Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print October 22, 2001, 10.1148/radiol.2213010455
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Quicktime Movies
Right arrow All Versions of this Article:
2213010455v1
221/3/827    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Shankaranarayanan, A.
Right arrow Articles by Duerk, J. L.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Shankaranarayanan, A.
Right arrow Articles by Duerk, J. L.
(Radiology. 2001;221:827-836.)
© RSNA, 2001

Technical Developments

Segmented k-Space and Real-Time Cardiac Cine MR Imaging with Radial Trajectories1

Ajit Shankaranarayanan, PhD, Orlando P. Simonetti, PhD, Gerhard Laub, PhD, Jonathan S. Lewin, MD and Jeffrey L. Duerk, PhD

1 From the Departments of Radiology (Magnetic Resonance Imaging) (A.S., J.S.L., J.L.D.), Biomedical Engineering (A.S., J.L.D.), and Oncology (J.S.L.), Case Western Reserve University and University Hospitals of Cleveland, 11100 Euclid Ave, Cleveland, OH 44106; and Siemens Medical Systems, Chicago, Ill (O.P.S., G.L.). Received February 14, 2001; revision requested March 28; revision received May 18; accepted June 5. Supported in part by research collaborations with Siemens Medical Systems (Chicago, Ill), Radionics (Burlington, Mass), and Guidant (Santa Clara, Calif) and grants from the Whitaker Foundation, the American Cancer Society, the Mary Ann S. Swetland Fund, the M. E. and F. J. Callahan Foundation and by National Institutes of Health, National Cancer Institute grants 1R01CA81431-01A1 and R33-CA88144-01. Address correspondence to J.L.D. (e-mail: duerk@uhrad.com).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Materials and Methods
Results
 Discussion
 REFERENCES
 
The authors developed and evaluated two cine magnetic resonance (MR) imaging sequences with a radial rather than a rectilinear k-space coordinate frame: segmented k space and real-time true fast imaging with steady-state precession, or FISP. The two radial k-space segmentation (or view sharing) techniques, which were interleaved or continuous, were compared, and the feasibility of their application in cardiac cine MR imaging was explored in phantom and volunteer studies. Images obtained with the radial sequences were compared with those obtained with two-dimensional Fourier transform, or 2DFT, sequences currently used in cine MR imaging. Temporal resolution of 55 msec was achieved with the real-time radial sequences, which allowed acquisition of almost 19 high-quality images per second.

Supplemental material: radiology.rsnajnls.org/cgi/content/full/2213010455/DC1.

Index terms: Heart, MR, 51.121412 • Magnetic resonance (MR), cine study, 51.121412 • Magnetic resonance (MR), image processing, 51.121412 • Magnetic resonance (MR), technology, 51.121412


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Materials and Methods
Results
 Discussion
 REFERENCES
 
In any magnetic resonance (MR) imaging application, spatial resolution, contrast-to-noise ratio (CNR), and signal-to-noise ratio (SNR) must be balanced to meet the demands encountered in the clinical setting. Cine MR imaging has the additional requirements of high spatial resolution, high temporal resolution, and short acquisition times. In some cases, temporal resolution as low as 100 msec per image can provide diagnostically useful information, but rapid contraction or relaxation of the myocardium may not be faithfully represented. The k-space segmentation techniques (1,2), echo sharing (3), and interpolation have been used to improve temporal resolution in cine MR imaging while keeping the acquisition time within a reasonable breath-hold period (<20 seconds).

Most previous cine acquisitions have made use of rectilinear k-space sampling on a cartesian grid and two-dimensional Fourier transform (2DFT) methods for reconstructing the images, despite the fact that n-dimensional Fourier transform sequences are susceptible to motion artifacts (eg, image ghosts) that may interfere with visualization of the region of interest. One of the ways to achieve improved temporal resolution in rectilinear cartesian sampling is to reduce the number of phase-encoding lines and acquire a rectangular field of view (FOV) to maintain isotropic spatial resolution. The choice of the FOV in the phase-encoding direction becomes critical, since small FOVs can lead to aliasing (also known as wraparound) artifacts from objects outside the FOV. These artifacts cause the objects outside the FOV to wrap into the imaging area; this may reduce the diagnostic value of the image.

Historically, the first MR images were obtained by using radial k-space sampling and backprojection reconstruction (4), but technical challenges (eg, eddy currents, chemical shift) led to the later widespread adoption of n-dimensional Fourier transform MR imaging (5). With recent improvements in technology and the investigation of new applications for MR imaging (eg, interventional MR), however, projection reconstruction methods (6–10) are once again gaining popularity, since they offer advantageous motion artifact immunity (11,12) when compared with 2DFT methods while they provide comparable temporal resolution (13–18). Radial sampling techniques also offer a different set of trade-offs than are encountered in standard Fourier imaging: Undersampling in the angular direction primarily gives rise to reduced SNR and, to a lesser extent, to a decrease in spatial resolution.

