|Home | About | Journals | Submit | Contact Us | Français|
To implement and examine the feasibility of a 3D ultra-short TE (UTE) sequence on a 7T clinical MR scanner in comparison with 3T MRI at high isotropic resolution.
Using an in-house built saddle coil at both field strengths we have imaged mid-diaphysial sections of five fresh cadaveric specimen of the distal tibia. An additional in vivo scan was performed at 7 Tesla using a quadrature knee coil.
Using the same type of saddle coil at both field strengths a significant increase in SNR at 7T compared to 3T (factor 1.7) was found. Significantly shorter T2* values were found at the higher field strength (T2*=552.2±126µs at 7T versus T2*=1163±391µs at 3T).
UHF MRI at 7T has great potential for imaging tissues with short T2.
Ultra-high field (UHF) magnetic resonance imaging (MRI) at 7 Tesla has recently seen increased interest in the MR community as clinical scanners become more widely available. Since the signal to noise ratio (SNR) is expected to increase linearly with field strength, an enhancement in image quality is expected compared to scanners operating at lower field strengths. This gain in SNR can be used to either improve spatial image resolution or cut down the scan time. The utility of UHF MRI for imaging of the musculoskeletal system has been demonstrated in numerous publications (1–6). Usually, only soft tissues with transverse (T2) relaxation times greater than a few milliseconds are visualized with conventional MRI at all field strengths. The reason for this lies in constraints regarding the minimum echo time possible. These limitations are due to the need for prephasing gradient lobes for conventional Cartesian acquisitions as well as the duration of the radiofrequency (RF) pulses and required slice-selection rephasing gradients. Thus, conventional imaging methods show little to no signal from tissues with shorter T2 relaxation times and these tissues are only observed by negative contrast, i.e. by looking at the absence of signal or other indirect methods. The basis of T2 relaxation is the interaction between neighboring spins known as dipolar (or spin-spin) coupling. If the spins are in a mobile environment, motion (or tumbling) of the spins reduces this coupling and increases T2. In a more structured environment, the coupling is stronger, resulting in a shorter T2. Thus tissues that are more solid and structured, such as tendons, ligaments, menisci, cortical and trabecular bone, have shorter T2 relaxation times as compared to other tissues.
Imaging of these solid and semi-solid tissues would be of great value to further enhance the diagnostic capabilities of MRI. As previously described in great detail (7), the proton NMR signal of cortical bone exhibits a broad distribution of transverse relaxation components attributable to several bone proton sources. An ultra short component (~12µs) from collagen backbone methylene protons (80%) and a small percentage form NWEPs (amide/hydroxide) and adsorbed mineral water. An ultra short component (~60µs) derived predominantly from collagen side-chain or otherwise mobile methylene protons and a longer component (400µs) derived predominantly from water bound to the collagen. Finally, a T2 component in the order of some milliseconds derived from pore water (60%) and from lipid methylene protons (40%). Thus, as pointed out by Horch et al (7), modern ultrashort echo time MRI of cortical bone is dominated by signal from water bound to bone matrix collagen.
By employing suitable MR imaging techniques such as radial readouts in conjunction with short-duration RF pulses the echo time (TE) can be considerably shortened (8–10). Previously, ultra-short TE (UTE) of cortical bone was performed at 1.5T (11) and more recently at 3T (12,13). In this study, we investigated the feasibility to use a 3D UTE sequence on a 7T human MR scanner. We have implemented the sequence on both a 7T and a 3T MR system. In order to conduct a fair comparison, we have also built two small identical RF coils for both field strengths. Comparisons of SNR and T2* between different field strengths using the same hardware have not been performed yet. Furthermore, T2* maps of cortical bone have not been published previously. In this work, we hypothesized that UHF-MRI at 7T will provide a significant increase in SNR and that T2* relaxation time of solid cortical bone might differ between field strengths. Furthermore, we will discuss limitations and encountered problems of 3D UTE at higher field strengths.
Mid-diaphysial sections of five fresh cadaveric tibiae stored at −80 C were used for UTE imaging which was performed on a 7T and 3T Signa MRI systems (General Electric, Milwaukee, WI) with gradient systems capable of a maximum amplitude of 40 mT/m and a maximum slew rate of 150 mT/m/msec. Transmit gain and shimming were manually adjusted. To operate at high frequency with high efficiency, two coaxial transmission-line saddle RF coils were built for each field strength. The coil design is based on the transmission line technique that provides distributed circuits and reduces the phase variation along the coil conductors at high frequencies. This method also alleviates problems related to coil efficiency degradation usually encountered at high and ultra-high field strengths. The two coils had the same size of 57mm in diameter and 57mm in length for a meaningful performance comparison at the two field strengths. Both coils were well matched to the system’s 50 Ohm when loaded with water phantoms and specimens. The loaded Q factors measured were similar for both coils (78 at 3T and 72 at 7T). B1 mapping through transmission coefficient S21 measurement indicated a less than 0.5 dB variation of the B1 field within a 30mm-diameter imaging area for both coils.
