OBJECTIVE. The aim of this study was to investigate diffusion-weighted MR imaging for differentiation of benign fracture edema and tumor infiltration with and without accompanying fracture.
SUBJECTS AND METHODS. In 10 volunteers, diffusion-weighted spin-echo, fat-suppressed spin-echo, and stimulated-echo sequences were optimized on a clinical 1.5-T scanner. In 34 patients, MR imaging with and without diffusion-sensitizing gradients (b = 598 sec/mm2 in spin-echo and fat-suppressed spin-echo, b = 360 sec/mm2 in stimulated-echo) was performed. Thirty-five lesions were analyzed, with 18 caused by acute (≤10 days old) osteoporotic or traumatic fractures and 17 caused by untreated malignant vertebral infiltration including nine fractures. Signal attenuation in diffusion-weighted images and contrast-to-noise ratio were calculated. The diffusion-weighted images were analyzed by two radiologists.
RESULTS. Images from three of 34 patients were excluded because of motion artifact. In osteoporotic and traumatic fractures, a strong signal attenuation of bone marrow edema was seen. In contrast to this, malignant-tumor infiltration caused only minor signal attenuation (p < 0.05), independent of accompanying pathologic fracture. All sequences showed identical changes of signal intensities. In four patients, initial diagnosis was changed by the findings in the diffusion-weighted images.
CONCLUSION. Diffusion-weighted spin-echo, fat-suppressed spin-echo, and stimulated-echo sequences are equally suitable for imaging of the spine. Calculation of signal attenuation and observation of signal characteristics allowed differentiation of benign fracture edema and tumor infiltration and provided excellent distinction between benign and malignant vertebral fractures in our series.
Diffusion-weighted MR imaging is a well-known technique in neuroradiology, especially for stroke patients [1,2,3,4]. Diffusion-weighted sequences are sensitive to molecular motion because random motion of water molecules in gradient fields produces phase dispersion and, therefore, signal attenuation . This information can be used for tissue characterization. Diffusion of water depends on many different factors like restriction, hindrance, membrane permeability, and tissue inhomogeneity [1, 6]. Restriction, as seen in cellular tissues, results in lower phase shift and lower signal attenuation .
Recently, new applications have been developed for liver  and pancreas imaging . First results using diffusion-weighted imaging of vertebral compression fractures showed a reduced water mobility in pathologic fractures . In that study, a steady-state free precession technique, which does not allow a quantification of the diffusion coefficient, was used [2, 11,12,13]. The lack of quantification reduced the usefulness of the steady-state free precession technique . Further limitations of the steady-state free precession technique are the lack of cardiac triggering, T2-contamination, and other confounding relaxation effects . These limitations are minor in spin-echo sequences . With such diffusion-weighted spin-echo sequences, Lang et al.  could differentiate necrotic and vital tumor parts in an animal osteosarcoma model because of lower water mobility in viable cellular structures. However, Ward et al.  showed an increased diffusion of interstitial water in posttraumatic bone marrow edema. From these results, we hypothesized that diffusion-weighted spin-echo sequences could differentiate benign fracture edemas and fractures caused by tumor infiltration due to higher restriction of water mobility in tumor cells.
In our study, we compared three optimized diffusion schemes. On the basis of research by Stejskal and Tanner , we used a fat-suppressed spin-echo sequence and a stimulated-echo sequence, in addition to a standard diffusion-weighted spin-echo sequence. We compared the different signal intensities of benign vertebral fracture edema caused by osteoporosis or trauma and fractures caused by malignant tumor infiltration. In addition to the fractures caused by malignancy, a subgroup with vertebral body tumor masses, but without accompanying fracture, was investigated as a control group.
Subjects and Methods
Sequence optimization was performed in 10 healthy volunteers (age range, 19-33 years) by subjective evaluation of the image quality (edge definition of the vertebral body and intervertebral disk at the level of the thoracicolumbar spine), motion artifacts, and signal attenuation of the intervertebral disk and normal fatty bone marrow due to changes of sequence parameters such as TR, TE, b-factor, image spatial resolution, and flip angle. For volunteers, we applied four b-factors: 0, 89, 355, and 798 sec/mm2 in the spin-echo and fat-suppressed spin-echo sequence and 0, 51, 204 and 460 sec/mm2 in the stimulated-echo sequence.
On the basis of routine examination with short tau inversion recovery (STIR) T1- and T2-weighted images in 34 patients (age range, 29-85 years), diffusion-weighted MR imaging of the spine was performed. Images of three patients were excluded from analysis because of motion artifact. In the remaining 31 patients, overall, we analyzed 35 vertebral body lesions: 18 acute traumatic or osteoporotic vertebral fractures and 17 untreated tumor infiltrations (plasmacytoma [n = 2], leukemia [n = 1], and metastases [n = 14]). In nine of 17 vertebral tumor infiltrations, an acute pathologic fracture was revealed on radiography and on conventional MR imaging. The time delay between onset of clinical symptoms (pain) and MR imaging in patients with fractures (benign and malignant) was 1-10 days. Informed consent was obtained from all patients participating the study. Diagnosis was proven with biopsy (nine lesions) or follow-up MR imaging (26 lesions). The follow-up criteria in which a resolution of the bone marrow edema was considered benign for diagnosis were described by Moulopoulos et al. . A malignant nature of the fractures or lesions was established by an increased edema size with additional soft-tissue masses or with new metastases. In four patients, two lesions each were examined (two osteoporotic fractures [one woman and one man] and two patients with multiple metastasis [amelanotic malignant melanoma and bronchogenic cancer]).
Diffusion-weighted MR imaging was performed on a 1.5-T ACS-NT scanner (Philips, Best, The Netherlands) with a gradient strength of 23 mT/m. We used a circular surface coil (diameter, 17 cm), which was placed directly posterior to the region of interest (ROI). An elastic body belt was fixed around the upper abdomen to reduce breathing artifacts. No respiratory triggering was used. After lesion localization with a sagittal STIR sequence (TR/TE, 2000/70; inversion time, 130 msec; slice thickness, 4 mm), we selected the ROI for sagittal diffusion-weighted imaging. In the volunteer study, we used the following sequence parameters: 3 slices, 6- to 8-mm thick; field of view, 260 mm; matrix size, 256 × 128; cardiac triggering, TR 1 R-R interval. The total TE was 96 msec in the spin-echo and fat-suppressed spin-echo sequence. For the stimulated-echo sequence, we used a total TE of 116 msec, including a time delay of 75 msec between the second and third 90° pulse. The b-factor, which describes the effective diffusivity of a sequence, was calculated by the equation b = γ2G2δ2(Δ - δ/3) (γ = gyromagnetic ratio, G = gradient strength, δ = gradient length, Δ = diffusion time). For the patient study, we applied two b-factors: 0 and 598 sec/mm2 in the spin-echo and fat-suppressed spin-echo sequences (diffusion time, 40 msec; diffusion gradient length, 39 msec) and 0 and 360 sec/mm2 in the stimulated-echo sequence (total diffusion time, 95 msec; diffusion gradient length, 14 msec). Fat-suppression in the fat-suppressed spin-echo sequences was performed by a spectral inversion recovery technique.
Because of strong motion artifact in diffusion-weighted images, special motion-artifact reduction techniques were necessary. We employed a navigator echo motion correction technique [3, 18, 19], which uses a pair of 180° pulses, one with and one without phase-encoding gradients, to correct phase errors. In our sequences, the first 180° pulse was acquired without and the second one with phase-encoding gradients. The diffusion-sensitizing gradients and the navigator echo were set in phase-encoding direction (feet-head) to obtain the best navigator echo artifact correction. The application of multidirectional diffusion gradients was not possible.
Four regional saturation slabs were positioned to reduce artifact from surrounding tissue (one ventrally; one dorsally close to the border of the vertebral body or the dorsal elements of the vertebrae, depending on lesion localization; one caudally; and one cranially at the margin of the field of view). The slices were angulated so that the phase-encoding gradient direction was parallel to the ventral border of the vertebral bodies.
Identical sequences were performed with and without diffusion-sensitizing gradients. Identical geometric parameters were used to ensure accurate comparison of the same regions. A peripheral fingertip pulse sensor was used for cardiac triggering. The trigger interval was set to shortest (180 msec) to reduce artifact originating from the aortic pulse wave. The acquisition time for all three diffusion-weighted sequences was approximately 20 min, depending on the heart rate. Diffusion-weighted spin-echo, fat-suppressed spin-echo, and stimulated-echo imaging was performed in all except one patient for whom the stimulated-echo sequence was aborted because of pain.
MR Data Analysis
Signal intensity data of the edematous regions were obtained by the implemented software of the scanner. A user-defined ROI was placed in the diffusion-weighted spin-echo images where the lesion or the edematous region was optimally visible. The diffusion-weighted images were chosen because of lack of differentiation of potential accompanying fracture edema without tumor infiltration on conventional MR images. For lesions that appeared hypointense, the largest fitting circle or ellipse was positioned centrally. For hyperintense or edematous regions with hyperintense signal intensity and additional accompanying hypointense regions, the largest fitting circle or ellipse in the hyperintense region was chosen. Because of our hypothesis that diffusion-weighted imaging can differentiate between tumor infiltration and fracture edema without tumor infiltration, this data analysis avoided potential mixture with pixels of the accompanying fracture edema. Subsequently, the chosen ROI was copied to all investigated images to compare identical edematous areas.
For quantification of diffusion effects, typically the apparent diffusion coefficient is computed from the slope of the semilog plot of the signal intensity as a function of the b-factor. This slope can only be reliably calculated from multiple b-factors . Because in our study for each sequence only two b-factors could be applied, we calculated normalized signal attenuation. Using identical sequences allows the comparison of different lesion entities.
Normalized signal attenuation was calculated by dividing the signal intensity of the diffusion-weighted images and the signal intensity of the images without diffusion-sensitizing gradients and by setting the signal intensity of the sequences without diffusion-sensitizing gradients to 100% (Fig. 1). Contrast-to-noise ratio (CNR) was calculated by CNR = [│S(normal) - S(lesion)│] / noise. Noise was defined as a standard deviation in a region of air (in frequency-encoding direction). S(lesion) was the signal intensity as measured for the signal attenuation calculation. S(normal) was measured from a user-defined ROI in an adjacent vertebral body without signal enhancement on STIR images. In two patients with diffuse bone marrow infiltration (leukemia [n = 1] and diffuse infiltration of myeloma [n = 1]), no CNR was calculated because of lack of normal bone marrow signal intensity.
The data of the malignant lesions were separated into two subgroups: one with and one control group without accompanying fracture.
The diagnosis based on the diffusion-weighted images was made by two radiologists who achieved unanimous agreement in all cases. Hereby, diffusion-weighted images were presented in conjunction with the STIR images, but without information about the calculated signal intensities.
For statistical analysis of the signal attenuation (interindividual), we used the Student's t test for unpaired groups. For comparison of the CNR with and without diffusion-sensitizing gradients (intraindividual), the paired Student's t test was calculated. A p value of < 0.05 was considered significant.
In 31 of 34 patients, image quality was judged to be diagnostic, and 35 lesions could be examined. Images of three patients were excluded from data analysis because of motion artifact. Lesions were located in the thoracic (n = 22) or lumbar (n = 13) spine.
All vertebral lesions showed increased signal intensity on STIR images. In each of the non-diffusion-weighted sequences, benign fracture edema and tumor infiltration showed similar signal intensity; spin-echo sequences without diffusion-sensitizing gradients showed a moderately higher signal intensity in benign vertebral body fractures and in tumor edema compared with normal fatty bone marrow. In non-diffusion-weighted stimulated-echo images only a minimal contrast was found (minimally hypointense signal intensity in benign and minimally hyperintense signal intensity in malignant lesions). In fat-suppressed spin-echo sequences, a strong hyperintense signal was found in benign and malignant lesions (Figs. 2,3A,3B,3C,3D,3E,3F,4A,4B,4C,4D,4E,4F). In non-diffusion-weighted images, no statistically different CNR for the two entities of vertebral fractures (benign and malignant) was calculated for all three investigated sequences.
After application of additional diffusion-sensitizing gradients, the CNR dramatically changed in benign vertebral fracture edema (p < 0.05 in all three investigated sequences). In spin-echo and stimulated-echo images a markedly hypointense signal intensity occurred (Figs. 2 and 3A,3B,3C,3D,3E,3F), and in fat-suppressed spin-echo images only a slightly hyperintense signal remained. In contrast to this phenomenon, no significant change of CNR was found in tumor lesions (p < 0.05 in all three investigated sequences) (Figs. 2 and 4A,4B,4C,4D,4E,4F). This finding was independent of accompanying fracture.
Unlike non-diffusion-weighted images, images with diffusion-sensitizing gradients showed a statistically different CNR for the two entities of vertebral fractures (benign and malignant) with the spin-echo and fat-suppressed spin-echo sequences (p < 0.05). For the stimulated-echo sequence, no statistically significant difference was calculated (Fig. 2).
Normalized Signal Attenuation
The calculated signal attenuation showed a significant difference between tumor lesions with or without accompanying fracture and benign fracture edema in all three sequences (p < 0.05). In osteoporotic and traumatic fracture a strong reduction of signal intensity was seen (44% in spin-echo, 52% in stimulated-echo, 38% in fat-suppressed spin-echo) (Figs. 1 and 3A,3B,3C,3D,3E,3F). In contrast to this finding, in tumor lesions without vertebral fracture and with accompanying fracture, no or only minor signal attenuation was found (for all malignant lesions on average, 96% in spin-echo, 95% in stimulated-echo, 91% in fat-suppressed spin-echo) (Figs. 1 and 4A,4B,4C,4D,4E,4F).
In five of nine patients showing pathologic fractures, diffusion-weighted images showed an area of attenuated signal adjacent to an area of unsuppressed signal intensity. We hypothesized that this signal attenuation was due to accompanying fracture edema, whereas the unsuppressed signal represented malignant tumor edema (Fig. 5A,5B).
In our series, diffusion-weighted images led to the correct diagnosis in four patients with vertebrae collapse, for whom the diagnosis was unclear after routine examination (n = 1) or was changed because of diffusion-weighted MR imaging (n = 3). A 62-year-old woman with a history of lung carcinoma and brain metastasis presented with acute fracture of three vertebral bodies (T10-T12) without trauma. Clinical symptoms (back pain) had occurred 8 days before MR imaging. Routine MR imaging suggested the diagnosis of metastasis. Radiotherapy was planned. Diffusion-weighted images showed strong signal attenuation (Fig. 6A,6B), and the diagnosis was changed to an osteoporotic fracture. The patient declined biopsy or therapy except for wearing a corset. Follow-up imaging 6 months later showed strong decrease of edema in all three vertebrae. Pain had gradually diminished and totally vanished after 6 months. Another 61-year-old woman presented with weight loss and reported a minor trauma. Routine MR imaging showed geographic bone marrow edema in the fractured vertebral body. Edema extended to the pedicle, and a pathologic fracture was suggested. Diffusion-weighted images showed strong signal attenuation in all sequences. Surgery was performed to stabilize the segment, and operative biopsy found edema and hemorrhage but no tumor cells. A third patient with known osteoporosis due to long-term medication of steroids suffered from an acute vertebral fracture after minor trauma. Diffusion-weighted images revealed a small hyperintense lesion in the dorsal part of the fracture, whereas the ventral part showed marked signal reduction (Fig. 5A,5B). The diagnosis of a pathologic fracture due to tumor mass was made after reviewing additional examinations that showed further osseous metastasis of a formerly occult lung cancer. A vertebral body fracture in a fourth patient with urothelial carcinoma was correctly diagnosed as osteoporotic fracture, whereas routine MR examination could not differentiate benign and malignant cause.
MR imaging has been applied for several years to differentiate benign and malignant vertebral body collapse. Several signs have been described [17, 20, 21], but neither a single one nor a combination of several signs yields a sufficiently high specificity, in light of the inherent clinical importance of differentiating benign and malignant diseases [2, 20,21,22,23,24]. Signal enhancement in tumor masses on T2-weighted images reflects mainly intracellular water, whereas fracture edema reflects interstitial water. These two bone marrow edema entities cannot be differentiated with conventional MR imaging because of similar signal intensities . Therefore, a new highly specific criteria for differentiation of benign and malignant vertebral fractures would be helpful [2, 10, 17, 22]. Water in vital tumor cells shows lower mobility as a result of cellular structures . In the presence of diffusion-sensitizing gradients, this finding should result in a lower signal attenuation compared with stronger dephasing of more mobile extracellular water with extensive signal attenuation. Accordingly, diffusion-weighted imaging should differentiate benign (osteoporotic and traumatic) and malignant vertebral fractures. We examined the feasibility of distinguishing benign fracture edema and malignant tumor infiltration with diffusion-weighted spin-echo, stimulated-echo, and fat-suppressed spin-echo sequences. In addition to vertebral fractures caused by tumor infiltration, a further control subgroup with tumor lesions without accompanying fractures was investigated.
In our study, we compared benign acute osteoporotic or traumatic fracture edema with untreated tumor lesions with and without accompanying fracture (malignant bone marrow edema). Our results showed an extensive signal attenuation (38-52%) in benign fracture edema, whereas in tumor infiltration independent of accompanying fractures, no significant signal reduction was found. These differences were subjectively visible by reviewing the MR images and by yielding a statistically significant difference for normalized signal attenuation and CNR. This finding indicates that diffusion-weighted spin-echo, stimulated-echo, and fat-suppressed spin-echo MR imaging can differentiate benign fracture edema and malignant vertebral fractures caused by tumor infiltration. We found no difference of signal change in malignant vertebral fractures and in the control subgroup with tumor infiltration but without accompanying fracture.
This finding indicates that the lack of signal reduction in malignant vertebral fractures is caused by tumor cell infiltration.
Because diffusion effects in biologic tissues are complex and the mechanism that allows differentiation of cellular edema with intracellular water in tumor infiltration and interstitial edema in benign fractures remains speculative, further studies are needed to investigate this issue. But our results are in close correlation with first differentiation of pathologic and traumatic or osteoporotic vertebral compression fractures with diffusion-weighted steady-state free precession techniques described by Baur et al. . We agree with their theory that the different diffusion effect is caused by more restriction or hindrance in densely packed tumor cells compared with more mobile water in extracellular volume fractions in fractures. This theory was drawn from biologic models [6, 26], in vitro studies, and theoretic models [27, 28] that describe a different mobility of intracellular and extracellular water.
In the model of two entities of bone marrow edema, which can be differentiated by diffusion-weighted imaging depending on cellular structures, cells other than tumor cells or cell death must be taken into account. We avoided influences caused by cell immigration and granulation tissue in later states of fracture healing [29, 30] or cell death and necrosis in treated tumor masses, which also influence diffusion-weighted images [10, 31, 32], by including only acute fractures (≤ 10 days old) and untreated tumor infiltration.
All three investigated sequences (spin-echo, stimulated-echo, and fat-suppressed spin-echo) enabled us to differentiate malignant tumor infiltration and benign fracture edema in routine diffusion-weighted imaging of the spine. We compared the signal intensities of identical sequences with and without diffusion-sensitizing gradients to exclude T2 and other confounding relaxation effects. To reduce perfusion effects , we used peripheral pulse-wave triggering and high b-factors. According to the intravoxel incoherent motion and perfusion model [5, 26], perfusion effects can be theoretically reduced by comparison of a lower and higher b-factor. In our study, only one high and no b-factor were applied because a further b-factor would have prolonged the acquisition time beyond their tolerance in patients with vertebral fractures. But by using higher b-factors like those used in our study (598 sec/mm2 for the spin-echo and fat-suppressed spin-echo sequence and 360 sec/mm2 for the stimulated-echo sequence), we found that these sequences were already relatively less sensitive to perfusion effects.
Because quantifying signal changes is helpful in diffusion-weighted imaging , we calculated normalized signal attenuation. Apparent diffusion coefficient calculation that quantifies diffusion effects related to diffusion and perfusion in biologic tissue  was not applied because no reliable data can be achieved from two b-factors . But in the future, by using only one of the presented diffusion schemes and multiple b-factors, we believe that sufficient apparent diffusion coefficient calculation of bone marrow edemas may be available.
Diffusion-weighted imaging is sensitive to motion artifact [4, 11]. To overcome this problem, echo-planar imaging techniques have been introduced in brain and abdominal imaging [4, 8, 9, 33] at the cost of high susceptibility artifact, as shown in spinal cord imaging . In preparation for our study, in a previous volunteer study, we also observed these effects in vertebral body imaging and, therefore, did not apply an echo-planar imaging technique. In our study, for motion artifact correction, a navigator echo technique was used as previously described [3, 18]. Further motion artifact reduction was obtained by using a peripheral pulse-wave triggering and regional saturation slabs. With these techniques, diagnostic image quality was obtained in 31 of 34 patients.
In contrast to spin-echo and stimulated-echo sequences used in our study, in the steady-state free precession technique used by Baur et al. , navigator echo, cardiac triggering, and diffusion quantification were not available. Further limitations for steady-state free precession are the T2-contamination and other confounding relaxation effects  and the difficulty in quantification of diffusion effects due to steady-state conditions [2, 11,12,13]. Therefore, we did not use a steady-state technique.
In our study, we used a stimulated-echo technique as a second alternative for diffusion-weighted imaging. With this technique, an identical signal change, as in the spin-echo sequences, that confirmed the reliability of both diffusion-weighted imaging techniques was found. A limitation in stimulated-echo techniques, however, is a lower signal-to-noise ratio . The main advantages of this technique are a low T2 contamination and the possibility of combining stimulated-echo with further motion artifact correction techniques. Stimulated-echo techniques use three 90° pulses before an echo with a 180° pulse is obtained from the so-called stimulated-echo. Because of a lack of T2-relaxation effects, during the phase between the second and third 90° pulse, long diffusion times without contribution to the T2-decay can be performed for high b-factors. Furthermore, a triggering of the third 90° pulse with the cardiac cycle or breathing is possible. These possible advantages could not be exploited in our study because longer acquisition times would have increased examination time too much to compare three sequences.
Despite the lower signal-to-noise ratio of the stimulated-echo compared with the other sequences, diffusion-weighted stimulated-echo MR imaging has been proven possible. Application of long diffusion times with stimulated-echo techniques holds promise to investigate tissue microstructures and apparent diffusion coefficient [6, 12, 31].
In our study, in four (12.9%) of 31 patients with bone marrow edema of unknown origin, only diffusion-weighted images revealed the correct diagnosis (Figs. 5A,5B and 6A,6B). The original diagnosis based on routine MR sequences was changed because of diffusion-weighted imaging in three patients (9.7%). In one patient, standard MR imaging could not differentiate benign and malignant disease, whereas diffusion-weighted imaging allowed us to make the diagnosis of benign osteoporotic fracture despite a known malignant tumor history. Diffusion-weighted imaging proved correct in these patients and all other patients, in whom it yielded the same diagnosis as routine MR imaging. Despite the relatively low number of patients, the sensitivity and specificity of 100% is promising. Although our results showed that diffusion-weighted imaging can contribute essential treatment information that is not available in conventional MR techniques, the nondiagnostic image quality of three (8.8%) of 34 patients and the inclusion criteria of acute fractures and untreated malignant disease limit the value of our study. Furthermore, the overall reduced signal-to-noise ratio due to the strong diffusion gradients requires the use of surface coils, which limit the field of view. This problem can be circumvented if the application of synergy coils for navigated diffusion-weighted imaging becomes possible.
As stated by Le Bihan , in pathologic fractures a mixture of interstitial edema and cell infiltration occurs and may reduce the specificity of diffusion-weighted imaging. Our study included nine vertebral fractures due to tumor infiltration. In five patients with accompanying fracture edema, it was possible to differentiate an area of signal attenuation adjacent to an edematous area that showed no signal attenuation on diffusion-weighted images. Relying on our hypothesis that tumor cell edema shows no signal reduction on diffusion-weighted images, we placed the ROIs accordingly. This placement allowed us to correctly diagnose all pathologic fractures by diffusion-weighted imaging. Despite the lack of histologic proof, the overall data of our study support the theory that diffusion-weighted imaging can reliably differentiate tumor lesions and accompanying bone marrow edema due to pathologic fracture. This finding might be useful in guiding bone biopsies to avoid false-negative results. Further investigations with larger numbers of patients with different types of malignancies, like diffuse bone marrow infiltration, and examination of necrotic tumor masses must be performed. Also, further proper quantification studies to fully understand the underlying mechanism and to define the full range of indication for diffusion-weighted imaging of vertebral bone marrow are needed.
In conclusion, our results show that navigated diffusion-weighted spin-echo, fat-suppressed spin-echo, and stimulated-echo sequences are feasible in vertebral body imaging on a clinical MR scanner. Fractures caused by untreated tumor infiltration can be differentiated from acute osteoporotic and traumatic fractures by comparison of identical sequences without and with diffusion-sensitizing gradients. Thereby, additional information that yielded a correct change of the initial diagnosis in four of 31 patients in our series was obtained. All three sequences provide the same information. We recommend the fat-suppressed spin-echo technique because the bone marrow edema is easily identified and the spin-echo technique is commonly available.
Address correspondence to E. Spuentrup.
Le Bihan D, Turner R, Douek P, Patronos N. Diffusion MR imaging: clinical applications. AJR 1992; 159:591-599
Le Bihan D. 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
Gass A, Gaa J, Sommer A, et al. Echo-planar diffusion-weighted MRI in the diagnostics of acute ischemic stroke: characterization of tissue abnormalities and limitations in the interpretation of imaging findings. (in German) Radiologe 1999; 39:695-702
Le Bihan D, Breton E, Lallemand D, Grenier P, Cabanis E, Lavall-Jeanet M. MR imaging of intravoxel inchoerent motions: application to diffusion and perfusion in neurologic disorders. Radiology 1986; 161:401-407
Le Bihan D, Basser PJ. Molecular diffusion and nuclear magnetic resonance. In: Le Bihan D, ed. Diffusion and perfusion magnetic resonance imaging: application to functional MRI. New York: Raven, 1995: 5-17
Buxton RB. Fast diffusion sensitive imaging with steady-state free precession. In: Le Bihan D, ed. Diffusion and perfusion magnetic resonance imaging: application to functional MRI. New York: Raven, 1995: 41-49
Baker LL, Goodman JB, Perkash I, et al. Benign versus pathologic compression fractures of vertebral bodies: assessment with conventional spin-echo, chemical shift, and STIR MR imaging. Radiology 1990; 174:495-502
Schiffenbauer YS, Tempel C, Abramovitch R, Meir G, Neeman M. Cyclocreatine accumulation leads to cellular swelling in C6 glioma multicellular spheroids: diffusion and one-dimensional chemical shift nuclear magnetic resonance microscopy. Cancer Res 1995; 55:153-158