Fast low-angle shot, or FLASH, techniques, also known as spoiled gradient-recalled techniques, have generally been used for cine cardiac MR imaging (19,20). However, fast low-angle shot techniques rely on inflow enhancement for blood or myocardium contrast. The inflow enhancement decreases with shorter repetition times, which causes saturation of the blood signal and, thus, reduction of the blood or myocardium contrast. True fast imaging with steady-state precession (FISP) (21), with its exceptionally high signal at short repetition time, high inherent blood or myocardium contrast, and motion insensitivity, has been shown to have better performance for rapid cine MR imaging in the heart (22–25). At a repetition time much less than T2, the resultant image signal intensity is inversely proportional to the ratio of T1 to T2. The ratio is approximately 5 in arterial blood and 18 in normal myocardium. This large difference has been exploited to generate high-contrast cine MR images of the heart at short repetition times.

Our hypothesis was that the properties of radial data acquisition would offer advantages over 2DFT techniques in temporal and spatial resolution when combined with the high SNR and CNR with the true FISP pulse sequence. Therefore, the objective of this study was to develop segmented and real-time radial true FISP sequences and to compare them among themselves and also with the 2DFT techniques being used currently for cardiac cine MR imaging.


   Materials and Methods
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
 Discussion
 REFERENCES
 
MR Imaging System
The radial k-space trajectory pulse sequences were implemented with a 1.5-T MR imager (Magnetom Sonata; Siemens Medical Systems, Erlangen, Germany) with a maximum gradient strength of 40 mT/m and maximum slew rate of 200 mT/m/msec. Volunteer MR imaging was performed with electrocardiographic monitoring of the heart activity for triggered data acquisition. The integrated body coil was used for radio-frequency transmission while a circularly polarized four-channel body array was used for signal reception. The pulse sequences were written to permit flexibility in selecting arbitrary section positions, variable matrix size, and variable number of projections for each of the radial sequences.

Developed Sequences
All sequences written and used in this project were for single-section acquisitions at multiple phases in the cardiac cycle to permit image review in cine format.

k-Space segmentation.—Two types of radial k-space segmentation, based on strategies developed for interventional MR imaging, were implemented (26). The two segmentation strategies are called interleaved and continuous segmentation. Figure 1 describes their k-space coverage. An interleaved segmentation is one in which each segment rotationally covers all k space, albeit coarsely. The first segment of acquired projections may include projection angles, for example, 0°, 2°, 4°, 6°, ... 178°, while the second segment acquires projections at angles 1°, 3°, 5°, ... 179°. This is also analogous to the interleaved segmentation implemented in an earlier 2DFT fast low-angle shot cine MR angiographic sequence (27,28). A continuous segmentation is one in which each segment contains projections that are adjacent, for example 0°, 1°, 2°, ... 89° in the first segment and then 90°, 91°, 92°, ... 179° in the second. This scheme is also analogous to a segmentation scheme implemented in an earlier 2DFT true FISP cine MR angiographic sequence (25). Subsequent segments would acquire sectors from different angular regions of radial k space.



View larger version (34K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Schematics show continuous (A) and interleaved (B) segmentation schemes in terms of the projections selected in each pass. A, Segments in continuous acquisitions contain adjacent projections (1, 2, 3, ...). B, Segments in interleaved acquisitions contain alternate projections (1, 3, 5, ...). Dashed lines indicate views collected for the first segment. Solid lines indicate views collected for the second segment.

 
Radial true FISP.—A regular 2DFT unspoiled steady-state free-precession sequence with balanced gradients along all imaging axes (true FISP) was modified to achieve a radial k-space data acquisition. This sequence was implemented to study some of the basic properties of angularly undersampled radial acquisitions. Despite the radial trajectory, the gradients were balanced to maintain the transverse magnetization in steady state from one repetition time to another. The magnetization was brought into steady state before the start of data acquisition by using a preparation scheme proposed by Diemling and Heid (29) with the addition of approximately 10 dummy pulses.

Segmented radial true FISP.—A segmented radial true FISP sequence was developed by modifying a breath-hold electrocardiography triggered segmented rectilinear k-space acquisition. The two segmentation schemes were implemented in this sequence. Figure 2 shows a schematic representation of this sequence. The electrocardiographic signal was used to trigger the data acquisition. Steady state was maintained throughout the acquisition by applying the radio-frequency pulse throughout the R-R interval, even after acquisition of the data from the last phase was completed. The sequence was developed for 10, 16, and 32 projections per segment, which corresponded to fast, normal, and slow human heart rates; the true FISP module consisted of repetition time msec/echo time msec of 3.0/1.5, which provided effective temporal resolution of 30, 48, and 96 msec, respectively. Table 1 shows the MR imaging parameters used for 16 projections per segment.