For imaging, a three-dimensional UTE pulse sequence (14) as depicted in Figure 1 was implemented on both systems. In order to shorten the TE and acquisition window, UTE uses radial trajectories that begin at the center of k-space. With this technique, the echo time is considerably shortened and there is no need for any prephasing gradients, which would increase the minimum TE. Center-out radial trajectories achieve the best possible resolution in the shortest acquisition window. Furthermore, data is acquired on the gradient ramp to keep a minimal TE. The data was reconstructed in MATLAB (Matlab 7.4, MathWorks Inc.) using the gridding method (15) with a radially quadratic density compensation function adjusted near the center of k-space for the ramp sampling. For 3D UTE imaging excitation, a constant amplitude RF pulse can be applied. This rectangular pulse excites the entire imaging volume and thus requires no slice-selection or refocusing gradients. Furthermore, a hard pulse at the maximum amplitude will maximally excite short-T2 components. Applying these techniques, a minimum TE, as measured from the end of the RF pulse to the start of the acquisition, of 64µs was achieved at both field strengths. On our system, this was limited by the hardware switching time from transmit to receive mode. The number of projections used was 11251 and 300 samples per projection were acquired in 2µsec per sample. Further imaging parameters included: TR = 20ms, 1.0mm isotropic resolution, 6×6×6 cm FOV, excitation angle α=15°, hard pulse duration = 82µs, 2µs sampling interval, and 600µs readout duration. Two excitations were averaged and combined to create the final image.
The cortical bone was manually segmented in order to determine regions of interest for subsequent data analysis. The mean signal intensity was computed for each echo time TE. Noise was determined from regions with no signal. Signal to noise was measured as the mean intensity value of the shortest TE divided by the standard deviation of the noise. A least-squares fit was performed at each voxel from different TEs assuming a mono-exponential T2 decay. The following range of TE values were acquired in order to sample the signal decay TE = 64µsec, 128µsec, 256µsec, 512µsec, 1024µsec, 1500µsec and 2048µsec. A least square fit was performed to the data assuming a mono exponential T2* decay using (Matlab 7.4, MathWorks Inc.). A baseline correction was applied. 3D image rendering was performed with OsiriX (Open Source)x.
In vivo 7T UTE imaging was performed using a similar 3D sequence, but instead of global excitation using a hard pulse, only a 10 cm slab was excited using a 20° 800 µs sinc excitation. A quadrature T/R knee coil with two elements (Nova Medical, Wilmington, MA) was positioned around the lower leg of a volunteer. Fat suppression was achieved using a spectrally-selective RF pulse, applied every 4 TRs. The sequence parameters were TE = 132µs, TR = 20ms, 1ms readout duration, 0.8mm isotropic resolution, 8×8×12cm FOV, NEX = 2, and a 14 min acquisition time.
Examples of 3D volume rendered isotropic UTE images acquired at 7T MRI are shown in Figure 2. The rendering was made possible by using an isotropic voxel size of 1mm. Additional coronal and sagittal sections are depicted in Figure 3 for each field strength. The signal gained from the ultra short T2 components are demonstrated. This image emphasizes the contour of the cortical bone of the tibia. Using exactly the same saddle coil at both field strengths, we found a significant increase in SNR of factor 1.7 from 3T to 7T field strength. An example of a T2* map at each field strengths is shown in Figure 4. The original images at the shortest echo time were under laid to the color maps. The mean and standard deviation of the measured T2* values are depicted in Table 1 for each of the specimens. The T2* values correlated with R=0.91 between the two different field strengths.
Axial sections from an 3D UTE in vivo acquisition at 7T are shown in Figure 5. The fat suppression significantly improves the contrast of the cortical bone, but the fat suppression quality varied across the slab due to B1 inhomogeneity. This result demonstrates coverage of a large 3D volume (8×8×12 cm) with fine spatial resolution (0.8mm isotropic) in a reasonable scan time (14 mins).
Like most solid and semi-solid tissues, cortical bone has an ultra short T2 relaxation time of less than 1.0ms and thus shows no signal in conventional MRI. Hence, it can only be measured by means of negative contrast to a background signal against which the bone image can be interpreted. However, to fully characterize normal cortical bone and its properties, signal has to be captured from the tissue itself. Through the use of UTE sequences, the TE can be considerably reduced and the signal detected before it has totally decayed. Constraints and imperfections in the MRI system significantly affect UTE imaging. The rapid acquisition methods push the limits of the gradient system and constraints on the RF amplitude limit the short-T2 excitation. The rapid switching between transmit and receive modes (between 60 and 300µs for clinical systems (16) and transient effects, such as eddy currents and coil ringing, cause artifacts. There are also other residual currents that can occur in the gradients, believed to be the result of the amplifiers not being completely switched off. All these residual currents dephase the magnetization for a short period following the gradient and thus distort the acquired data and the slice profile (17). Small timing errors between the different gradients, RF pulses and acquisition can also result in detrimental artifacts. The hard pulse excitation we used required no accompanying gradients, eliminating problems associated with residual gradients from the slice selection. Delays between the different gradients and the acquisition were manually calibrated on phantoms. The hard pulse also provided maximal short-T2 excitation. The use of this pulse was feasible because of the specimen size and the small sensitive volume of the coil, and this may not be practical for clinical applications.
The quality of the exponential fit was hampered by multiple short and long T2 values (7). For one cortical specimen, the mono exponential fit was inconsistent on both systems mainly due to increased fat contamination. For these scans, the center frequency was set incorrectly, causing the off-resonant fat to accumulate significant phase during the readout. For radial trajectories, this phase accrual results in radial ringing image artifacts, as can be seen with point spread function analysis, causing the fat signal in this specimen to intrude into the cortical bone. To reduce artifacts due to the fat resonance frequency shift, we applied a simple off-resonance correction scheme to the radially acquired data from this specimen. Prior to gridding, the data was multiplied by a complex exponential at several different frequencies, and the resulting images were evaluated qualitatively to determine the best correction to apply. As a result, the cortical regions were more clearly depicted and thus the quality of the mono exponential fit was improved. This allowed us to salvage this data, although ideally the correct center frequency would be used for the acquisition. However, curve fitting is complicated due to the presence of different ultra-short (collagen matrix), short (organic matrix and bound water) and long T2 (free water) components, which result in more complex decay curves. Previous studies (12) reported a T2* mean value of 576±38 µs for in vivo measurements, which are very similar to the values found here. Values found for 7T were significantly smaller than the values found at 3T and the SNR was clearly increased at the higher field strength for all specimens. Some variation in T1 values depending on porosity was previously shown (12). The three-dimensional UTE imaging pulse sequence used here features relatively short TR in order to be efficient. Thus, some T1 effects are expected to influence the SNR measurements as previously measured T1 of bone water at 3 Tesla indicated a range of 100 to 350 ms (18).
In this work, we have demonstrated the feasibility of ultra short TE imaging for cortical bone using UHF-MRI at high resolution and isotropic voxel size. We characterized and compared UTE images from 7T and 3T field strength using clinical scanners and exactly the same radiofrequency coils at both field strengths. We determined an increase in SNR of factor 1.7 and a significant change in T2* values. A stronger decrease of T2* is also expected at the higher field strength since magnetic field inhomogeneities and susceptibility effects are much more pronounced at 7T (19). These effects cannot be separated from intrinsic T2 characteristics with UTE sequences. Currently, we are expanding our research to 2D pulse sequences using two half pulses (8) and in vivo UTE at both 7T and 3T field strength to demonstrate its feasibility in vivo. The gain in SNR can potentially be used to further enhance spatial resolution.
This work was funded by UC Discovery Grant LSIT01-10107 ITL-BIO04-10148 And NIH grant award program numbers R01-AR057336, 1K99EB012064, R01-AR49701, R01-AG17762 and P30-AR058899.
Roland Krug, Department of Radiology and Biomedical Imaging, University of California, San Francisco.
Peder Eric Zufall Larson, Department of Radiology and Biomedical Imaging, University of California, San Francisco.
Chunsheng Wang, Department of Radiology and Biomedical Imaging, University of California, San Francisco.
Andrew J. Burghardt, Department of Radiology and Biomedical Imaging, University of California, San Francisco.
Douglas A. C. Kelley, GE Healthcare Technologies, San Francisco, California, USA.
Thomas M. Link, Department of Radiology and Biomedical Imaging, University of California, San Francisco.
Xiaoliang Zhang, Department of Radiology and Biomedical Imaging, University of California, San Francisco.
Daniel B. Vigneron, Department of Radiology and Biomedical Imaging, University of California, San Francisco.
Sharmila Majumdar, Department of Radiology and Biomedical Imaging, University of California, San Francisco.