High-Resolution Time-Resolved Contrast-Enhanced MR Abdominal and Pulmonary Angiography Using a Spiral- TRICKS Sequence

Transcription

1 High-Resolution Time-Resolved Contrast-Enhanced MR Abdominal and Pulmonary Angiography Using a Spiral- TRICKS Sequence Jiang Du* and Mark Bydder Magnetic Resonance in Medicine 58: (2007) Both high spatial resolution and high temporal resolution are desirable for contrast-enhanced magnetic resonance angiography (CE-MRA) in order to depict the arterial vasculature. In this work a fast MR pulse sequence (spiral time-resolved imaging with contrast kinetics (Spiral-TRICKS)) with spiral readout in-plane and Cartesian slice encoding was developed whereby the slices are partitioned into multiple regions and acquired in the order used with the TRICKS sequence. The combination of highly efficient spiral sampling with TRICKS acquisition significantly reduced imaging time requirements. A unit second temporal reconstructed frame rate could be achieved for threedimensional (3D) CE-MRA without undersampling of the spiral trajectories. Image quality was improved through spiral trajectory measurement and field-map correction. Phantom and volunteer studies were performed to demonstrate the feasibility of this technique. Magn Reson Med 58: , Wiley-Liss, Inc. Key words: gradient distortion; magnetic resonance angiography; spiral trajectories; three-dimensional; time-resolved MRA Three-dimensional (3D) contrast-enhanced magnetic resonance angiography (CE-MRA) has been established as a safe and reliable method for the evaluation of abdominal vasculature. Conventional CE-MRA examinations are typically performed using a single-phase 3D Fourier transform gradient-echo acquisition. It is critical to coordinate the acquisition of central k-space data with contrast arrival to generate arterial phase images and suppress venous contamination (1). Time-resolved acquisition provides contrast dynamics in the vasculature by capturing the passage of the contrast bolus through the artery and vein of interest (2 8). Vascular pathologies with complex enhancement kinetics, such as delayed filling or asymmetric filling, can be assessed on multiple successive phases of contrast bolus (6 8). Therefore, time-resolved acquisition eliminates timing considerations and reduces sensitivity to filling delays between vessels. However, conventional 3D Cartesian acquisition takes about sec to produce good image quality and thus is unable to provide high temporal resolution. To overcome this problem, several techniques have been developed to speed up the acquisition. The time-resolved imaging with contrast kinetics (TRICKS) sequence divides the 3D Cartesian k-space into several subvolumes located at increasing distance from the k-space center and samples the k-space center more often than outer sections to generate a high temporal reconstructed frame rate (6). The spatial resolution is traded for temporal resolution. Using a combination of undersampled projection reconstruction (PR) with TRICKS acquisition allows a high temporal reconstructed frame rate without spatial resolution degradation (7,8). Compared to radial PR, spiral trajectory allows longer sampling intervals and offers more efficient raw data sampling by circular coverage of k-space (9 11). The complete k-space can be covered with fewer excitations without undersampling artifact. Spiral trajectories permit sampling to begin at the k-space center, which significantly reduces the echo time (TE) and provides intrinsic flow compensation. The longer readout of spiral leads to a longer repetition time (TR), which increases the signal-to-noise ratio (SNR) due to more signal recovery and lowers the average specific absorption rate (SAR) delivered to the patient. But the long spiral readout also leads to reduced stationary spin suppression and higher sensitivity to B 0 inhomogeneities, resulting in image blurring (12). Gradient error and eddy-current effects may also contribute to increased image blurring. However, blurring due to B 0 -field inhomogeneities can be reduced by correcting the phase errors through field-map generation (4,12). Gradient distortion and eddy-current effects can be corrected through spiral trajectory measurement (13,14). In this work a fast MR pulse sequence (Spiral-TRICKS) was developed with spiral in-plane readout and Cartesian slice encoding. The 3D k-space was partitioned into multiple regions along the slice-encoding direction and acquired in the order used with the TRICKS sequence. A unit second temporal reconstructed frame rate was achieved for 3D CE-MRA. Gradient distortion and eddy-current effects were reduced through spiral