View larger version (47K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Schematics depict a true FISP segmented radial cine sequence. A, With continuous segmentation, the projections marked in each of the segments are continuous. B, With interleaved segmentation, the projections marked in each of the segments are not continuous, but each segment covers the entire k space. The electrocardiographic (ECG) signal was used to trigger the acquisition of data. Constant radio-frequency pulsing was used to maintain the steady state.

 

View this table:
[in this window]
[in a new window]

  		TABLE 1. Summary of Imaging Parameters

 
Shared segmented radial true FISP.—The breath-hold segmented radial true FISP sequence was modified to increase the temporal resolution by sharing the projections between the segments, as shown in Figure 3. View sharing, a concept developed for earlier real-time sequences (3), was performed for both interleaved and continuous segmented sequences. Fifty percent of the projections from the adjacent phases were shared to generate additional images at intermediate cardiac phases. The sequence was implemented for 10, 16, and 32 projections per segment. This provided an effective temporal resolution of 15, 24, and 48 msec, respectively, for the segmented radial true FISP sequence.



View larger version (27K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. Schematic shows implementation of the continuous radial shared segmented sequence. Sharing of projections was performed between each segment to obtain an increased number of cardiac phases. Sequences with 10, 16, and 32 projections per segment were implemented for the different heart rates. As with the radial segmented sequence, constant radio-frequency pulsing was used to maintain steady state.

 
Real-time radial true FISP.—We created a real-time radial true FISP sequence (2.2/ 1.1) by modifying a real-time ungated 2DFT true FISP sequence, which is used currently with echo sharing (3), to improve the temporal resolution. Fifty percent of the projections were shared with prior and subsequent data. This gave an effective temporal resolution of 55 msec for a 50-projection acquisition. Both interleaved and continuous sharing methods were implemented and compared for their performance in dynamic conditions. Figure 4 shows the schematic of the implementation of the real-time radial sequence. Steady state was again achieved and maintained by applying the gradients and the radio-frequency pulses for one preparatory heartbeat and then at constant repetition time throughout the acquisition.



View larger version (25K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4. Schematic depicts the radial real-time ungated sequence (2.2/1.1; effective temporal resolution, 55 msec). View sharing was performed to increase the temporal resolution. Continuous and interleaved sharing were implemented. Steady state was achieved and maintained by applying the gradients and the radio-frequency pulses for one preparatory heartbeat and then constantly applying them throughout the acquisition.

 
Image reconstruction.—Image reconstruction was perfomed with a variant of the regridding method (30). The coordinates for one quadrant of the k-space matrix were calculated prior to the start of the acquisition; the rest of the coordinates were then estimated from those calculated, on the basis of the symmetry properties of a radially symmetric k space. A simple ramp filter was used for density compensation. Vector processing, rather than scalar methods, was used for the regridding operations to improve the reconstruction time. The radial points were oversampled by two to fill the regridded k space. This provided two advantages. First, twice the number of points was used to fill the radial direction of k space (regridding was performed onto a x2 matrix) in the same amount of time. This made the acquisition more efficient. Second, circular artifacts at the edge of the image were eliminated by using the additional points because the circular artifacts now lay at the edge of the x2 image matrix and not in the central image matrix. The central image matrix was used for viewing the final image.

Image quality measures.—The SNR, CNR, and temporal and spatial resolutions were measured on all the images. In phantom studies, SNR was defined as mean signal amplitude in a uniform region divided by the SD in a region of noise outside the object. For images from volunteer studies, the mean signal amplitude was measured in blood. The sizes of the regions of interest for signal amplitude and noise measurements were 100 and 50 mm2, respectively. CNR was calculated from the mean difference between the blood and myocardium pixel amplitudes and the SD of the noise measured outside the body in a region free of artifacts at simple visual inspection. The regions of interest used to determine the mean signal amplitude of the blood were at least 100 mm2 while those for the myocardium were at least 50 mm2. Regions of interest for noise were at least 100 mm2. One author (A.S.) placed the regions of interest.

Phantom Studies
To confirm successful pulse sequence development, a series of phantom (1,000 g H2O, 1.25 g NiSO4, 5.00 g NaCl [Siemens Medical Systems]) tests were performed prior to evaluation trials to numerically compare the SNR of the developed radial true FISP acquisitions of static objects with that obtained for a 2DFT acquisition with identical echo time, repetition time, flip angle, acquisition matrix, and section thickness. The spatial resolution phantom was then used to study the basic properties of radial sampling, such as the dependence of the SNR and spatial resolution on the number of projections. Data were acquired in the spatial resolution phantom with the radial true FISP sequence by keeping all the factors constant except the number of projections. Acquisition parameters were 3.0/1.5; FOV, 300 mm; flip angle, 60°; bandwidth, 975 Hz/pixel; matrix size, 256 x 256; section thickness, 10 mm; one signal acquired; effective temporal resolution, 55 msec.

