Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved Garry E. Gold 1 Eric Han 2 Jeff Stainsby 3 Graham Wright 3 Jean rittain 2 Christopher eaulieu 1 Received November 2, 23; accepted after revision February 21, 24. 1 Department of Radiology, Stanford University, 3 Pasteur Dr., Grant ldg. S-68, Stanford, C 9435-515. ddress correspondence to G. E. Gold (gold@stanford.edu). 2 GE pplied Science Laboratory West, Menlo Park, C. 3 Department of Medical iophysics, University of Toronto, Toronto, ON, Canada. JR 24;183:343 351 361 83X/4/1832 343 merican Roentgen Ray Society Musculoskeletal MRI at 3. T: Relaxation Times and Image Contrast OJECTIVE. The purpose of our study was to measure relaxation times in musculoskeletal tissues at 1.5 and 3. T to optimize musculoskeletal MRI methods at 3. T. MTERILS ND METHODS. In the knees of five healthy volunteers, we measured the T1 and T2 relaxation times of cartilage, synovial fluid, muscle, marrow, and fat at 1.5 and 3. T. The T1 relaxation times were measured using a spiral Look-Locker sequence with eight samples along the T1 recovery curve. The T2 relaxation times were measured using a spiral T2 preparation sequence with six echoes. ccuracy and repeatability of the T1 and T2 measurement sequences were verified in phantoms. RESULTS. T1 relaxation times in cartilage, muscle, synovial fluid, marrow, and subcutaneous fat at 3. T were consistently higher than those measured at 1.5 T. Measured T2 relaxation times were reduced at 3. T compared with 1.5 T. Relaxation time measurements in vivo were verified using calculated and measured signal-to-noise results. Relaxation times were used to develop a high-resolution protocol for T2-weighted imaging of the knee at 3. T. CONCLUSION. MRI at 3. T can improve resolution and speed in musculoskeletal imaging; however, interactions between field strength and relaxation times need to be considered for optimal image contrast and signal-to-noise ratio. Scanning can be performed in shorter times at 3. T using single-average acquisitions. Efficient higher-resolution imaging at 3. T can be done by increasing the TR to account for increased T1 relaxation times and acquiring thinner slices than at 1.5 T. T he intrinsic signal-to-noise ratio (SNR) available in an MRI study is a function of the strength of the main magnetic field, the volume of tissue being imaged, and the radiofrequency coil being used. Most conventional MRI of the musculoskeletal system is done at 1.5 T. Systems with higher field strength, typically 3. T, are now becoming more prevalent in the clinical setting. ecause the available magnetization varies linearly with field strength, imaging at 3. T should provide twice the intrinsic SNR of imaging at 1.5 T if the coils and the subject are equivalent [1]. However, field-dependent changes in tissue relaxation times and in the chemical shift difference between fat and water may limit the SNR benefit seen at 3. T. The spin-spin relaxation time, T2, is fairly constant at different field strengths, decreasing slightly at higher field strength [2]. The spinlattice relaxation time, T1, increases as the field strength increases [2]. Prior measurements of relaxation times at 4. T showed increases in T1 relaxation time of 7 9% and decreases in T2 relaxation time of 1 2% compared with times at 1.5 T [3]. lthough relaxation times at 3. T also change compared with those at 1.5 T, in vivo values have not been available in the literature to our knowledge. The changes in these parameters affect the choice of TR and TE that are appropriate for 3. T, and ultimately affect the contrast and SNR of the images produced. t 3. T, the chosen TR and TE must reflect the underlying tissues being imaged and the contrast desired. Musculoskeletal imaging protocols typically consist of several 2D multisection scans, with or without fat saturation. t 3. T, because the T1 relaxation times have increased, the TR must be longer to maximize the SNR gain. t 3. T, TR must also be longer to achieve the same type of contrast on T1-weighted images as achieved with 1.5 T. Similarly, the TE should JR:183, ugust 24 343