trajectory measurement and field-map correction. Phantom and volunteer studies were performed to demonstrate the feasibility of this technique. MATERIALS AND METHODS Pulse Sequence Department of Radiology, University of California San Diego, San Diego, California, USA. Grant sponsor: Bracco Diagnostic, Inc. *Correspondence to: Jiang Du, Ph.D., Department of Radiology, University of California San Diego, 200 West Arbor Drive, San Diego, CA jiangdu@ucsd.edu Received 1 June 2006; revised 3 April 2007; accepted 8 April DOI /mrm Published online in Wiley InterScience ( Wiley-Liss, Inc. 631 A fast dynamic imaging sequence (spiral-tricks, shown in Fig. 1) was implemented whereby a stack of spiral 3D trajectories is integrated with TRICKS acquisition to provide high-resolution CE-MRA. The Spiral-TRICKS sequence employs spiral readout trajectory in-plane (kx and ky), and conventional Fourier encoding in the slice direction (kz). In each slice-encoding step, 72 (pulmonary MRA) or 48 (abdominal MRA) interleaved spirals were

2 632 Du et al. FIG. 1. a: k-space sampling using a 3D stack of spirals trajectory with kz slices partitioned into four regions (regions A D). b: The interleaved spiral trajectory used for in-plane sampling. c: Schematics of data acquisition. Mask data were acquired for all regions, followed by dynamic data acquisition in the orders used in TRICKS. A sliding-window reconstruction was used to generate high-spatialresolution time-frame images. acquired. The number of spiral interleaves and readout points were chosen as a compromise between spatial resolution, off-resonance blurring, and sampling efficiency. The slice encodings were partitioned into multiple regions (regions A D, with region D being three times the size of regions A C) from low-frequency slice encodings to highfrequency slice encodings and sampled with a TRICKS acquisition scheme (A 1 B 2 A 3 C 4 A 5 D 6 A 7 B 8...), where the subscript refers to the acquisition order (8). Furthermore, spiral interleaves from region A were divided into three subsets of interleaves, with each acquisition of region A covering only one subset of spiral interleaves, a strategy similar to the selective line acquisition mode (SLAM) technique (15). In the slice-encoding direction, asymmetric k-space sampling was employed to shorten the total acquisition time. Only slices from region A were fully sampled, while slices from regions B D were half sampled. As a result, only 58% of the slice partitions were acquired. Homodyne reconstruction along the slice-encoding direction was carried out to recover the slice resolution (16). The combination of a spiral trajectory, TRICKS acquisition, SLAM technique, and partial slice encoding was expected to significantly improve the temporal reconstructed frame rate of 3D dynamic imaging. Image Reconstruction and Correction A sliding-window reconstruction algorithm was used for dynamic image reconstruction, where data sharing of different regions among neighbor time frames provides high temporal reconstructed frame rate 3D images without spatial resolution degradation (6 8). View sharing will result in temporal resolution blurring, which is regarded as acceptable for dynamic CE-MRA. This reconstruction strategy has been widely used in fluoro-triggering and other dynamic image reconstruction methods (17 19). Spiral acquisition is susceptible to gradient distortion, eddy currents, and off-resonance effects that result in image artifacts such as blurring, distortion, and SNR degradation (9,16). Many algorithms have been developed to address these issues (9,16,20,21). Here a two-step correction algorithm was adopted. First, eddy-current effects and gradient errors were minimized through k-space trajectory measurement (13,14), where the phase difference with and without readout gradients accurately maps the true k-space trajectory. Second, phase error due to off-resonance from B 0 -field inhomogeneities was corrected through field-map acquisition (3,4). A conjugate phase reconstruction was applied to correct the zeroth-order effects of the off-resonance (20,21). Frequency-segmented off-resonance corrections, such as those based on minimizing the imaginary part of the image signal, may offer higher-order correction (4,12), but the reconstruction time will be significantly longer. The entire reconstruction time, including calculating the spiral trajectory and field map, regridding the interleaved spiral interleaves onto Cartesian grids, and fast Fourier transform of the 3D dynamic images (17 time frames) acquired with an eight-channel phased-array coil, took about half an hour on a 3.2 GHz Pentium 4 processor. The