Volunteer Studies
All volunteer studies were performed with informed consent according to a protocol approved by the institutional review board for human investigation. The data were acquired in three healthy male volunteers (age range, 30–60 years; mean age, 40 years) with no known cardiac disease. The volunteers were placed in the MR imager, and the short-axis views of the heart were located with use of standard scout sequences. The electrocardiographic signal was used to trigger the MR imaging data acquisition.

Contrast between the blood and myocardium was quantified by calculating the CNR for these regions. Each of the radial cine sequences was then performed at the same section position. The 2DFT segmented true FISP sequence was also performed at the same position with the same MR imaging parameters. The only difference was the k-space trajectories. The MR imaging parameters for these sequences are given in Table 1. Note that the MR imaging parameters given for the segmented and shared segmented sequences are for 16 projections per segment. A currently used real-time 2DFT true FISP sequence, with equivalent MR imaging parameters (Table 1), was also performed for comparison with the real-time radial sequence. SNR and CNR were calculated for all images. Finally, the images were displayed in cine mode so all the authors could determine the appearance of radial streakings (due to motion of the heart) and could visually observe the apparent heart wall motion.


   Results
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
Discussion
 REFERENCES
 
Results in the phantom studies are shown in Figure 5. Note the increase in SNR and the reduction in the amount of streaking artifacts (ie, angular aliasing artifacts in radial k-space acquisitions) with an increasing number of projections. Also note that the spatial resolution remained approximately constant as the number of projections was increased from 64 to 256.



View larger version (195K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5a. True FISP radial MR images (3.0/1.5; flip angle, 60°; FOV, 300 x 300 mm) were acquired in a spatial resolution phantom study of some of the basic properties of radial acquisitions. Two hundred fifty-six radial points were collected with (a) 32, (b) 64, (c) 90, (d) 128, (e) 180, and (f) 256 views. Note that the radial streak artifacts (arrows in a and b) decrease as the number of views increases. Similarly, the SNR increases as the number of views increases. Spatial resolution, however, remains more or less the same as the number of views increases.

 


View larger version (190K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5b. True FISP radial MR images (3.0/1.5; flip angle, 60°; FOV, 300 x 300 mm) were acquired in a spatial resolution phantom study of some of the basic properties of radial acquisitions. Two hundred fifty-six radial points were collected with (a) 32, (b) 64, (c) 90, (d) 128, (e) 180, and (f) 256 views. Note that the radial streak artifacts (arrows in a and b) decrease as the number of views increases. Similarly, the SNR increases as the number of views increases. Spatial resolution, however, remains more or less the same as the number of views increases.

 


View larger version (189K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5c. True FISP radial MR images (3.0/1.5; flip angle, 60°; FOV, 300 x 300 mm) were acquired in a spatial resolution phantom study of some of the basic properties of radial acquisitions. Two hundred fifty-six radial points were collected with (a) 32, (b) 64, (c) 90, (d) 128, (e) 180, and (f) 256 views. Note that the radial streak artifacts (arrows in a and b) decrease as the number of views increases. Similarly, the SNR increases as the number of views increases. Spatial resolution, however, remains more or less the same as the number of views increases.

 


View larger version (197K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5d. True FISP radial MR images (3.0/1.5; flip angle, 60°; FOV, 300 x 300 mm) were acquired in a spatial resolution phantom study of some of the basic properties of radial acquisitions. Two hundred fifty-six radial points were collected with (a) 32, (b) 64, (c) 90, (d) 128, (e) 180, and (f) 256 views. Note that the radial streak artifacts (arrows in a and b) decrease as the number of views increases. Similarly, the SNR increases as the number of views increases. Spatial resolution, however, remains more or less the same as the number of views increases.

 


View larger version (186K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5e. True FISP radial MR images (3.0/1.5; flip angle, 60°; FOV, 300 x 300 mm) were acquired in a spatial resolution phantom study of some of the basic properties of radial acquisitions. Two hundred fifty-six radial points were collected with (a) 32, (b) 64, (c) 90, (d) 128, (e) 180, and (f) 256 views. Note that the radial streak artifacts (arrows in a and b) decrease as the number of views increases. Similarly, the SNR increases as the number of views increases. Spatial resolution, however, remains more or less the same as the number of views increases.

