Three-Dimensional Magnetization-Prepared Rapid Gradient-Echo Imaging (3D MP RAGE)

A new three-dimensional imaging technique which is applicable for 3D MR imaging throughout the body is introduced. In our preliminary investigations we have acquired high-quality 3D image sets ofthe abdomen showing minimal respiratory artifacts in just over 7 min (voxel size 2.7 X 2.7 X 2.7 mm3), and 3D image sets of the head showing excellent gray/white contrast in less than 6 min (voxel size 1.0 X 2.0 X 1.4 mm').


INTRODUCTION
Three-dimensional magnetization-prepared rapid gradient-echo imaging ( 3D MP RAGE) is a novel technique applicable for three-dimensional imaging throughout the body. A schematic representation of 3D MP RAGE is shown in Fig. 1. The sequence begins with a magnetization preparation (MP) period as employed by Haase et al. (1, 2) in conjunction with the snapshot FLASH technique. For example, an inversion pulse followed by a delay yields TI-dependent contrast, and a 90"-180"-90" pulse set gives TZdependent contrast.
After the longitudinal magnetization is encoded with the desired contrast during the MP period, a rapid gradient-echo (RAGE) sequence is used to sample the prepared magnetization. The sampling sequence might be any of the well-known rapid gradient-echo techniques such as FLASH (3), snapshot FLASH ( 1 , 2), GRASS, FISP ( 4 ) , FAST,or FFE ( 5 ) . A fraction of the desired 3D k-space volume is acquired by a given RAGE acquisition. The amount of k space sampled with a given acquisition depends on the following: 1. The temporal nature of the motion in the region of the body being imaged. For example, depending on the TR of the RAGE acquisition, a smaller fraction of k space might be sampled when the heart is imaged than when the liver is imaged.
2. The degree to which the acquisition degrades the prepared contrast. This of course depends on the TR and flip angle for the RAGE acquisition as well as the T1 relaxation properties for the tissues of interest.
In Fig. 1  A magnetization recovery period follows each RAGE acquisition. The duration of the recovery period is determined by the desired contrast properties of the image, the TI relaxation properties of the tissues, and the state of the longitudinal magnetization at the end of the gradient-echo acquisition. The two limiting cases for the magnetization recovery period are zero duration and a duration which is relatively long compared to the T 1 's of interest. The second case, that of a relatively long recovery period, is particularly important since in this case a given preparation-acquisition-recovery (P-A-R) cycle is decoupled from other cycles, and relative variations in the recovery periods therefore do not adversely affect the image quality.
The complete 3D MP RAGE data set is formed by repeating the P-A-R cycle described above until the desired k-space volume is covered. Note from Fig. 1 that a trigger signal may be used to initiate each cycle for imaging structures subject to periodic motion such as the liver or heart.
The structure of the 3D MP RAGE sequence provides several inherent advantages over existing 3D acquisition methods which operate in a steady-state mode. (By steady state we mean sequences which are based on either a steady state of the longitudinal component of the magnetization (e.g., standard spin-echo or FLASH) or a steady state of the complete magnetization vector (e.g., FISP or GRASS) .) The advantages of 3D MP RAGE include: 1. The cyclic nature ofthe sequence makes it naturally applicable to imaging structures subject to periodic motion such as the liver or heart using respiratory or cardiac triggering, respectively. When the magnetization recovery period is sufficiently long with respect to the tissue T 1 values, variations in the total cycle length as may occur with changes in heart or respiratory rates do not adversely affect the images. To the best of our knowledge, this is the first technique that can produce 3D image sets of the abdomen with minimal respiratory artifacts in an imaging period acceptable for routine clinical use.
2. By employing a separate magnetization preparation period, given tissue contrast properties can be obtained in a much shorter imaging time than is possible using existing steady-state acquisition schemes.
3. The dead times in the preparation and recovery periods can be used for secondary magnetization preparations such as spatial or chemical presaturation.   The total imaging time was 7.18 min. The magnetization preparation consisted of an inversion pulse followed by a 350-ms delay which produced strong T 1 weighting in the image. Each RAGE acquisition acquired 128 lines in 1024 ms (TR/TE 8/3.3, FLASH type sequence, lo" flip angle) and was performed at end expiration. The recovery period was 2 s. Because respiratory triggering was not yet available on our machine, the subject voluntarily respired in synchrony with the sequence. Figures  2b-2d show coronal ( b and c) and transverse (d) images reformatted from the original sagittal acquisition. Since each RAGE acquisition was only 1 s, image artifacts from stomach, bowel, and cardiac motions appear predominantly in one phase-encoding direction (horizontal in Fig. 2a). Note the sharp definition of the upper edge of the liver in Figs. 2a-2c. Examination of the anterior subcutaneous fat in Figs. 2a and 2d reveals only minor artifacts from respiration. Note the relatively black appearance of flowing blood as seen in Fig. 2c. This feature may prove very valuable in relation to studies of vessels diseased with atherosclerosis. Figure 3 shows images from a 3D MP RAGE acquisition of the head of a normal volunteer acquired in the sagittal orientation. The image matrix was 128 ( 180 mm) by 128 (250 mm) by 256 (250 mm), interpolated to 128 X 256 X 256. This yields voxels with dimensions of 1.4 by 1.0 (interpolated) by 1.0 mm. The total imaging time was 5.92 min. The magnetization preparation consisted of an inversion pulse followed by a 500-ms delay which produced strong T 1 weighting in the image. Each RAGE acquisition acquired 128 lines in 1280 ms (TR/TE 10/4.15, FLASH type sequence, lo" flip angle). The recovery period was 1 s. Figures 3b-3d show coronal (band c) and transverse (d) images reformatted from the original sagittal acquisition. The images display excellent gray/white contrast compared to the standard T l-dependent imaging sequences we currently employ (400/ 15 2D spin echo and 30/ 5 3D FLASH). Due to the very short TE of the RAGE acquisition and the small voxel sizes, the images do not show any significant susceptibility artifacts at air/ soft tissue and bone/soft tissue interfaces. In Fig. 3c, note the bright appearance of the blood in the arteries surrounding the pituitary. During the 500-ms delay period, fully magnetized blood flows into the transmitlreceive head coil resulting in an inflow enhancement effect for the arteries. Blood that experiences the inversion pulse, such as that in the venous structures, appears dark in the images.

CONCLUSIONS
Initial experience with the 3D MP RAGE sequence shows promise that it will be a noteworthy addition to the existing arsenal of MR imaging techniques. Particularly exciting is the possibility of generating high-quality three-dimensional image sets of the abdomen.
There remain many important issues to be addressed in connection with the optimum parameter values for given imaging situations. Since the magnetization is sampled during a transient, many aspects of optimizing this sequence may be even more difficult than was the case for existing steady-state imaging techniques. We have begun theoretical and experimental investigation (6) into the general problem of optimizing the contrast for a prepare-acquire imaging sequence such as 3D MP RAGE. For the RAGE acquisition, we have implemented and are currently investigating modifications to the standard acquisition procedure, including (i) RF pulse phase offsets between excitations ( 7) to eliminate the formation of steady-state transverse magnetization and the resulting artifacts, and (ii) modification of the sampling order in k space (e.g., sampling a 2D k-space plane from the respective k = 0 component outward) to improve the characteristics of the point spread function and provide access to a wider range of image contrast properties.