multiple coil data were combined using the square root of sum of squares (22). Experimental Methods Phantom and volunteer studies were performed on a standard 1.5T MR scanner (Signa LX; GE Medical System, Waukesha, WI, USA). A phantom was scanned to test the effects of gradient distortion, eddy current, and B 0 -inhomogeneity correction. Time-resolved imaging of the pulmonary and renal vasculature was performed on four volunteers (32 63 years old, kg; two for pulmonary and two for renal MRA) using an eight-element cardiac phasedarray coil (MRI Devices, Milwaukee, WI, USA). The acquisition parameters for pulmonary/renal imaging are were FOV cm, TR 5.1/7.4 ms, TE 0.6/0.8 ms, flip angle 30, BW 125 khz, readout 512/1024 points per interleaf, spiral interleaves 72/48, 36 kz slice encodings in the anterior posterior direction, slice thickness 3.0/2.0 mm, and reconstruction matrix size Region A was updated every sec. Other regions and all of k-space were updated every 7.5 sec. The total scan time was 30 sec, including mask time. The injection of 30 ml Gd-BOPTA (MultiHance ; Bracco Imaging SpA, Mila, Italy) followed by a 20-ml saline flush was applied to each dynamic acquisition. A computer-controlled power injector (Spectris; Medrad, Indianola, PA, USA) was used to ensure a precise injection rate of 3.0 ml/sec for both contrast material and flush. Written informed consent was obtained before the imaging procedures in accordance with institutional review board rules. RESULTS k-space distortion was measured as shown in Fig. 2. Figure 2a shows the ideal and measured trajectories along the kx direction. Figure 2b shows the k-space deviation from the ideal trajectory for both kx and ky directions. As shown in the image, the distortion is nonlinear and thus cannot be corrected through simple gradient delay or k-space shift. The phantom images shown in Fig. 3 demonstrate the

3 Dynamic Imaging Using Spiral-TRICKS 633 FIG. 2. a: Measured and ideal spiral trajectories (only the first 100 of 512 points are displayed for better depiction of the difference). b: k-space deviation of the measured spiral trajectory for the Gx and Gy gradients. efficacy of this k-space distortion correction. Image artifacts were significantly reduced with the measured trajectory instead of the ideal trajectory. Off-resonance correction using a field map further reduced the image artifact. Object details located at the superior and inferior sides of the imaging FOV were recovered through this correction. The Spiral-TRICKS sequence was applied to CE pulmonary imaging (Fig. 4). In total, 17 time frames were generated within 30 sec (including mask), resulting in a high temporal reconstructed frame rate of 1.0 sec/frame, which captured the complete contrast dynamics in the lung, including the arterial, arterial venous transition, and venous phases. The high in-plane spatial resolution of 1.56 mm, which was zero-padded to 0.63 mm, provided an excellent depiction of the pulmonary vessels. Figure 5 shows multiple projections of two consecutive arterial phases of the renal vasculature, which demonstrate the high spatial resolution ( mm 3 ) and rapid renal parenchymal enhancement. A high temporal reconstructed frame rate of 1 sec/frame captured the segmental arteries without parenchymal contamination. DISCUSSION Our preliminary data demonstrate that the spiral-tricks sequence can provide CE-MRA of the abdominal (pulmonary and renal) vasculature with both high spatial resolution and a high temporal reconstructed frame rate. The 4D imaging capability (3D spatial resolution plus 1D temporal resolution) provides detailed contrast enhancement kinetics and a solution for imaging delayed filling or asymmetric filling of the contrast bolus. FIG. 3. Phantom images without any correction (a), with k-space trajectory measurement (b), and k-space trajectory measurement plus B 0 field-map correction (c). The B 0 field map only corrects the phase errors, and k-space distortion was corrected using a spiral trajectory measurement. Eddy-current and gradient distortion was significantly reduced after the trajectory measurement (thick arrows). B 0 field-map correction helps to recover signal at the periphery of the imaging FOV (thin arrows), where B 0 inhomogeneity is most significant. FIG. 4. Complete contrast dynamics of the pulmonary vasculature is depicted through 15 consecutive time-frame maximum intensity projection (MIP) images with a temporal reconstructed frame rate of 1.0 sec/frame. The acquisition FOV of mm 3, spiral readout of 512, and 72 spiral interleaves result in an acquired voxel size of mm 3, which was zero-padded to mm 3 in the reconstruction.