 


View larger version (172K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5f. True FISP radial MR images (3.0/1.5; flip angle, 60°; FOV, 300 x 300 mm) were acquired in a spatial resolution phantom study of some of the basic properties of radial acquisitions. Two hundred fifty-six radial points were collected with (a) 32, (b) 64, (c) 90, (d) 128, (e) 180, and (f) 256 views. Note that the radial streak artifacts (arrows in a and b) decrease as the number of views increases. Similarly, the SNR increases as the number of views increases. Spatial resolution, however, remains more or less the same as the number of views increases.

 
Images from the volunteer studies obtained with the different sequences are shown in Figures 6–9. Results with the segmented radial true FISP sequence are shown in Figure 6. Note an obvious decrease in the spatial resolution of the images in Figure 6 (bottom) when compared with Figure 6 (top). Also note, however, that the SNR of the images in Figure 6 (bottom) is higher (with lower spatial resolution and fewer samples) than that shown in Figure 6 (top). Similarly, the spatial resolution and SNR of the 256 x 256 and 128 x 128 images obtained with the segmented radial cine sequence with interleaved segmentation (Fig 7) behave in a way that is similar to those obtained with continuous segmentation (Fig 6). This can also be seen from the SNR and CNR values shown in Table 2. However, note a slight increase in the amplitude of the angular undersampling artifacts (ie, streaking) in Figure 7 (top) when compared with Figure 6 (top). The images acquired with both strategies performed in cine mode (Movies 1–4, radiology.rsnajnls.org/cgi/content/full/2213010455/DC1) also showed no difference in the apparent motion of the heart.



View larger version (202K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 6. Short-axis continuous segmented radial cine MR images (3.0/1.5; FOV, 300 x 300 mm; flip angle, 55°) of the heart were acquired at two time points with use of 256 radial points (top, 256 x 256 matrix) and 128 radial points (bottom,128 x 128 matrix) with the same number of projections (n = 128). Note the higher spatial resolution of the top images compared with the bottom images. SNR, however, is higher on the bottom images compared with the top images.

 


View larger version (177K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 7. Short-axis interleaved segmented radial cine MR images (3.0/1.5; FOV, 300 x 300 mm; flip angle, 55°) of the heart were acquired at two time points with use of 256 radial points (top, 256 x 256 matrix) and 128 radial points (bottom, 128 x 128 matrix) with the same number of projections (n = 128). Note the higher spatial resolution of the top images compared with the bottom images. SNR, however, is higher on the bottom images compared with the top images. Also note the radial streaking in the top images (arrows), which was not seen with continuous segmentation (Fig 6).

 


View larger version (107K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 8. Short-axis continuous shared segmented MR images (3.0/1.5; FOV, 300 x 300 mm; flip angle, 55°; 256 x 256 matrix) of the heart were acquired at two time points with 16 views per segment. Fifty percent of the views were shared. One hundred forty-four projections were used to obtain images of 36 cardiac phases within the same acquisition time as was used in the radial segmented sequence. Note the higher SNR in comparison with that in Figures 6 (top) or 7 (top).

 


View larger version (181K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 9. Real-time ungated radial MR images (2.2/1.1; FOV, 320 x 320 mm; flip angle, 45°; 128 x 128 matrix) were acquired with continuous (top) and interleaved (bottom) sharing. Fifty percent of the projections were shared to give an effective temporal resolution of 55 msec with 50% sharing. Note the radial streaking (arrows) due to angular undersampling, but no noticeable effect is seen in the region of interest (the two chambers and the walls).

 

View this table:
[in this window]
[in a new window]

  		TABLE 2. Summary of SNR and CNR Values for Cine Sequences

 
Improved temporal resolution was obtained with the shared segmented radial true FISP sequence, as can be seen from the increase in the number of cardiac phases acquired (Table 2) for the same acquisition time, with 50% sharing of projections. This resulted in apparently smoother motion of the heart when performed in cine mode (Movie 5, radiology.rsnajnls.org/cgi/content/full/2213010455/DC1). Also note in Table 2 that 144 projections were used to acquire the data. This resulted in greater than 28% improvement in image SNR for images obtained with the shared segmented sequence (Fig 8) compared with images obtained with the segmented sequence (acquired with 128 projections) (Fig 6, top).

Figure 9 illustrates the images obtained with the real-time radial ungated cine sequence. Images obtained with both continuous (Fig 9, top) and interleaved (Fig 9, bottom) view sharing are shown. Note that the radial streaking inside the region of interest (the two chambers of the heart) appears negligible. However, radial streaking can be seen outside the region of interest. Dynamic visual inspection of the apparent contraction of the heart, when the images from continuous sharing were displayed in cine mode (Movies 6, 7, radiology.rsnajnls.org/cgi/content/full/2213010455/DC1), showed discontinuous motion of the heart wall during systole. However, the motion was more continuous when the images obtained with interleaved sharing were displayed in cine mode. The SNR and CNR for the images are tabulated in Table 2.

Figure 10 shows images obtained with real-time 2DFT and radial true FISP sequences with nearly identical MR imaging parameters (Table 1). Note from Table 1 that a rectangular FOV was used in the 2DFT cine sequence to obtain almost isotropic spatial resolution. This leads to aliasing of the object into the region of interest on the 2DFT true FISP images (Fig 10, bottom; Movie 9, radiology.rsnajnls.org/cgi/content/full/2213010455/DC1), while this is not seen on the radial images (Fig 10, top; Movie 8, radiology.rsnajnls.org/cgi/content/full/2213010455/DC1). Radial streaking artifacts can be seen clearly on the image though the effect of the artifacts in the region of interest seems to be reduced.



View larger version (177K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 10. Real-time radial cine MR images (top [2.2/1.1; flip angle, 45°; square FOV, 300 x 300 mm]) and real-time 2DFT cine MR images (bottom [2.2/1.1; flip angle, 45°; rectangular FOV, 188 x 300 mm]) depict aliasing artifacts (bottom, arrows). No wraparound artifacts are depicted on the top images. Note that the spatial resolution is almost the same on the top and bottom images. Streaking artifacts (top, arrows) on the top images were caused by angular undersampling.

 
The SNR, CNR, number of phases, effective temporal resolution, and total MR imaging duration for the images obtained with each of the radial sequences are summarized in Table 2. Note that an average 39% decrease in the SNR of images obtained with the radial cine sequences was seen when compared with images obtained with the regular 2DFT segmented true FISP sequence. However, the CNR was approximately 19% lower when compared with that of images obtained with the standard 2DFT cine sequence at the same time point with respect to the R wave. Approximately 28% improvement in the SNR and 17% improvement in the CNR were seen when the images obtained with the shared segmented radial cine sequence were compared with images obtained with the segmented radial cine sequences. Also note that the effective temporal resolution of the standard real-time 2DFT sequence was 77 msec, while that for the radial real-time sequence was 55 msec; this allowed collection of an increased number of cardiac phases with the radial sequence for the same acquisition time (Table 2). The CNR and SNR values with the interleaved and continuous methods showed less than 5% difference for both 256 x 256 and 128 x 128 matrix size. However, comparison of SNR between a 128 x 128 image and a 256 x 256 image showed that the SNR for 128 x 128 images was 54% higher, when compared with that for 256 x 256 images, for the same number of projections.


   Discussion
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
 Discussion
REFERENCES
 
In cardiac cine MR imaging, where temporal resolution and spatial resolution are critical parameters, radial sampling techniques can provide advantages over conventional techniques because image spatial resolution depends primarily on the number of samples in the radial direction and less on the number of views acquired. The lower SNR is partially compensated by combining radial sampling with the true FISP sequence, which inherently has higher SNR and CNR than the 2DFT fast low-angle shot sequences currently in widespread use.

Higher SNR and fewer streaking artifacts on images acquired with 256 projections when compared with images acquired with 90 projections is due to the fact, already described in the literature (31,32), that undersampling in the angular direction results in a decrease in SNR and increase in the extent of streaking artifacts, with the intensity of the streaks depending on the degree of undersampling.

The higher SNR observed on the 128 x 128 images compared with the 256 x 256 images results from the dependence of SNR on the voxel size and the degree of undersampling (when all the other MR imaging parameters are kept constant). The degree of undersampling for a 128 x 128 image acquired with 128 projections is, therefore, less than that for a 256 x 256 image acquired with 128 projections, which results in the higher SNR. However, the image spatial resolution obtained on the 256 x 256 image is higher and, more importantly, is more uniform in all directions. Thus, the trade-offs between the SNR and spatial resolution in radial sampling, different from cartesian sampling, provide the opportunity to obtain images with higher spatial resolution per unit of time when compared with current cartesian-based techniques.

A decrease in SNR observed on images obtained with comparable radial and 2DFT cine sequences is partly due to the inefficiency of radial sampling when compared with cartesian sampling. The inefficiency results from the fact that {pi}/2 times the number of samples is needed to achieve similar SNR, as is obtained with cartesian sampling with identical FOV and spatial resolution. Reconstruction methods also play an important role in determining the final SNR with radial sampling. Another reason for the decreased SNR may be the use of a simple ramp filter for density compensation, which results in higher image pixel variance (33,34). Increased SNR values can also be achieved by using backprojection methods (ie, the reconstruction algorithm normally used in computed tomography) instead of the regridding methods used in our implementation (33). The fact that the process of backprojection needs to be performed separately for both the real and imaginary parts of the complex raw data, however, results in an unacceptably long reconstruction time with our current hardware. With hardware improvements in the future, other reconstruction methods will also be implemented and evaluated.

As the images in Figure 8 and the SNR and CNR values in Table 2 show, the shared segmented sequence, in comparison with the regular segmented sequence, provides better image quality by using more projections, which substantially improves the temporal resolution. Thus, the concept of shared segmentation helps to increase either the number of phases for the same amount of acquisition time or the number of projections per segment to acquire the same number of phases. Therefore, the improved temporal and spatial resolutions with higher SNR can be used in breath-hold cardiac cine MR imaging for a fast and accurate cardiac function study.

While the comparison of SNR values between the continuous and interleaved segmented cine MR images did not show any substantial difference, an increase in radial streaking artifacts was observed on interleaved segmented cine MR images. This may be so because each segment was a severely undersampled data set, which caused its own streaking artifacts. Thus, any discrepancies between the segments will result in increased streaking (16) on the final image. Clinically, therefore, the radial segmented cine sequence with continuous segmentation would be expected to perform better than the interleaved segmented cine sequence.

With the real-time radial sequence, continuous sharing, which is analogous to the sliding window reconstruction previously described (15,16), results in alternate updating of data in horizontal and vertical directions. Since contraction and relaxation of the heart are simultaneous in all directions, the corresponding k space has to be continuously updated in all directions to faithfully reproduce the motion. Data acquired for every phase with continuous sharing will result in updating of only a part of k space and will cause the motion of the heart to appear discontinuous and asynchronous. With interleaved sharing, k space is continuously updated in all directions. Thus, in a short-axis view, radial motion of the heart will appear to be smooth when images obtained with interleaved sharing are displayed.

As seen in Table 1, the real-time radial sequence made use of only 50 views to generate the images in Figures 9 (top) and 10 (top). As explained in the literature (31), the point-spread function of discrete angular undersampling shows a circular uniform region at its center; the radius of the uniform region is dependent on the degree of undersampling. For a degree of undersampling of 50/200 and 300 x 300-mm FOV, the radius of the circular uniform area for the region of interest was 7.5 cm (15-cm diameter). When a 300 x 300-mm FOV (normally used in the real-time sequence) was used, the maximum horizontal extent across the two chambers when the heart was in the diastolic phase was 11 cm, and the maximum vertical extent was 7 cm. Thus, the uniform area that spans an image plane region is sufficient to cover the heart without producing radial streaking within the region of interest. Moreover, the temporal resolutions achieved with the sequences were substantially higher than those currently afforded with 2DFT methods, without degrading image quality. For cardiac applications involving short-axis views, the number of projections could be decreased further to optimize the size of the uniform area, which would simultaneously increase the temporal resolution.

In conclusion, combining true FISP with undersampled radial k-space sampling techniques for application in real-time cardiac cine MR imaging and breath-hold segmented cine MR imaging is feasible and appears to be promising. Temporal resolution was improved considerably for both segmented and real-time radial cine sequences by sharing views, without affecting the spatial resolution and without introducing noticeable artifacts. No substantial difference in SNR was seen between the images obtained with the interleaved or continuous segmented radial true FISP sequences. However, the use of interleaved segmentation led to increased streaking on images when compared with the use of continuous segmentation. Results with the real-time radial sequence show that interleaved sharing is better suited for real-time cardiac cine studies of ventricular wall motion because of the method of sharing.


   ACKNOWLEDGMENTS
 
We thank Oliver Heid (Siemens Medical Systems, Erlangen, Germany) for providing the code for vector regridding, which helped us write faster reconstruction code.


   FOOTNOTES
 
Abbreviations: CNR = contrast-to-noise ratio, FISP = fast imaging with steady-state precession, FOV = field of view, SNR = signal-to-noise ratio, 2DFT = two-dimensional Fourier transform

Author contributions: Guarantor of integrity of entire study, J.L.D.; study concepts and design, A.S., O.P.S., G.L.; literature research, A.S., O.P.S.; clinical studies, J.S.L.; experimental studies, A.S., O.P.S., G.L.; data acquisition and analysis/interpretation, A.S., O.P.S., G.L.; manuscript preparation, definition of intellectual content, editing, and revision/review, all authors; manuscript final version approval, J.L.D.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
 Discussion
 REFERENCES
 