Gold et al. Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved be slightly shorter to account for decreases in T2 relaxation times. The number of slices and the spatial resolution required may also influence the choice of TR and TE. We report the in vivo measurements of T1 and T2 for various musculoskeletal tissues at 3. T and illustrate the interactions of these parameters with the selection of TR and TE to achieve a given image contrast. Materials and Methods Subjects The right knees of five healthy volunteers (two men, three women; ages, 27 38 years) were imaged at both 1.5 T (Signa TwinSpeed, GE Healthcare) and 3. T (Signa VH/i, GE Healthcare). The local institutional review board first approved all MRI examinations. t 1.5 T, a four-channel transmit receive phased array knee coil was used (MRI Devices). t 3. T, a quadrature transmit receive knee coil was used (GE Healthcare). Sets of images at a single axial location containing muscle, cartilage, synovial fluid, marrow, and subcutaneous fat were acquired at different points on the T1 and T2 relaxation curves. T1 or T2 relaxation times of a given tissue were calculated by performing a monoexponential fit of the mean pixel intensities of a selected region of interest (ROI) at the different sampling times. In each subject, five measurements per tissue type were used to determine average relaxation times for each tissue. Phantom Validation The T1 and T2 relaxation time measurement techniques were validated using a phantom of known relaxation times. Seven phantom components of known relaxation times (Eurospin) were chosen to cover the expected range of relaxation times in musculoskeletal tissue. Five separate T1 and T2 relaxation time measurements were made at 1.5 and 3. T to determine the accuracy and repeatability of the measurement techniques. The protocols and coils used for the phantoms were the same as those used for the in vivo measurements. T1 Relaxation Time Measurement Protocol Images for the T1 measurements in our five volunteers were obtained using a Look-Locker method [4]. lpha pulses of 1 were used to acquire eight images representing eight equidistant samples along the T1 recovery curve. The sampling period was tailored to provide sampling times with adequate coverage of the T1 recovery curve for each tissue of interest (Table 1). cquisitions with the inversion pulse are subtracted from non inversion recovery steady-state acquisitions, yielding an understood asymptote of zero [5]. ecause the asymptote of the exponential is known, acquiring the last sample at a time approximately equal to the T1 relaxation time of the tissue is sufficient. TR of 5, msec was selected to provide adequate longitudinal recovery. TLE 1 Note. Longer TEs are used to sample tissues, such as synovial fluid, with relaxation times that are longer. T2 Relaxation Time Measurement Protocol Images for the T2 measurements in our five volunteers were obtained using a T2 preparation sequence [5 8]. The T2 preparation sequence consists of a 9 tip-down pulse, a train of equally spaced 18 pulses, and a ( 9 ) tip-up pulse. Different TEs are generated by varying the number of 18 pulses during the train of 18 pulses or by varying the space between those 18 pulses (Table 1). Images were acquired at six TEs with a TR of 3 sec. cquisition and Postprocessing TEs for Protocols for Measuring T1 and T2 Relaxation Times oth T1- and T2-weighted image sets were acquired using a spectral spatial excitation with spiral readout consisting of 4,96 points and 8 arms. Other imaging parameters included a field of view of 18 cm, slice thickness of 3 mm, and 4 averages for T1- weighted and 2 averages for T2-weighted. In-plane resolution was.5 mm. For fat and marrow measurements, the spectral spatial excitation was centered on the lipid resonance. Using a custom software tool, ROIs were placed on each tissue of interest and a monoexponential fit was calculated for each ROI [9]. ROIs were placed in the patella cartilage, lateral gastrocnemius muscle, femoral bone marrow, patellofemoral joint fluid, and medial subcutaneous fat; ROIs in the fluid and cartilage were relatively small because of the smaller volumes of tissue present. To preserve the integrity of the relaxation curve fit, points below or near the noise floor were discarded. Relaxation times were averaged across all subjects and the SDs of the measurements were calculated. Values for p were calculated using a paired Student s t test. TEs (msec) for T1-Weighted Imaging In Vivo Verification TEs (msec) for T2-Weighted Imaging Muscle, cartilage 5.4, 26, 46, 66, 86, 1,6, 1,26, 1,46 5.4, 17, 27.6, 38.3, 54.3, 12 Synovial fluid 5.4, 56, 1,6, 1,56, 2,6, 2,56, 3,6, 3,56 5.4, 53, 99.6, 193, 39, 774 Marrow, subcutaneous fat 5.4, 16, 26, 36, 56, 66, 76 5.4, 25, 43.6, 81, 166, 326 Two of the original five volunteers