4 634 Du et al. FIG. 5. Rotation of two consecutive time frames. Frames 4 (a c) and 5(d f) depict the fast parenchyma and venous enhancement in the renal vasculature with high spatial resolution ( mm 3 ) and a high temporal reconstructed frame rate (1 sec/ frame). A spiral trajectory provides higher acquisition efficiency than a Cartesian or radial trajectory, although a quantitative comparison among the three trajectories was not performed in this study. The combination of spiral trajectory with TRICKS acquisition provides a temporal reconstructed frame rate of 1 sec/frame. Conventional multiphase acquisitions typically provide a temporal resolution of 3 6 sec/frame by using a short TR and partial echo, partial slice-encoding acquisition (2,5). In a spiral-tricks acquisition the low spatial frequencies are slightly oversampled, which helps to reduce motion. The high spatial frequencies were sampled according to the Nyquist criteria. Data-sharing was performed in the slice-encoding direction. This differs from the sliding-window reconstruction algorithm used in 3D PR (19), in which view-sharing is performed for high-frequency projection data to suppress undersampling streak artifact. The combination of spiral-tricks with the SLAM technique further improves the frame update rate at the cost of slightly more temporal blurring (15). The major problem associated with spiral-tricks acquisition is its susceptibility to gradient distortion, eddy currents, and B 0 inhomogeneity (3,16). Spiral images typically suffer from blurring, especially at the periphery of the imaging FOV, where B 0 inhomogeneity is substantially higher than that at the center FOV. The two-step correction algorithm used here significantly suppressed these artifacts. Gradient anisotropy leads to different gradient delays along each physical gradient. The k-space trajectory measurement automatically corrects the gradient anisotropy, as well as the effects of eddy currents and gradient nonlinearities. Its combination with field-map correction further improves the image quality, especially at the periphery of the imaging FOV, by correcting both the phase errors due to center frequency offset and linear inhomogeneity (through k-space measurement). In our study a low-spatial-resolution field map was acquired only once before the dynamic image acquisition. Recent studies have shown that the field map may change substantially upon contrast arrival and departure, which may result in residual blurring artifacts when using a global field map correction rather than the actual field map (3). Therefore, it may be necessary to periodically acquire low-spatial-resolution field maps during the dynamic image acquisition. A simpler improvement is to acquire field maps before and after the contrast injection (4). The mask and images before contrast arrival can be corrected using the first field map, while postcontrast images can be corrected using the second field map. In our study the k-space deviation was corrected using the trajectory measurement algorithm (13,14). As shown in Fig. 2, the k-space deviation is nonlinear, which makes correction by a simple k-space shift inadequate. However, a k-space shift can suppress some of the spiral artifact. In summary, we have described a time-resolved 3D spiral-tricks sequence that generates 3D CE-MRA images with both high spatial resolution and a high temporal reconstructed frame rate. REFERENCES 1. Prince MR, Grist TM, Debatin JF. 