MRI of the Spine: Image Quality and Normal Neoplastic Bone Marrow Contrast at 3 T Versus 1.5 T

Transcription

1 Medical Physics and Informatics Original Research Zhao et al. 3-T Versus 1.5-T MRI of the Spine Medical Physics and Informatics Original Research FOCUS ON: Jian Zhao 1,2 Roland Krug 1 Duan Xu 1 Ying Lu 1 Thomas M. Link 1 Zhao J, Krug R, Xu D, Lu Y, Link TM Keywords: 1.5 T, 3 T, bone marrow, MRI techniques, spine DOI: /JR Received ugust 28, 2008; accepted after revision October 30, Department of Radiology, University of California, San Francisco, 400 Parnassus ve., -367, ox 0628, San Francisco, C ddress correspondence to T. M. Link (tmlink@radiology.ucsf.edu). 2 Department of Radiology, The Third Hospital of Hebei Medical University, Shijiazhuang, China. JR 2009; 192: X/09/ merican Roentgen Ray Society MRI of the Spine: Image Quality and Normal Neoplastic one Marrow Contrast at 3 T Versus 1.5 T OJECTIVE. The objectives of our study were to compare image quality of the spine and visualization of spine abnormalities at 3 T and 1.5 T as well as to evaluate differences in quantitative assessment of normal and neoplastic vertebral bone marrow. MTERILS ND METHODS. One hundred nine MR examinations of the spine were performed at 1.5 T and 3 T in the same patients within a time interval of less than 3 months. Visualization of anatomic and pathologic structures was analyzed by two radiologists. Normal and pathologic bone marrow was assessed on T1-weighted fast spin-echo (FSE) sequences. The signal intensity contrast of neoplastic bone marrow versus normal vertebral bone marrow was measured at 1.5 T versus 3 T. Sensitivity, specificity, and accuracy with 95% CIs were computed to assess the performance of muscle and disk as standards to differentiate between neoplastic and normal bone marrow on T1-weighted sequences at 1.5 T and 3 T. RESULTS. For all anatomic structures evaluated, image quality was rated significantly higher at 3 T than at 1.5 T, with 71.6% of the studies overall being superior at 3 T. The contrast between normal and pathologic bone marrow was significantly larger at 3 T (mean ± SD, 0.33 ± 0.13) than at 1.5 T (0.27 ± 0.11). The highest accuracy was found using muscle signal at 3 T to differentiate between normal and pathologic bone marrow. CONCLUSION. The use of 3-T MRI improves visualization of anatomic structures in the spine over 1.5-T MRI. s an internal standard on T1-weighted FSE images, skeletal muscle can be used to differentiate between infiltrative and normal bone marrow with higher accuracy at 3 T than at 1.5 T. T he introduction of 3-T MRI for more sophisticated clinical applications has yielded important benefits for imaging the musculoskeletal system [1 3]. The ability to increase the signal-to-noise ratio (SNR) and spatial resolution for musculoskeletal imaging has provided better visualization of anatomic and pathologic structures, including cartilage, bones, and ligaments [1, 4 8]. For the spine, better visualization of the nerve roots, facet joints, spinal cord, and disks and better characterization of the bone marrow using 3-T MRI could potentially improve diagnostic performance in detecting clinically relevant abnormalities [2, 9 11]. Field strength also affects the imaging of bone marrow because of increased susceptibility effects originating from the trabeculae at higher field strengths [8] that result in lower signal of overall bone marrow in non-fat-saturated sequences. Nondegenerated disk and skeletal musculature in T1-weighted fast spin- echo (FSE) sequences are accepted calibration standards that are used to differentiate between normal and abnormal bone marrow composition at 1.5 T [12]; however, whether the same standards can be applied for imaging at 3 T is not yet known. Therefore, the goals of this study were, first, to compare image quality of the spine using 1.5-T and 3-T MRI in the same patients; and, second, to evaluate differences in quantitatively assessing normal and neoplastic vertebral bone marrow at 1.5 T and 3 T. Materials and Methods Patients Sixty-four patients (29 males and 35 females) undergoing 3-T and 1.5-T MRI within a time interval of less than 3 months (average, 1.2 ± 0.60 months) were included in this study. The mean age (± SD) of the study group at the first examination was 42.9 ± 19.8 years, and the age range was years. total of 109 spine examinations were in cluded: 43 cervical, 23 thoracic, and 43 lumbar spine examinations. The JR:192, pril

2 Zhao et al. clinical indi cations for MRI of the spine included neoplastic lesions, such as metastases and multiple myeloma (n = 41); back pain (n = 48); degenerative disk disease (n = 21); and follow-up after spine surgery (n = 8). Some patients had more than one clinical indication for undergoing MRI. In 41 examinations, pathologic bone marrow changes were diagnosed that were confirmed by biopsy or clinical and imaging follow-up. The high percentage of neoplastic lesions re flected the fact that consecutive patients were selected in a time interval of less than 3 months. MRI scans were obtained in a consecutive fashion. None of the patients in the study group had metallic hardware; the policy at our institution is to avoid examination at 3 T of patients with large amounts of metal because of increased artifacts and potentially increased energy deposition in the hardware, which could lead to overheating and jeopardize patient safety. No trauma patients were included in the study group because abnormalities may have changed during the time period between the two examinations. None of the patients received radiation therapy during the period between the two examinations because radiation therapy may limit comparability of bone marrow signal changes. Patients with lesions of unknown histology and sclerotic metastases were excluded from the study. The size and body mass index of patients were not documented; however, given the fact that we examined the same patients at 1.5 T and 3 T in a relatively short time interval, this lack of data should not affect the results of this study. MRI Imaging was performed using 1.5-T and 3-T MR scanners (Excite, GE Healthcare). n 8-channel cervical thoracic lumbar phased-array coil was used at 3 T and an 8-channel phased-array coil at 1.5 T. Protocols were tailored for 3-T imaging by increasing the TR and decreasing the TE slightly as previously suggested by Gold et al. [13]. Cervical spine imaging at 3 T The following weighted FSE sequence with a TR/TE range of 640/12 18, a field of view (FOV) of 18 cm, and a slice thickness of 3 mm; second, a sagittal fatsaturated T2-weighted FSE sequence (TR/TE, 3,500/102; FOV, 18 cm; slice thickness, 3 mm); third, an axial T2-weighted FSE sequence (5,000/90; FOV, 16 cm; slice thickness, 4 mm); and fourth, an axial 3D gradient-echo sequence (gradient-recalled acquisition in the steady state [GRSS]; 35/15; FOV, 16 cm; flip angle, 5 ; slice thickness, 2 mm). Cervical spine imaging at 1.5 T The following weighted FSE sequence (TR/TE range, 600/12 18; FOV, 18 cm; slice thickness, 3 mm); second, a sagittal fat-saturated T2-weighted FSE sequence (TR/TE, 3,000/102; FOV, 18 cm; slice thickness, 3 mm); third, an axial T2-weighted FSE sequence (5,000/90; FOV, 16 cm; slice thickness, 4 mm); and fourth, an axial 3D GRSS sequence (35/15; FOV, 16 cm; flip angle, 5 ; slice thickness, 4 mm). Thoracic spine imaging at 3 T The following weighted FSE sequence (TR/TE range, 640/12 18; FOV, 32 cm; slice thickness, 3 mm); second, an axial T1-weighted FSE sequence (600/12 18; FOV, 18 cm; slice thickness, 4 mm); third, a sagittal fat-saturated T2-weighted FSE sequence (TR/TE, 3,200/90; FOV, 32 cm; slice thickness, 3 mm); and fourth, an axial T2-weighted FSE sequence (3,500/60; FOV, 16 cm; slice thickness, 4 mm). Thoracic spine imaging at 1.5 T The following weighted FSE sequence (TR/TE range, 600/12 18; FOV, 32 cm; slice thickness, 3 mm); second, an axial T1-weighted FSE sequence (500/12 18; FOV, 16 cm; slice thickness, 4 mm); third, a sagittal fat-saturated T2-weighted FSE sequence (TR/TE, 3,000/90; FOV, 32 cm; slice thickness, 3 mm); and fourth, an axial T2- weighted FSE sequence (3,500/60; FOV, 16 cm; slice thickness, 4 mm). Lumbar spine imaging at 3 T The following weighted FSE sequence (TR/TE range, 640/12 18; FOV, 24 cm; slice thickness, 4 mm); second, an axial T1-weighted FSE sequence (600/12 18; FOV, 16 cm; slice thickness, 4 mm); third, a sagittal fat-saturated T2-weighted FSE sequence (TR/TE, 4,200/60; FOV, 24 cm; slice thickness, 4 mm); and fourth, an axial T2-weighted FSE sequence (3,500/60; FOV, 16 cm; slice thickness, 4 mm). Lumbar spine imaging at 1.5 T The following weighted FSE sequence (TR/TE range, 600/12 18; FOV, 24 cm; slice thickness, 4 mm); second, an axial T1-weighted FSE sequence (500/12 18; FOV, 16 cm; slice thickness, 4 mm); third, a sagittal fat-saturated T2-weighted FSE sequence (TR/TE, 4,000/60; FOV, 24 cm; slice thickness, 4 mm); and fourth, an axial T2-weighted FSE sequence (3,500/60; FOV, 16 cm; slice thickness, 4 mm). The bandwidth varied according to field strength or the use of fat saturation. bandwidth of MHz was used at 3 T, with a higher bandwidth for non-fat-saturated images, and a bandwidth of MHz was chosen at 1.5 T. Non-fat-saturated T1-weighted and 3-T images had a higher bandwidth to account for increased chemical shift artifacts. ecause increasing bandwidth lowers SNR, band width was not increased above MHz. Qualitative nalysis ll images were evaluated by two board-certified radiologists with expertise in musculoskeletal MRI on PCS workstations. t the time of the analysis, the radiologists were blinded to the clinical history of and previous reports about the study subjects. During the reading session, ambient light was minimized and no time constraints were used. Radiologists reviewed the 3-T and 1.5-T images side-by-side, which were set up randomly, while blinded to all image information and sequence parameters. oth radiologists re viewed the images independently. Radiologists were asked to grade image quality using the following criteria: edge sharpness, amount of blurring, artifacts, and amount of noise. natomic structures including facet joints, endplates, nerve roots, spinal cord, and disks were evaluated separately. 4-level scale was used, in which 4 indicated optimal image quality. If one or two of the criteria listed above were not optimal, image quality was graded as 3; this, however, did not affect image evaluation. If analysis of the images was limited by the criteria listed above and diagnostic evaluation was affected, image quality was graded as 2. If diagnosis was substantially limited and images were nondiagnostic, image quality was graded as 1. Finally, the radiologists were asked to assess the overall image quality of the 3-T images compared with the 1.5-T images. The image quality of both studies was compared and after unblinding was scored as superior if 3-T MR image quality overall was better than 1.5 T, as the same if 3-T MR image quality was similar to 1.5-T MR image quality, or as inferior if 3-T MR image quality was inferior to 1.5-T image quality. Clinical ssessment The MR examinations were compared con cerning visualization of clinically relevant abnormal findings, such as synovial cysts at the facet joints and nerve root compression, that were detected on only one of the studies. bnormal findings were documented and compared concerning the absence or presence of visualization on the 3-T and 1.5-T images. Great care was taken to avoid including abnormalities in this analysis that may have developed during the time interval between the studies; potentially new lesions were excluded from the analysis. In addition, radiology reports were analyzed concerning the diagnosis of findings detected in the 3-T examinations but not in the 1.5-T examinations. Quantitative nalysis The signal intensity of normal bone marrow and of neoplastic vertebral bone marrow infil tration was measured on T1-weighted sagittal images with regions of interest (ROIs) of the same size placed in the same location on 1.5-T and 3-T MR images. In normal bone marrow, rectangular ROIs were placed in the superior part of the vertebral body to avoid areas with signal abnormality below the endplate. The size of the ROI was adjusted according to the size of the vertebral 874 JR:192, pril 2009

3 3-T Versus 1.5-T MRI of the Spine Fig. 1 Sagittal T1-weighted MR image obtained at 3 T in 27-year-old woman shows placement of rectangular regions of interest (ROIs) in superior part of normal vertebral bodies (L1 L5). Note that we carefully avoided including chemical shift artifact at endplate (arrows) in ROI. body as shown in Figure 1. In abnormal bone marrow, the size of the ROI was adjusted according to the size of the lesion and all ROIs were located in the center of the lesions as shown in Figure 2. The average values for each examination were calculated using ROIs from all the vertebral bodies assessed. Homogeneous bone marrow with signal intensity on T1-weighted images higher than disk and muscle and with no abnormal signal changes on fat-saturated T2-weighted studies was defined as normal. In patients with susp ected malignancy, biopsy, imaging, and clinical follow-up were available. Contrast (C) between lesions and normal bone marrow was calculated according to the following formula: C = (SI nbm SI bmi ) / (SI nbm + SI bmi ), where SI nbm is the signal intensity of normal bone marrow and SI bmi indicates the signal intensity of bone marrow infiltration. This so-called Michelson contrast formula has been used for patterns in which both bright and dark features are equivalent and take up similar fractions of the area [14]. It has also been used for comparing 1.5-T and 3-T MRI in a previous study [15]. Signal intensity was also measured in nondegenerated intervertebral disks and skeletal muscles in sagittal T1-weighted images. ROIs were placed at the center of the psoas muscles with 1-cm diameter at the lumbar spine (Fig. 3) within the rhomboideus major muscles at the thoracic spine and within the semispinalis colli and semispinalis capitis muscles at the cervical spine. ll degenerated intervertebral disks were excluded from the analysis. Nondegenerated disks were selected on sagittal fatsaturated T2-weighted images on the basis of normal signal intensity and height. ROIs were drawn manually according to the shape of the non degenerated disks in T1-weighted images (Fig. 4). Quantitative signal intensity differences between muscle and disk versus the average signal intensity of normal bone marrow and diseased bone marrow were computed separately to assess the performance of each as a calibration standard by which to differentiate normal and abnormal bone marrow composition. Statistical nalysis Differences in image quality measurements of all evaluated anatomic structures between 3 T and 1.5 T were compared using Wilcoxon s signed-rank test with a significance threshold of p < The Fig. 2 Sagittal T1-weighted fast spin-echo images obtained at 3 T. and, 55-year-old woman with metastatic melanoma () and 36-year-old woman with metastatic breast cancer (). oth metastases were proven by vertebral bone biopsy. Regions of interest (octagon, ; squares, ) were placed in center of lesion carefully to avoid partial volume effects. McNemar test was used to assess differences in direct comparisons of image quality. Contrast differences of pathologic and normal bone marrow between 3-T and 1.5-T images were assessed using paired Student s t tests with a significance thresh old of p < Sensitivity, specificity, and accuracy with 95% CIs were calculated to estimate the diagnostic performance of muscle or disk as standards on T1-weighted FSE sequences for detecting diseased bone marrow. ll statistical computations were processed using statistics software (JMP, version 7, SS Institute). Results Image Quality Using all available imaging sequences, both radiologists scored the quality of the 3-T MR images to be significantly higher overall (p < ) than of those obtained at 1.5 T and for each anatomic structure evaluated at the spine (Table 1). Significantly higher image quality scores at 3 T were also found at the cervical, thoracic, and lumbar spine when computed separately. For all structures, an average score of 3.51 ± 0.62 (SD) was calculated for the 3-T images, whereas a corresponding score of 2.79 ± 1.23 JR:192, pril

4 Zhao et al. Fig. 3 Sagittal T1-weighted MR images of cervical, thoracic, and lumbar spine in 40-year-old man (), 35-year-old woman (), and 55-year-old woman (C) obtained at 3 T show muscle tissue with region of interest (ROI) used as standard of reference. C, ROIs at cervical spine were placed within semispinalis colli and semispinalis capitis muscles (circle, ), at thoracic spine within rhomboideus muscles (circle, ), and at lumbar spine at center of iliopsoas muscles (circle, C). reas with fatty infiltration of muscles were carefully avoided. Fig. 4 Region-of-interest (ROI) placement in normal disks. and, Sagittal fat-saturated T2-weighted () and T1-weighted () fast spin-echo (FSE) MR images in 47-year-old woman with left leg pain. ased on findings in, nondegenerated intervertebral disks were identified and ROIs were placed in (ovals). Ovalshaped ROIs were placed manually in T1-weighted FSE images according to shape of nondegenerated disks. C 876 JR:192, pril 2009

5 3-T Versus 1.5-T MRI of the Spine TLE 1: Image Quality at 3 T and 1.5 T Image Quality Score (mean ± SD) Cervical Spine Thoracic Spine Lumbar Spine Complete Spine Structure 3 T 1.5 T 3 T 1.5 T 3 T 1.5 T 3 T 1.5 T Disk 3.58 ± ± ± ± ± ± ± ± 0.53 Nerve root 3.47 ± ± ± ± ± ± ± ± 0.64 Endplate 3.53 ± ± ± ± ± ± ± ± 0.63 Facet joint 3.33 ± ± ± ± ± ± ± ± 0.64 Spinal cord 3.60 ± ± ± ± ± ± ± ± 0.55 Note n excellent score is defined as 4 and a poor score as 1. Using Wilcoxon s signed-rank test, we found that differences for all structures were statistically significant (p < 0.05). TLE 2: Comparison of Image Quality at 3 T Versus 1.5 T for Cervical, Thoracic, and Lumbar Spine Separately and for the Complete Spine Image Quality Superior image quality at 3 T compared with 1.5 T Same image quality at 3 T and 1.5 T Inferior image quality at 3 T compared with 1.5 T was calculated for the 1.5-T images. Visualization of the endplate showed the highest average score of 3.59 ± 0.56 at 3 T when compared with the other anatomic structures. The overall image quality score was superior at 3 T compared with 1.5 T in 71.6% of the examinations, the same in 23.9%, and inferior in 4.6% (p < ) (Table 2). Figures 5 7 show representative examples of the higher image quality obtained at 3 T versus 1.5 T, with a larger amount of image noise and indistinctness of the structural edges noted at 1.5 T. Decreased visualization of nerve roots and of facet joints in particular No. (%) of Studies Cervical Spine a Thoracic Spine b Lumbar Spine a Complete Spine a 33 (76.7) 17 (73.9) 28 (65.1) 78 (71.6) 8 (18.6) 5 (21.7) 13 (30.2) 26 (23.9) 2 (4.7) 1 (4.3) 2 (4.7) 5 (4.6) Overall a p < , McNemar test. b p = , McNemar test. Fig year-old woman with low back pain associated with degenerative disk disease and degenerative disease of facet joints. and, xial T2-weighted fast spin-echo images obtained at 1.5 T () and 3 T (). Note differences in image quality between two images. Facet joints (black arrow) are shown in both images but are better delineated in 3-T image. Nerve roots (white arrow) are well visualized in 3-T image but not as well in 1.5-T image. Note, however, also that slice position is not entirely identical and that chemical shift artifacts appear more prominent at 3 T. is noted at 1.5 T (Fig. 5). The largest difference between individual scores was found at the endplate (Fig. 6). In 14 of the 109 examinations, additional clinical findings were detected on 3-T images when compared with 1.5-T images (Fig. 7); these additional findings included facet joint degeneration (n = 5), ligamentum flavum hypertrophy (n = 3), narrowing of the neural foramina (n = 3), pathology of the spinal cord (n = 1), and a synovial cyst at the facet joints (n = 2). In only one examination, an additional clinical finding that is, facet hypertrophy was detected at 1.5 T compared with 3 T. None of the additional findings affected the therapeutic management of these patients. Quantitative nalysis Of the 109 examinations, neoplastic vertebral bone marrow was found in 41 examinations. The image contrast calculated with the Michelson contrast formula between normal and neoplastic bone marrow at 3 T was significantly higher than that at 1.5 T (0.33 ± 0.13 vs 0.27 ± 0.11, respectively; p < 0.05). Significant differences were also found in differentiating between normal and pathologic bone marrow of the cervical, thoracic, JR:192, pril

6 Zhao et al. Fig year-old woman with low back pain. and, Sagittal T1-weighted fast spin-echo images obtained at 1.5 T () and 3 T (). Endplate is better delineated at 3 T than at 1.5 T, with better contrast between disk and vertebral body. lso note that chemical shift artifacts at these regions are more pronounced at 3 T than at 1.5 T. and lumbar spine separately at 3 T and 1.5 T. Using the signal of muscle in relation to that of bone marrow on T1-weighted FSE sequences as a standard to differentiate between neoplastic and normal bone marrow, the sensitivity, specificity, and accuracy were 85.4%, 92.7%, and 89.0% at 3 T, respectively, whereas these performance values were 80.5%, 80.5%, and 80.5% at 1.5 T. Using nondegenerative disk as a standard showed mostly lower values: The sensitivity, specificity, and accuracy were 59.4%, 96.9%, and 78.1%, respectively, at 3 T and 62.5%, 93.8%, and 78.1% at 1.5 T (Table 3). Figure 8 shows a metastatic bone marrow lesion in the lumbar spine that was visualized with higher contrast at 3 T than at 1.5 T. Discussion In this study, we were able to show that 3-T MRI provides better visualization of anatomic structures at the spine compared with 1.5 T and that additional imaging findings were seen at 3 T in 14 of 109 examinations. y quantitatively comparing muscle and intervertebral disk as calibration standards to differentiate neoplastic and normal bone marrow in T1-weighted images, we observed the highest diagnostic accuracy using the signal intensity of muscle at 3 T. The higher SNR that is available with 3 T provides the potential to increase diagnostic accuracy through improved image quality. In previous studies, investigators have compared 3-T versus 1.5-T MRI in evaluating different musculoskeletal regions: Saupe et al. [16, 17] showed a trend toward improved visualization of articular cartilage abnormalities of the wrist at 3 T with higher interobserver agreement but no significant increase in diagnostic performance. Comparing MR images of human calcaneus specimens obtained at 3-T and 1.5-T MRI, Phan et al. [8] found significantly higher SNR and better correlations with micro-ct derived parameters at 3 T than at 1.5 T. High-field-strength Fig year-old man with left leg pain. and, xial T2-weighted fast spin-echo images obtained at 1.5 T () and 3 T (). Note differences in image quality and noise between two images. Facet joints are better depicted at 3 T. lthough degenerative disease of left facet joint with small synovial cyst (arrow, ) was listed in report of 3-T study, it was not mentioned in report of 1.5-T study. 878 JR:192, pril 2009

7 3-T Versus 1.5-T MRI of the Spine TLE 3: Sensitivities, Specificities, and ccuracies with 95% CIs for Using Muscle and Intervertebral Disk as Standards to Differentiate Normal and Pathologic one Marrow Performance Measure MRI at 3 T was also shown to allow improved visualization of cartilage and focal cartilage lesions. The results of an in vitro study [4] showed that 3-T MRI with optimized high-resolution sequences had significantly improved diagnostic performance in assessing focal cartilage lesions compared with 1.5-T MRI. n additional in vitro study performed at the ankle showed similar results [3]. However, to the best of our knowledge no studies to date have compared clinical 1.5-T versus 3-T imaging in particular concerning diagnostic outcome. Our study performed at the spine interestingly showed additional imaging findings at 3 T in only 14 of 109 examinations and none of these findings would have altered the clinical management of the patients. Image contrast at 3 T depends on the relaxation times of the various musculoskeletal tissues and the SNR gain. ecause relaxation times at 1.5 T and 3 T are different, acquisition parameters must be adapted to obtain the best signal and contrast between normal and diseased tissues. Thus, our parameters were Muscle Intervertebral Disk 3 T 1.5 T 3 T 1.5 T Sensitivity, % (95% CI) 85.4 ( ) 80.5 ( ) 59.4 ( ) 62.5 ( ) Specificity, % (95% CI) 92.7 ( ) 80.5 ( ) 96.9 ( ) 93.8 ( ) ccuracy, % (95% CI) 89.0 ( ) 80.5 ( ) 78.1 ( ) 78.1 ( ) Fig year-old woman with metastatic melanoma. and, Sagittal T1-weighted fast spin-echo images obtained at 1.5 T () and 3 T (). Metastatic lesion in L3 vertebral body is better delineated and visualized with higher contrast at 3 T compared with 1.5 T. modified slightly according to a previously published study by Gold et al. [13] that is, the TR was increased slightly in the FSE sequences. Using these modified parameters in our study, we found that the contrast between normal and infiltrative bone marrow in all 41 examinations with pathologic bone marrow was significantly higher at 3 T than at 1.5 T. Carroll et al. [12] previously proposed two potential internal standards for detecting diseased bone marrow on T1-weighted MR images obtained at 1.5 T. The signal intensity of diseased and normal bone marrow relative to adjacent skeletal muscle or nondegenerated intervertebral disk in T1-weighted images was investigated in 71 cases. They found that skeletal muscle and nondegenerated intervertebral disk were accurate and statistically significant internal standards for differentiating between normal hematopoietic marrow and infiltrative marrow pathology on T1- weighted images, with an accuracy of 94% for muscle and 98% for disk. The results of their study suggest that an alteration in marrow signal intensity that is less than or equal to skeletal muscle or intervertebral disk on T1-weighted spin-echo MRl at 1.5 T should not be attributed to normal hematopoietic bone marrow. Our results are in accordance with those of Carroll and colleagues at 1.5 T and 3 T, although the performance of these standards was slightly lower in our study than in theirs. In our study, using muscle as a calibration standard at 3 T showed the highest accuracy value of 89.0% versus 80.5% at 1.5 T, whereas accuracies of the disk were 78.1% at both 3 T and 1.5 T. We therefore conclude that these standards can also be used at 3 T. These findings are important in particular because increased susceptibility effects at 3 T caused by trabecular bone induce signal loss in normal bone marrow. These internal standards are useful and can serve as simple tools to guide the practicing radiologist in differentiating normal and abnormal bone marrow on MRI. Specific absorption rate (SR) describes the energy that is deposited by a radiofrequency JR:192, pril

8 Zhao et al. field in a patient per unit of mass or weight and is increased by a factor of 4 at 3 T compared with 1.5 T. Given the fact that spine imaging includes large volumes, SR limits may be exceeded at 3 T. This factor must be considered as a caveat with 3 T, and steps to minimize SR such as reducing the acquisition flip angle, increasing the TR, and decreasing the number of phase-encoding steps in conjunction with parallel imaging must be considered. In addition, artifacts are more pronounced at 3 T than at 1.5 T [18 20]. lthough the bandwidth was adjusted, we noted increased chemical shift artifacts at 3 T in some studies compared with studies at 1.5 T that may have potentially unblinded the radiologists grading of the images and may have introduced a bias. The differences in the evaluation of the endplates between 1.5 T and 3 T may in part be due to increased chemical shift artifacts, which enhance the vertebral body contour. This is a potential limitation of the analysis. Increased susceptibility may also affect image quality at 3 T and may be addressed with an increase in bandwidth. Given that no patients with metal hardware were examined in our study, this issue did not substantially affect the results of this study. However, occasionally we thought that in T1- weighted images the bone marrow signal appeared to be lower at 3 T than at 1.5 T because of susceptibility artifacts from trabecular bone. This observation originally motivated this study. Given the increase in SNR at the same time, these findings could not be verified using quantitative measurements. To improve image quality at 3 T, Lin et al. [21] suggested maximizing echo-train length, decreasing slice thickness and TE, and increasing receiver bandwidth to reduce susceptibility artifacts. For patients with spine hardware, orienting the frequency-encoding direction parallel to the long axis of metal to diminish susceptibility artifacts has also been suggested. In conclusion, we found that 3-T MRI improves visualization of anatomic structures over 1.5-T MRI. Using a quantitative approach, skeletal muscle and nondegenerative disk can still be used as internal reference standards at 3 T to differentiate normal and neoplastic bone marrow, with superior results found for muscle at 3 T compared with 1.5 T. cknowledgments We thank Shoujun Zhao and Caixia Li for their help with the statistical analysis of the data in this manuscript. References 1. Schibany N, a-ssalamah, Marlovits S, et al. Impact of high field (3.0 T) magnetic resonance imaging on diagnosis of osteochondral defects in the ankle joint. Eur J Radiol 2005; 55: Shapiro MD. MR imaging of the spine at 3T. Magn Reson Imaging Clin N m 2006; 14: arr C, auer JS, Malfair D, et al. MR imaging of the ankle at 3 Tesla and 1.5 Tesla: protocol optimization and application to cartilage, ligament and tendon pathology in cadaver specimens. Eur Radiol 2007; 17: Link TM, Sell C, Masi JN, et al. 3.0 vs 1.5 T MRI in the detection of focal cartilage pathology: ROC analysis in an experimental model. Osteoarthritis Cartilage 2006; 14: Masi JN, Sell C, Phan C, et al. Cartilage MR imaging at 3.0 versus that at 1.5 T: preliminary results in a porcine model. Radiology 2005; 236: owers ME, Tung G, Trinh N, et al. Effects of CL interference screws on articular cartilage volume and thickness measurements with 1.5 T and 3 T MRI. Osteoarthritis Cartilage 2008; 16: Ludescher, Martirosian P, Lenk S, et al. Highresolution magnetic resonance imaging of trabecular bone in the wrist at 3 Tesla: initial results. cta Radiol 2005; 46: Phan CM, Matsuura M, auer JS, et al. Trabecular bone structure of the calcaneus: comparison of MR imaging at 3.0 and 1.5 T with micro-ct as the standard of reference. Radiology 2006; 239: Liney GP, ernard CP, Manton DJ, Turnbull LW, Langton CM. ge, gender, and skeletal variation in bone marrow composition: a preliminary study at 3.0 Tesla. J Magn Reson Imaging 2007; 26: Rodegerdts E, oss, Riemarzik K, et al. 3D imaging of the whole spine at 3T compared to 1.5T: initial experiences. cta Radiol 2006; 47: Phalke VV, Gujar S, Quint DJ. Comparison of 3.0 T versus 1.5 T MR: imaging of the spine. Neuroimaging Clin N m 2006; 16: , ix 12. Carroll KW, Feller JF, Tirman PF. Useful internal standards for distinguishing infiltrative marrow pathology from hematopoietic marrow at MRI. J Magn Reson Imaging 1997; 7: Gold GE, Han E, Stainsby J, Wright G, rittain J, eaulieu C. Musculoskeletal MRI at 3.0 T: relaxation times and image contrast. JR 2004; 183: Michelson, Harvey. Plotnick collection of the history of quantum mechanics and the theory of relativity. In: Michelson. Studies in optics. Chicago, IL: The University of Chicago Press, 1927:ix, Morakkabati-Spitz N, Gieseke J, Kuhl C, et al. 3.0-T high-field magnetic resonance imaging of the female pelvis: preliminary experiences. Eur Radiol 2005; 15: Saupe N, Prussmann KP, Luechinger R, osiger P, Marincek, Weishaupt D. MR imaging of the wrist: comparison between 1.5- and 3-T MR imaging preliminary experience. Radiology 2005; 234: Saupe N, Pfirrmann CW, Schmid MR, Schertler T, Manestar M, Weishaupt D. MR imaging of cartilage in cadaveric wrists: comparison between imaging at 1.5 and 3.0 T and gross pathologic inspection. Radiology 2007; 243: Eckstein F, Charles HC, uck RJ, et al. ccuracy and precision of quantitative assessment of cartilage morphology by magnetic resonance imaging at 3.0T. rthritis Rheum 2005; 52: Meyer JS. Musculoskeletal imaging at 3 T: imaging and optimization. Pediatr Radiol 2008; 38[suppl 2]: Schmidt GP, Wintersperger, Graser, aur- Melnyk, Reiser MF, Schoenberg SO. High-resolution whole-body magnetic resonance imaging applications at 1.5 and 3 Tesla: a comparative study. Invest Radiol 2007; 42: Lin W, n H, Chen Y, et al. Practical consideration for 3T imaging. Magn Reson Imaging Clin N m 2003; 11: , vi 880 JR:192, pril 2009