were imaged at 1.5 and at 3. T to verify T1 relaxation times of muscle and marrow fat. spin-echo sequence was used with a TE of 14 msec and TRs of 8, 2,, 4,, and 6, msec at 1.5 and 3. T. Pixel bandwidth was ± 16 khz; field of view, 16 cm; and matrix, 512 192. SNR values were measured from the images. Using the MRI signal equation, S = K (1 e TR / T1 ) e TE / T2 where K is a constant depending on the proton density), and the measured relaxation times, we compared the calculated signal levels based on the relaxation time results with the measured signal level (S) results for muscle and marrow fat. Calculated and measured values were normalized so that S = 1 at a TR of 6, msec. These measurements provided verification of the accuracy of T1 measurements made in all five volunteers using the T1 measurement protocol. One of the original five volunteers was imaged at both 1.5 and 3. T to verify SNR measurements and contrast-to-noise ratio (CNR) values. sagittal proton density weighted fast spin-echo sequence was used with a TR/minimum TE of 4,/14. coronal T1-weighted spin-echo sequence was used with a TR/minimum TE of 8/14. The imaging parameters were identical at each field strength. SNR was calculated as signal divided by the SD of the noise in each tissue in five locations. Significance of the SNR differences was determined using a paired Student s t test. CNR was calculated for cartilage and fluid at each TR for 1.5 and 3. T and compared. Comparison of 1.5-T Protocol for T2-Weighted Imaging of Knee with Two TLE 2 High-Resolution 3.-T Protocols That Take into ccount Changes in Relaxation Times Parameter 1.5 T 3. T, Protocol 3. T, Protocol Section thickness (mm) 3. 1.8 2.2 Skip (mm).5.3.3 Matrix 512 192 512 92 512 192 4, 8, 6, TE (msec) 52 48 48 Echo-train length 8 16 8 verages 3 3 2 Note. Longer TR and shorter TE at 3. T account for increased T1 and decreased T2 relaxation times. Scanning time for all sequences was 4 min 56 sec. t 3. T, using TE and TR above, we can decrease signal averages or increase echo-train length and get signal-to-noise ratio comparable to that at 1.5 T with higher resolution (thinner sections). 344 JR:183, ugust 24
Musculoskeletal MRI at 3. T Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved Protocol Design Example We used our relaxation time measurements to design a high-resolution T2-weighted imaging protocol for 3. T. The goals of our 3.-T protocol were to achieve a higher spatial resolution than that achieved with 1.5 T with comparable SNR of cartilage and muscle in a similar scanning time (Table 2). Receiver bandwidth was identical at both field strengths (± 16 khz). Matrix size at both field strengths was 512 192 to achieve a relatively high spatial resolution with a total scanning time of less than 5 min. To increase resolution at 3. T, the TR was increased to account for increases in T1 relaxation times, the TE was slightly decreased to account for decreases in T2 relaxation times, and the slice thickness was decreased. In one 3.-T protocol, the number of signal averages was decreased to keep imaging time constant, and in the other the echo-train length was increased. T2-weighted protocol was chosen because it is part of most knee imaging protocols and, as a low-snr sequence, it illustrates the SNR differences between 1.5 and 3. T. Results The phantom validation study shows that the methods for acquiring the T1 and T2 relaxation times were highly accurate and reproducible at both 1.5 and 3. T (Figs. 1 and 2, respectively). For the range of expected T1 and T2 relaxation times, our measurements were within the range of the phantom accuracy as reported by the manufacturer (± 3%). The SD of the average of our measurements is also small, indicating good reproducibility. Measured T1 Relaxation Time (msec) 1,5 1,25 1, 75 5 25 25 TLE 3 ctual T1 Relaxation Time (msec) 5 75 1, 1,25 1,5 In Vivo T1 Relaxation Times at 1.5 and 3. T Relaxation Times (msec) for T1-Weighted Imaging % Increase p a t 1.5 T t 3. T Muscle 1,13 ± 91.7 1,42 ± 38.1 2.4.2 Cartilage 1,6 ± 155 1,24 ± 17 14.5.4 Synovial fluid 2,85 ± 279 3,62 ± 32 2.1 Subcutaneous fat 288 ± 8.42 371 ± 7.94 22.3.1 Marrow fat 288 ± 5.27 365 ± 9. 21.1.1 Note. Relaxation times are average ± SD. a Differences between relaxation times at 1.5 and 3. T are significant. Measured T2 Relaxation Time (msec) 4 3 2 1 1 Our in vivo results show that T1 values increased when moving from 1.5 to 3. T (Table 3), and T2 values decreased slightly (Table 4). These results are statistically significant, as indicated by the calculated p values, and are consistent with results at other field strengths and using tissue samples [2, 3, 1]. The T2 relaxation time of synovial fluid decreased sharply from 1.5 to 3. T. To confirm the accuracy of the measured T1 relaxation times in vivo, we measured the signal levels of muscle and marrow fat in two volunteers using various TRs at 1.5 and 3. T (Figs. 3 and 4, respectively). The signal levels predicted from the MR signal equation using the measured relaxation times compared well with measured signal levels in muscle and fat. Measured signal levels in cartilage and synovial fluid showed high variability, probably because of individual variation or partial volume effects. Using the measured relaxation times and values of the proton density from the literature, we calculated the signal levels for various tissues at a fixed TE (14 msec) versus TR (Fig. 5). Signal levels were divided by the square root of the TR to fix total scanning time and were normalized to 1 for the maximum signal at 3. T. Values of proton density or ρ relative to water used for the calculated signal levels were taken from the literature and were as follows: synovial fluid, 1. [11, 12]; marrow fat and subcutaneous fat, 1. [13]; cartilage,.7 [3, 1]; and muscle,.6 [14 16]. Image contrast at a given TR is based on the signal difference between two tissues. Knee images obtained at 1.5 and 3. T in a healthy volunteer show the changes in contrast indicated by our calculations (Figs. 6 and 7). Measurements of the CNR show that the relative contrast between fluid and cartilage at a long TR of 4, msec is increased at 3. T compared with 1.5 T (CNR of 37.5 vs 16.2), which is predicted from the relaxation times. t a shorter TR of 8 msec, the 2 3 4 ctual T2 Relaxation Time (msec) Fig. 1. ccuracy and repeatability of measurements of T1 and T2 relaxation times at 1.5 T in a phantom (Eurospin). and, Solid lines indicate precision (± 3%) of relaxation times guaranteed by manufacturer. Diamonds are averages of five measurements of T1 () and T2 () relaxation times, and error bars are two times SD of those measurements. Note excellent accuracy and repeatability of T1 and T2 measurements at 1.5 T. JR:183, ugust 24 345
Gold et al. Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved TLE 4 In Vivo T2 Relaxation Times at 1.5 and 3. T Relaxation Times (msec) for T2-Weighted Imaging % Decrease p a t 1.5 T t 3. T Muscle 35.3 ± 3.85 31.7 ± 1.9 1.1.2 Cartilage 42.1 ± 7.5 36.9 ± 3.81 12.3.7 Synovial fluid 1,21 ± 14 767 ± 48.8 36.6.2 Subcutaneous fat 165 ± 5.5 133 ± 4.43 19.3.1 Marrow fat 165 ± 4.96 133 ± 6.14 19.3.1 Note. Relaxation times are average ± SD. a Differences between relaxation times at 1.5 and 3. T are significant. Measured T1 Relaxation Time (msec) 1,6 1,4 1,2 1, 8 6 4 2 2 4 6 8 1, 1,2 1,4 1,6 ctual T1 Relaxation Time (msec) Measured T2 Relaxation Time (msec) 4 3 2 1 cartilage-to-fluid CNR is also greater at 3. T than at 1.5 T (11.9 vs 4.8). SNR measurements also show the approximately twofold increase in SNR at 3. T, which accounts for much of the increase in contrast between fluid and cartilage at a long TR (Fig. 7). The SNR gain was statistically significant for both a TR of 4, msec and a TR of 8 msec (p <.2), although generally less than twofold at a TR of 8 msec. Images from the 1.5 to 3. T comparison (Fig. 8) show that comparable SNR can be ob- 1 2 3 4 ctual T2 Relaxation Time (msec) Fig. 2. ccuracy and repeatability of measurements of T1 and T2 relaxation times at 3. T in a phantom (Eurospin). and, Solid lines indicate precision (± 3%) of relaxation times guaranteed by manufacturer. Diamonds are averages of five measurements of T1 () and T2 () relaxation times, and error bars are two times SD of those measurements. Note excellent accuracy and repeatability of T1 and T2 measurements at 3. T. Normalized Signal-to-Noise Ratio 1..8.6.4.2. 1, 2, 3, 4, 5, 6, 7, Normalized Signal-to-Noise Ratio 1..8.6.4.2. 1, 2, 3, 4, 5, 6, 7, Fig. 3. Calculated and measured signal levels at 1.5 T based on measured T1 relaxation times. and, Measured ( ) and calculated (dotted lines) signal levels in muscle () and marrow fat () were normalized to 1 at TR of 6, msec. Note excellent agreement between predicted and calculated values, indicating measured relaxation times are accurate. 346 JR:183, ugust 24
Musculoskeletal MRI at 3. T Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved Normalized Signal-to-Noise Ratio 1..8.6.4.2.. 1, 2, 3, 4, 5, 6, 7, tained at higher resolution at 3. T with a TE and TR that take the changes in relaxation times into account. The SNR of cartilage at 1.5 T with a slice thickness of 3.5 mm was 1.4, and at 3. T with a section thickness of 1.8 mm was 11.5. Using a section thickness of 2.2 mm at 3. T, the cartilage SNR was 15.7. Use of a thin section (1.8 mm) at 1.5 T resulted in a low cartilage SNR of 3.8. The muscle SNR at 3. T, using a section thickness of 1.8 mm, was more than twice the muscle SNR at 1.5 T (8.5 vs 3.). The cartilage SNR was similarly increased at 3. T (11.5 vs 3.8), with Normalized Signal-to-Noise Ratio 1..8.6.4.2 more variability in the measurement. The greater-than-twofold increase here in SNR at 3. versus 1.5 T is caused by differences in the acquisition protocol and the increase in magnetization at 3. T. Discussion The phantom validation results show the accuracy of the T1 and T2 measurement techniques we used. The T1 measurement technique has been used extensively in the brain [17 2] but was modified in our study 1, 2, 3, 4, 5, 6, 7, Fig. 4. Calculated and measured signal levels at 3. T based on measured T1 relaxation times. and, Measured ( ) and calculated (dotted lines) signal levels in muscle () and marrow fat () were normalized to 1 at TR of 6, msec. Note excellent agreement between predicted and calculated values, indicating measured relaxation times are accurate. Signal.6.5.4.3.2.1. Fluid cartilage contrast (TR, 1,2) = Muscle Cartilage Fluid Fat Fluid cartilage contrast (TR, 4,) =.7. 1, 2, 3, 4, 5, 6, Signal 1.4 1..8.6.4.2 for the expected relaxation times in the musculoskeletal system. The T2 measurement technique has also been used to measure transverse relaxation times in blood, muscle, and the myocardium [7, 8, 21, 22]. The combination of these techniques with the spiral readout enables us to measure the shorter T2 relaxation times that are found in the musculoskeletal system. In our volunteers, measured T1 and T2 relaxation times were in the range of published values [2, 3, 1]. The sharp decrease of the measured T2 relaxation time for synovial Fluid cartilage contrast (TR, 1,6) = Muscle Cartilage Fluid Fat Fluid cartilage contrast (TR, 4,) =.12 1, 2, 3, 4, 5, 6, Fig. 5. Calculated tissue signal levels at fixed TE (14 msec) for 1.5 and 3. T using measured relaxation times. and, Signal levels at 1.5 T () and 3. T () were divided by square root of TR and plotted versus TR with maximum signal level at 3. T normalized to 1.. Difference between signal levels at any given TR determines contrast between tissues. Subcutaneous fat and bone marrow are plotted together as fat for simplicity. Note that at 3. T (), because of increased T1 relaxation times, TR of 1,6 msec is required to achieve same signal level (zero contrast) of cartilage and fluid, which requires TR of only 1,2 msec at 1.5 T (). t TR of 4, msec, fluid-to-cartilage contrast is greater at 3. T (.12 vs.7) because of increase in magnetization. JR:183, ugust 24 347
Fig. 6. Images of knee of healthy volunteer obtained at 1.5 and 3. T. and, Sagittal proton density weighted images at 1.5 T () and 3. T () with TR/TE of 4,/14. Increase in signal-to-noise ratio at 3. T is evident. Visually, image contrast appears similar; quantification is shown in Figure 7. rrow in indicates increased chemical shift at 3. T. C and D, Coronal T1-weighted images at 1.5 T (C) and 3. T (D) with TR/TE of 8/ 14. Image contrast also appears similar in these images, with cartilage having higher signal than fluid at short TR. Note increase in chemical shift at 3. T (arrow, D). C D 2 Signal-to-Noise Ratio (TR, 8) Signal-to-Noise Ratio (TR, 4,) Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved Gold et al. 18 16 14 12 1 8 6 4 2 Fluid Cartilage Muscle 2 18 16 14 12 1 8 6 4 2 Fluid Fat Cartilage Muscle Fat Fig. 7. Measurements of signal-to-noise ratio (SNR) in one healthy volunteer at both 1.5 (black bars) and 3. T (white bars) with TE of 14 msec. SNR differences between 1.5 and 3. T were statistically significant (asterisk indicates p <.2)., Difference of SNR values or contrast-to-noise ratio (CNR) for fluid and cartilage at TR of 4, msec is 16.2 at 1.5 T and 37.5 at 3. T., CNR for cartilage and fluid at TR of 8 msec is 4.8 at 1.5 T and 11.9 at 3. T, also showing increase in contrast at 3. T. 348 JR:183, ugust 24
Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved Musculoskeletal MRI at 3. T C Fig. 8. T2-weighted images in 28-year-old healthy woman volunteer at 1.5 and 3. T using protocols based on measured relaxation times., Sagittal T2-weighted image at 1.5 T with section thickness of 3.5 mm (Table 2) shows cartilage signal-to-noise ratio (SNR) is 1.4 and muscle SNR is 5.8., Sagittal T2-weighted image at 3. T using section thickness of 1.8 mm and parameters adjusted for relaxation times (Table 2) shows cartilage SNR is 11.5 and muscle SNR is 8.5. C, Sagittal T2-weighted image at 1.5 T using section thickness of 1.8 mm shows much lower overall image SNR. Cartilage SNR is 3.8 and muscle SNR is 3.. Muscle SNR at 1.5 T is slightly less than one-half SNR at 3. T using these parameters. D, Sagittal T2-weighted image at 3. T with section thickness of 2.2 mm and parameters adjusted for relaxation times (Table 2) shows cartilage SNR is 15.7 and muscle SNR is 9.5. D fluid was also observed at 4 T [3]. The T2 relaxation times of fat and marrow are longer than the values reported by Duewell et al. [3] but similar to the values reported by ottomley et al. [2]. Cartilage and synovial fluid relaxation times had relatively high SDs that may be the result of small ROIs, spatial or individual variations, or pathology [23 25]. Reported relaxation times in synovial fluid vary greatly [1 12]. We confirmed our measurements of T1 relaxation times by measuring signal levels in muscle and marrow fat in two volunteers. The predicted signal levels and the measured levels correlated well (Figs. 3 and 4), indicating that our T1 relaxation times for these tissues are accurate in vivo. The results of this correlation with measured values in cartilage and synovial fluid are not as accurate, perhaps because of individual variability or small ROIs. JR:183, ugust 24 Image contrast at 3. T depends on the relaxation times of the various musculoskeletal tissues and the SNR gain. For example, cartilage and fluid have the same signal (zero contrast) at a TR of approximately 1,6 msec at 3. T, compared with a TR of approximately 1,2 msec at 1.5 T (Fig. 5). This fact shows that contrast comparable to that of 1.5 T can be achieved with a longer TR at 3. T because of the increase in T1 relaxation times. The SNR gain at a short TR is also less than a factor of 2 because of the increase in T1 relaxation times, which was also found at 4. T [3]. Comparison of the maximum signal levels of fat at a short TR (Fig. 5) shows that the signal increase at 3. T will be about 1.7 times that at 1.5 T. To achieve a similar image contrast to that at 1.5 T and maximize the SNR gain at 3. T, the TR should be longer for T1-weighted imaging at 3. T. Image contrast between cartilage and fluid in proton density weighted or T2-weighted imaging may be improved at 3. T. Contrast between cartilage and fluid is increased at 3. T (Fig. 5), with the increase being greater at a longer TR because of the T1 recovery of the fluid. t a long TR, the twofold gain in SNR is achieved because of greater longitudinal relaxation. s with 4. T [3], we found no major qualitative differences in tissue contrast at 3. T compared with 1.5 T (Fig. 6). The impact of the decrease in T2 relaxation times at 3. T will be less significant for spin-echo imaging but may be more important for gradient echo imaging. We did not measure relaxation times in ligaments and menisci. These tissues generally have short T2 relaxation times at 1.5 T [26] and have little signal on routine MRI sequences. With the shorter T2 relaxation times seen at 3. T, normal menisci and liga- 349
Gold et al. Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved ments should have even less signal. Contrast between these tissues and synovial fluid will be important in the visualization of pathology at 3. T. On the basis of our results, contrast between fluid and tissues with little signal should be increased at 3. T compared with 1.5 T (Fig. 5). The resonance frequency at 3. T is twice that at 1.5 T, about 125 MHz, and the radiofrequency power for excitation at 3. T is four times higher than at 1.5 T [27, 28]. ecause the radiofrequency power deposited is a function of tissue volume excited, radiofrequency power deposition is more a problem with large body areas (hips) than with smaller areas (knees) [28]. Radiofrequency power deposition was carefully monitored during our study. Pulse sequences that are particularly affected by increased radiofrequency power deposition are spin-echo sequences at a short TR (T1-weighted) or fast spin-echo sequences with multiple 18 refocusing pulses. For this reason, manufacturers have begun to offer the option of fast spin echo with refocusing pulses of less than 18 [29]. These sequences are less SNR-efficient than conventional fast spin echo but offer less power deposition. Our study has three major limitations. First, we studied only healthy volunteers, so we do not have relaxation data on important pathology such as marrow edema. However, the imaging protocols designed for 3. T based on the relaxation measurements in our study have shown marrow edema in a few cases. Second, significant variation occurs in the T1 and T2 relaxation times of cartilage and synovial fluid in our data. However, the reported relaxation times of synovial fluid have considerable variability [1 12], and cartilage relaxation times are agedependent [12, 23]. Finally, we studied only a limited number of subjects for our relaxation time measurements, in vivo verification of T1 measurements, and SNR measurements. However, the relaxation measurements we obtained are within expected ranges from prior studies at other field strengths [3] and show significant changes from 1.5 to 3. T. Our data give us the needed starting point to design musculoskeletal imaging protocols at 3. T. Increased resolution may be helpful in several problem areas of musculoskeletal imaging [3]. On the basis of the measured changes in relaxation times, we designed two protocols for imaging at increased resolution at 3. T in the same imaging time as at 1.5 T (Table 2). In general, the principles behind the protocol design are as follows: First, decrease the TE to compensate for shorter T2 relaxation times; second, increase TR to compensate for longer T1 relaxation times; and third, decrease the number of signal averages or increases the echo-train length to keep the scanning time comparable. The images from these protocols (Fig. 8) show that the additional signal at 3. T allows the acquisition of more and thinner slices than 1.5 T at similar SNR and scanning times. Similar principles will allow the use of the additional signal at 3. T to increase imaging speed compared with 1.5 T, while keeping SNR and spatial resolution equivalent. One way to improve imaging speed is to reduce signal averages. ecause each additional signal average doubles the scanning time while providing an increase in SNR equal only to the square root of 2, the number of signal averages can be decreased by a factor of 2 4 at 3. T while maintaining comparable SNR to 1.5 T. In practice, other considerations, such as reducing phase wrap, may limit the reduction of signal averages. However, even with increasing the TR to account for T1 increases at 3. T, considerable saving in imaging time will be possible. Relaxation times and contrast between normal tissues are not the only factors that go into designing a clinical MRI protocol. Other factors such as contrast resolution between abnormal tissues, the effect of artifacts, and clinical throughput are all critically important in everyday practice. Contrast between fluid and tissues such as menisci and ligaments should increase at 3. T, but this hypothesis needs to be proven clinically. Increased chemical shift artifacts on non fat-saturated sequences may require the use of a higher receiver bandwidth at 3. T. The relaxation time measurements provided in this study give important information for designing clinical protocols at 3. T, but experience in the clinical setting is crucial to determine the best protocols for a given practice. References 1. Collins CM, Smith M. Signal-to-noise ratio and absorbed power as functions of main magnetic field strength, and definition of 9 degrees RF pulse for the head in the birdcage coil. Magn Reson Med 21;45:684 691 2. ottomley P, Foster TH, rgersinger RE, Pfeifer LM. review of normal tissue hydrogen NMR relaxation times and relaxation mechanisms from 1-1 MHz: dependence on tissue type, NMR frequency, temperature, species, excision, and age. Med Phys 1984;11:425 448 3. Duewell SH, Ceckler TL, Ong K, et al. Musculoskeletal MR imaging at 4 T and at 1.5 T: comparison of relaxation times and image contrast. Radiology 1995;196:551 555 4. Look DCL, Locker DR. Time saving in measurement of NMR and EPR relaxation times. Rev Sci Instrum 197;41:25 251 5. Wright G, Stainsby J. Preserving T1 or T2 contrast in magnetization preparation sequences. (abstr) In: Proceedings of the International Society for Magnetic Resonance in Medicine 4th annual meeting. New York, NY: ISMRM, 1996:1474 6. rittain JH, Hu S, Wright G, Meyer CH, Macovski, Nishimura DG. Coronary angiography with magnetization-prepared T2 contrast. Magn Reson Med 1995;33:689 696 7. Foltz WD, l-kwifi O, Sussman MS, Stainsby J, Wright G. Optimized spiral imaging for measurement of myocardial T2 relaxation. Magn Reson Med 23;49:189 197 8. Stainsby J, Wright G. Monitoring blood oxygen state in muscle microcirculation with transverse relaxation. Magn Reson Med 21;45:662 672 9. MacFall JR, Riederer SJ, Wang HZ. n analysis of noise propagation in computed T2, pseudodensity, and synthetic spin-echo images. Med Phys 1986;13:285 292 1. Henkelman RM, Stanisz GJ, Kim JK, ronskill MJ. nisotropy of NMR properties of tissues. Magn Reson Med 1994;32:592 61 11. Hardy P, Recht MP, Piraino D, Thomasson D. Optimization of a dual echo in the steady state (DESS) free-precession sequence for imaging cartilage. J Magn Reson Imaging 1996;6:329 335 12. Yao L, Sinha S, Seeger LL. MR imaging of joints: analytic optimization of GRE techniques at 1.5 T. JR 1992;158:339 345 13. Fernandez-Seara M, Song HK, Wehrli FW. Trabecular bone volume fraction mapping by low-resolution MRI. Magn Reson Med 21;46:13 113 14. Ehman RL, Kjos O, Hricak H, rasch RC, Higgins C. Relative intensity of abdominal organs in MR images. J Comput ssist Tomogr 1985;9: 315 319 15. Dunn JF, Zaim-Wadghiri Y. Quantitative magnetic resonance imaging of the mdx mouse model of Duchenne muscular dystrophy. Muscle Nerve 1999;22:1367 1371 16. Tsubahara, Chino N, Ishii H, kaboshi K, Takahashi H, Saitoh M. Objective parameters in magnetic resonance imaging (MRI) for neuromuscular diagnosis: preliminary findings. Disabil Rehabil 1992;14:163 167 17. Shah NJ, Zaitsev M, Steinhoff S, Zilles K. new method for fast multislice T(1) mapping. Neuroimage 21;14:1175 1185 18. Steinhoff S, Zaitsev M, Zilles K, Shah NJ. Fast T(1) mapping with volume coverage. Magn Reson Med 21;46:131 14 19. Karlsson M, Nordell. nalysis of the Look- Locker T(1) mapping sequence in dynamic contrast uptake studies: simulation and in vivo validation. Magn Reson Imaging 2;18:947 954 2. Jezzard P, Duewell S, alaban RS. MR relaxation times in human brain: measurement at 4 T. Radiology 1996;199:773 779 21. Stainsby J, Wright G. Partial volume effects on vascular T2 measurements. Magn Reson Med 1998;4:494 499 35 JR:183, ugust 24
Musculoskeletal MRI at 3. T Downloaded from www.ajronline.org by 46.3.25.67 on 12/5/17 from IP address 46.3.25.67. Copyright RRS. For personal use only; all rights reserved 22. Li KC, Wright G, Pelc LR, et al. Oxygen saturation of blood in the superior mesenteric vein: in vivo verification of MR imaging measurements in a canine model work in progress. Radiology 1995;194:321 325 23. Mosher TJ, Dardzinski J, Smith M. Human articular cartilage: influence of aging and early symptomatic degeneration on the spatial variation of T2 preliminary findings at 3 T. Radiology 2;214:259 266 24. Mosher TJ, Smith H, Dardzinski J, Schmithorst VJ, Smith M. MR imaging and T2 mapping of femoral cartilage: in vivo determination of the magic angle effect. JR 21;177:665 669 25. Smith HE, Mosher TJ, Dardzinski J, et al. Spatial variation in cartilage T2 of the knee. J Magn Reson Imaging 21;14:5 55 26. Gold GE, Pauly JM, Macovski, Herfkens RJ. MR spectroscopic imaging of collagen: tendons and knee menisci. Magn Reson Med 1995;34: 647 654 27. Shellock FG. Radiofrequency energy-induced heating during MR procedures: a review. J Magn Reson Imaging 2;12:3 36 28. rix G, Seebass M, Hellwig G, Griebel J. Estimation of heat transfer and temperature rise in partial-body regions during MR procedures: an analytical approach with respect to safety considerations. Magn Reson Imaging 22;2:65 76 29. Hennig J, Weigel M, Scheffler K. Multiecho sequences with variable refocusing flip angles: optimization of signal behavior using smooth transitions between pseudo steady states (TRPS). Magn Reson Med 23;49:527 535 3. Potter HG, Linklater JM, llen, Hannafin J, Haas S. Magnetic resonance imaging of articular cartilage in the knee: an evaluation with use of fast-spin-echo imaging. J one Joint Surg m 1998;8:1276 1284 Mark your calendar for the following RRS annual meetings: May 15 2, 25 New Orleans Hilton Riverside, New Orleans, L; pril 3 May 5, 26 Vancouver Convention and Exhibition Centre, Vancouver, C, Canada; May 6 11, 27 Grande Lakes Resort, Orlando, FL; pril 13 18, 28 Marriott Wardman Park Hotel, Washington, DC. JR:183, ugust 24 351