3D contrast-enhanced MR angiography. New York: Springer-Verlag; Schoenberg SO, Bock M, Knopp MV, Essig M, Laub G, Hawighorst H, Zuna I, Kallinowski F, Kaick G. Renal arteries: optimization of threedimensional gadolinium-enhanced MR angiography with bolus-timingindependent fast multiphase acquisition in a single breath hold. Radiology 1999;211: Amann M, Bock M, Floemer F, Schoenberg SO, Schad LR. Threedimensional spiral MR imaging: application to renal multiphase contrast-enhanced angiography. Magn Reson Med 2002;48: Zhu H, Buck DG, Zhang ZH, Zhang HL, Wang P, Steger VA, Prince MR, Wang Y. High temporal and spatial resolution 4D MRA using spiral data sampling and sliding window reconstruction. Magn Reson Med 2004;52: Carr JC, Laub G, Zheng J, Pereles SF, Finn JP. Time-resolved threedimensional pulmonary MR angiography and perfusion imaging with ultrashort repetition time. Acad Radiol 2002;9: Korosec FR, Frayne R, Grist TM, Mistretta MA. Time-resolved contrastenhanced 3D MR angiography. Magn Reson Med 1996;36: Vigen KK, Peters DC, Grist TM, Block WF, Mistretta CA. Undersampled projection-reconstruction imaging for time-resolved contrast-enhanced imaging. Magn Reson Med 2000;43: Du J, Carroll TJ, Wagner HJ, Vigen KK, Fain SB, Block WF, Korosec FR, Grist TM, Mistretta CA. Time-resolved, undersampled projection reconstruction imaging for high resolution CE MRA of the distal runoff vessels. Magn Reson Med 2002;48: Meyer CH, Hu BS, Nishimura DG, Macovski A. Fast spiral coronary artery imaging. Magn Reson Med 1992;28: Spielman DM, Pauly JM, Meyer CH. Magnetic-resonance fluoroscopy using spirals with variable sampling densities. Magn Reson Med 1995; 34: King K, Foo TKF, Crawford CR. Optimized gradient waveforms for spiral scanning. Magn Reson Med 1995;34: Noll DC, Pauly JM, Meyer CH, Nishimura DG, Macovski A. Deblurring for non-2d Fourier transform magnetic resonance imaging. Magn Reson Med 1992;25: Duyn JH, Yang YH, Frank JA, Veen JW. Simple correction method for k-space trajectory deviations in MRI. J Magn Reson 1998;132: Zhang Y, Hetherington HP, Stokely EM, Mason GF, Twieg DB. A novel k-space trajectory measurement technique. Magn Reson Med 1998;39: Rehwald WG, Kim RJ, Simonetti OP, Laub G, Judd RM. Theory of high-speed MR imaging of the human heart with the selective line acquisition mode. Radiology 2001;220: Noll D, Nishimura D, Makovski A. Homodyne detection in magnetic resonance imaging. IEEE Trans Med Imaging 1991;10: Riederer SJ, Tasciyan T, Farzaneh F, Lee JN, Wright RG, Herfkens RJ. MR fluoroscopy: technical feasibility. Magn Reson Med 1988;8:1 15.

5 Dynamic Imaging Using Spiral-TRICKS Foo TKF, Bernstein MA, Aisen AM, Hernandez RJ, Collick BD, Bernstein T. Improved ejection fraction and flow velocity estimates with use of view sharing and uniform repetition time excitation with fast cardiac techniques. Radiology 1995;195: Du J, Carroll TJ, Brodsky E, Lu AM, Mistretta CA, Block WF. Contrast enhanced peripheral magnetic resonance angiography using time-resolved vastly undersampled isotropic projection reconstruction. J Magn Reson Imaging 2004;20: Irarrazabal P, Meyer CH, Nishimura DG, Macovski A. Inhomogeneity correction using an estimated linear field map. Magn Reson Med 1996; 35: Kadah YM, Hu XP. Simulated phase evolution rewinding (SPHERE): a technique for reducing B 0 inhomogeneity effects in MR images. Magn Reson Med 1997;38: Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased array. Magn Reson Med 1990;16: