of up to 100% [5, 19]. In our study, 50% of the malignant vertebral fractures were false-negative according to the cutoff value of -1.44%, as best determined by ROC analysis. All of these lesions were true-positive for metastasis in PSIF imaging at a delta value of 3 ms.

Limitations

One limitation of our study was the presence of different types of primary tumors, which might have caused slightly divergent signal behavior in the different sequences. Different tumors have a different mixture of tissue elements, water, and extracellular matrix. These differences might lead to variability in water and proton mobility and thus to different ADCs, in- and opposed-phase ratios, and SIs in SSFP sequences. Another limitation of our study might be a lack of histologic correlation in some patients due to ethical reasons. Not every patient with an osteoporotic vertebral fracture had to be treated surgically or had to undergo vertebroplasty. Verification by biopsy is not routinely performed in patients with a clinically apparent osteoporotic fracture. However, clinical and imaging follow-ups were performed in all cases. Reduction of edema is a clear criterion for the benign cause of a fracture. Also, in patients with clinically and radiologically apparent malignant fractures or if a malignant fracture does not result in a change of treatment, bioptic verification is not performed. Three of the patients with malignant fractures did not undergo biopsy because they had multiple metastases and the malignant origin of the fractures was not clinically and radiologically in doubt. Another limitation might be the sample size of 46 cases in the study, but the initial sample size considerations show that at a sample size of 17 subjects per group, a sufficient power (α = 0.05, β = 0.9) is provided. The significant differences depicted in the examined pulse sequences justify the assumption that differentiation between osteoporotic and malignant fractures is possible.

Conclusions

DW-PSIF imaging with a delta value of 3 ms is a valuable tool for differentiating benign and malignant vertebral fractures. Quantitative chemical-shift imaging and ADC values (DW single-shot TSE) also showed significant differences between the two entities, but these differences were inferior to DW-PSIF because of a large overlap. Qualitatively assessed opposed-phase, DW-EPI, and DW single-shot TSE sequences and ADCs (DW-EPI) could not be used to accurately differentiate benign from malignant vertebral fractures.
Further studies in a larger cohort of patients should be encouraged.

Acknowledgments

We thank Fabian Bamberg for his invaluable assistance with the statistical part of the study.

Footnote

Supported by the Deutsche Forschungsgemeinschaft (DFG grant DI 1413/1-1).

References

1.
Fornasier VL, Czitrom AA. Collapsed vertebrae: a review of 659 autopsies. Clin Orthop Relat Res 1978; 131:261–265
2.
Yuh WT, Zachar CK, Barloon TJ, Sato Y, Sickels WJ, Hawes DR. Vertebral compression fractures: distinction between benign and malignant causes with MR imaging. Radiology 1989; 172:215–218
3.
Frager D, Elkin C, Swerdlow M, Bloch S. Subacute osteoporotic compression fracture: misleading magnetic resonance appearance. Skeletal Radiol 1988; 17:123–126
4.
Shih TT, Huang KM, Li YW. Solitary vertebral collapse: distinction between benign and malignant causes using MR patterns. J Magn Reson Imaging 1999; 9:635–642
5.
Erly WK, Oh ES, Outwater EK. The utility of in-phase/opposed-phase imaging in differentiating malignancy from acute benign compression fractures of the spine. Am J Neuroradiol 2006; 27:1183–1188
6.
Leeds NE, Kumar AJ, Zhou XJ, McKinnon GC. Magnetic resonance imaging of benign spinal lesions simulating metastasis: role of diffusion-weighted imaging. Top Magn Reson Imaging 2000; 11:224–234
7.
Baur A, Dietrich O, Reiser M. Diffusion-weighted imaging of bone marrow: current status. Eur Radiol 2003; 13:1699–1708
8.
Baur A, Stabler A, Bruning R, et al. Diffusion-weighted MR imaging of bone marrow: differentiation of benign versus pathologic compression fractures. Radiology 1998; 207:349–356
9.
Baur-Melnyk A. Malignant versus benign vertebral collapse: are new imaging techniques useful? Cancer Imaging 2009; 9(spec no A):S49–S51
10.
Biffar A, Baur-Melnyk A, Schmidt GP, Reiser MF, Dietrich O. Multiparameter MRI assessment of normal-appearing and diseased vertebral bone marrow. Eur Radiol 2010; 20:2679–2689
11.
Castillo M, Arbelaez A, Smith JK, Fisher LL. Diffusion-weighted MR imaging offers no advantage over routine noncontrast MR imaging in the detection of vertebral metastases. Am J Neuroradiol 2000; 21:948–953
12.
Karchevsky M, Babb JS, Schweitzer ME. Can diffusion-weighted imaging be used to differentiate benign from pathologic fractures? A meta-analysis. Skeletal Radiol 2008; 37:791–795
13.
Baur A, Huber A, Dürr HR, et al. Differentiation of benign osteoporotic and neoplastic vertebral compression fractures with a diffusion-weighted, steady-state free precession sequence [in German]. Rofo 2002; 174:70–75
14.
Tang G, Liu Y, Li W, Yao J, Li B, Li P. Optimization of b value in diffusion-weighted MRI for the differential diagnosis of benign and malignant vertebral fractures. Skeletal Radiol 2007; 36:1035–1041
15.
Oztekin O, Ozan E, Hilal Adibelli Z, Unal G, Abali Y. SSH-EPI diffusion-weighted MR imaging of the spine with low b values: is it useful in differentiating malignant metastatic tumor infiltration from benign fracture edema? Skeletal Radiol 2009; 38:651–658
16.
Disler DG, McCauley TR, Ratner LM, Kesack CD, Cooper JA. In-phase and out-of-phase MR imaging of bone marrow: prediction of neoplasia based on the detection of coexistent fat and water. AJR 1997; 169:1439–1447
17.
Eito K, Waka S, Naoko N, Makoto A, Atsuko H. Vertebral neoplastic compression fractures: assessment by dual-phase chemical shift imaging. J Magn Reson Imaging 2004; 20:1020–1024
18.
Zajick DC, Morrison WB, Schweitzer ME, Parellada JA, Carrino JA. Benign and malignant processes: normal values and differentiation with chemical shift MR imaging in vertebral marrow. Radiology 2005; 237:590–596
19.
Ragab Y, Emad Y, Gheita T, et al. Differentiation of osteoporotic and neoplastic vertebral fractures by chemical shift {in-phase and out-of phase} MR imaging. Eur J Radiol 2009; 72:125–133
20.
Balliu E, Vilanova JC, Peláez I, et al. Diagnostic value of apparent diffusion coefficients to differentiate benign from malignant vertebral bone marrow lesions. Eur J Radiol 2009; 69:560–566
21.
Chan JH, Peh WC, Tsui EY, et al. Acute vertebral body compression fractures: discrimination between benign and malignant causes using apparent diffusion coefficients. Br J Radiol 2002; 75:207–214
22.
Herneth AM, Friedrich K, Weidekamm C, et al. Diffusion weighted imaging of bone marrow pathologies. Eur J Radiol 2005; 55:74–83
23.
Herneth AM, Naude J, Philipp M, Beichel R, Trattnig S, Imhof H. The value of diffusion-weighted MRT in assessing the bone marrow changes in vertebral metastases [in German]. Radiologe 2000; 40:731–736
24.
Zampa V, Cosottini M, Michelassi C, Ortori S, Bruschini L, Bartolozzi C. Value of opposed-phase gradient-echo technique in distinguishing between benign and malignant vertebral lesions. Eur Radiol 2002; 12:1811–1818
25.
Sourbron S, Biffar A, Ingrisch M. Platform for research in medical imaging. (abstract) In:, Sourbron S, Biffar A, Ingrisch M, Fierens Y, Luypaert R (eds). ESMRMB09 Antalya, 2009 MAGMA Magn Reson Mater Phy 2009; 22(suppl 1):539
26.
Stejskal EO, Tanner JE. Spin diffusion measurements: spin echoes in the presence of a time-dependent field gradient. J Chem Phys 1965; 42:288–292
27.
Park SW, Lee JH, Ehara S, et al. Single shot fast spin echo diffusion-weighted MR imaging of the spine: is it useful in differentiating malignant metastatic tumor infiltration from benign fracture edema? Clin Imaging 2004; 28:102–108
28.
Baur A, Huber A, Ertl-Wagner B, et al. Diagnostic value of increased diffusion weighting of a steady-state free precession sequence for differentiating acute benign osteoporotic fractures from pathologic vertebral compression fractures. Am J Neuroradiol 2001; 22:366–372
29.
Vogler JB, Murphy WA. Bone marrow imaging. Radiology 1988; 168:679–693
30.
Cuénod CA, Laredo JD, Chevret S, et al. Acute vertebral collapse due to osteoporosis or malignancy: appearance on unenhanced and gadolinium-enhanced MR images. Radiology 1996; 199:541–549
31.
An HS, Andreshak TG, Nguyen C, Williams A, Daniels D. Can we distinguish between benign versus malignant compression fractures of the spine by magnetic resonance imaging? Spine (Phila Pa 1976) 1995; 20:1776–1782
32.
Moulopoulos LA, Yoshimitsu K, Johnston DA, Leeds NE, Libshitz HI. MR prediction of benign and malignant vertebral compression fractures. J Magn Reson Imaging 1996; 6:667–674
33.
Jung HS, Jee WH, McCauley TR, Ha KY, Choi KH. Discrimination of metastatic from acute osteoporotic compression spinal fractures with MR imaging. RadioGraphics 2003; 23:179–187
34.
Link TM, Guglielmi G, van Kuijk C, Adams JE. Radiologic assessment of osteoporotic vertebral fractures: diagnostic and prognostic implications. Eur Radiol 2005; 15:1521–1532
35.
Dietrich O, Biffar A, Reiser MF, Baur-Melnyk A. Diffusion-weighted imaging of bone marrow. Semin Musculoskelet Radiol 2009; 13:134–144
36.
Hackländer T, Scharwächter C, Golz R, Mertens H. Value of diffusion-weighted imaging for diagnosing vertebral metastases due to prostate cancer in comparison to other primary tumors [in German]. Rofo 2006; 178:416–424
37.
Le Bihan DJ. Differentiation of benign versus pathologic compression fractures with diffusion-weighted MR imaging: a closer step toward the “holy grail” of tissue characterization? Radiology 1998; 207:305–307

Information & Authors

Information

Published In

American Journal of Roentgenology
Pages: 1083 - 1092
PubMed: 23096183

History

Submitted: September 29, 2011
Accepted: February 16, 2012

Keywords

  1. chemical-shift imaging
  2. diffusion-weighted imaging
  3. metastatic fracture
  4. musculoskeletal imaging
  5. osteoporotic fracture
  6. spine
  7. vertebral body fractures

Authors

Affiliations

Tobias Geith
All authors: Institute of Clinical Radiology, LMU University of Munich, Campus Grosshadern, Marchioninistrasse 15, Munich, Bavaria 81377, Germany.
Gerwin Schmidt
All authors: Institute of Clinical Radiology, LMU University of Munich, Campus Grosshadern, Marchioninistrasse 15, Munich, Bavaria 81377, Germany.
Andreas Biffar
All authors: Institute of Clinical Radiology, LMU University of Munich, Campus Grosshadern, Marchioninistrasse 15, Munich, Bavaria 81377, Germany.
Olaf Dietrich
All authors: Institute of Clinical Radiology, LMU University of Munich, Campus Grosshadern, Marchioninistrasse 15, Munich, Bavaria 81377, Germany.
Hans Roland Dürr
All authors: Institute of Clinical Radiology, LMU University of Munich, Campus Grosshadern, Marchioninistrasse 15, Munich, Bavaria 81377, Germany.
Maximilian Reiser
All authors: Institute of Clinical Radiology, LMU University of Munich, Campus Grosshadern, Marchioninistrasse 15, Munich, Bavaria 81377, Germany.
Andrea Baur-Melnyk
All authors: Institute of Clinical Radiology, LMU University of Munich, Campus Grosshadern, Marchioninistrasse 15, Munich, Bavaria 81377, Germany.

Notes

Address correspondence to T. Geith ([email protected]).

Metrics & Citations

Metrics

Citations

Export Citations

To download the citation to this article, select your reference manager software.

Articles citing this article

Media

Figures

Other

Tables

Share

Share

Copy the content Link

Share on social media