  1. Finn JP, Edelman RR. Black-blood and segmented k-space magnetic resonance angiography. Magn Reson Imaging Clin N Am 1993; 1:349-357.
  2. Schulen V, Schick F, Loichat J, et al. Evaluation of k-space segmented cine sequences for fast functional cardiac imaging. Invest Radiol 1996; 31:512-522.
  3. Laub G. Breath-hold cine MR imaging with a shared and reordered gradient echo technique (abstr) In: Book of abstracts: Society of Magnetic Resonance in Medicine 1993. Berkeley, Calif: Society of Magnetic Resonance in Medicine, 1993; 478.
  4. Lauterbur PC. Image formations by induced local interactions: examples employing nuclear magnetic resonance. Nature 1973; 242:190-191.
  5. Kumar A, Welti D, Ernst RR. NMR Fourier zeugmatography. J Magn Reson 1975; 18:69-83.
  6. Bergin CJ, Pauly JM, Macovski A. Lung parenchyma: projection reconstruction MR imaging. Radiology 1991; 179:777-781.
  7. Brossmann J, Frank LR, Pauly JM, et al. Short echo time projection reconstruction MR imaging of cartilage: comparison with fat-suppressed spoiled GRASS and magnetization transfer contrast MR imaging. Radiology 1997; 203:501-507.
  8. Noll DC, Pauly JM, Nishimura DG, Macovski A. Magnetic resonance reconstruction from projections using half the data. SPIE Med Imaging V: Image Physics 1991; 1443:29-36.
  9. Hafner S. Fast imaging in liquids and solids with the back-projection low angle shot (BLAST) technique. Magn Reson Imaging 1994; 12:1047-1051.
  10. Cremillieux Y, Briguet A, Deguin A. Projection reconstruction methods: fast imaging sequences and data processing. Magn Reson Med 1994; 32:23-32.
  11. Glover GH, Pauly JM. Projection reconstruction technique for reduction of motion effects in MRI. Magn Reson Med 1986; 4:359-376.
  12. Glover GH, Lee AT. Motion artifacts in fMRI: comparison of 2DFT with PR and spiral scan methods. Magn Reson Med 1995; 33:624-635.
  13. Rasche V, Holz D, Schepper W. Radial turbo spin echo imaging. Magn Reson Med 1994; 2:629-638.
  14. Rasche V, Holz D, Proksa R. MR fluoroscopy using radial multi-gradient echo (rMGE) MRI. Magn Reson Med 1995; 34:754-761.
  15. Rasche V, Holz D, Kohler J, Proksa R, Roschmann RP. Catheter tracking using continuous radial MRI. Magn Reson Med 1997; 37:963-968.
  16. Rasche V, de Boer RW, Holz D, Proksa R. Continuous radial data acquisition for dynamic MRI. Magn Reson Med 1995; 34:754-761.
  17. Bucker A, Neurberg JM, Adam GB, et al. Stent placement with real time MRI guidance: initial animal experiment experiences. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1998; 169:655-657[German].
  18. Shimizu K, Mulkern RV, Oshio K, et al. Rapid tip tracking with MRI by a limited projection reconstruction technique. J Magn Reson Imaging 1998; 8:262-264.
  19. Miller S, Hahn U, Bail DM, et al. Cardiac MRI for determining functional left ventricular parameters. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1999; 170:47-53[German].
  20. Onishi S, Fukui S, Atsumi C, et al. Clinical evaluation of regurgitant blood flow by rapid cine magnetic resonance imaging in patients with valvular heart disease. J Cardiol 1989; 19:571-582[Japanese].
  21. Duerk JL, Lewin JS, Wendt M, Petersilge C. Remember true FISP? a high SNR, near 1-second imaging method for T2-like contrast in interventional MRI at .2 T. J Magn Reson Imaging 1998; 8:203-208.
  22. Heid O. True FISP cardiac fluoroscopy (abstr) In: Proceedings of the Fifth Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1997; 320.
  23. Bundy J, Simonetti O, Laub G, Finn JP. Segmented true FISP cine imaging of the heart (abstr) In: Proceedings of the Seventh Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1999; 1282.
  24. Carr J, Simonetti O, Kroeker R, Bundy J, Pereles S, Finn JP. Segmented true FISP: an improved technique for cine MR angiography (abstr) In: Proceedings of the Eighth Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2000; 199.
  25. Carr JC, Simonetti O, Bundy J, Li D, Pereles S, Finn JP. Cine MR angiography of the heart with segmented true fast imaging with steady-state precession. Radiology 2001; 219:828-834.
  26. Shankaranarayanan A, Wendt M, Aschoff AJ, Lewin JS, Duerk JL. Radial keyhole sequences for low field projection reconstruction interventional MRI. J Magn Reson Imaging 2001; 13:142-151.
  27. Sakuma H, Caputo GR, Steffens JC, et al. Breath-hold MR cine angiography of coronary arteries in healthy volunteers: value of multiangle oblique imaging planes. AJR Am J Roentgenol 1994; 163:533-537.
  28. Atkinson DJ, Edelman RR. Cineangiography of the heart in a single breath hold with a segmented turbo FLASH sequence. Radiology 1991; 178:357-360.
  29. Diemling M, Heid O. Magnetization-prepared true FISP imaging (abstr) In: Proceedings of the First Meeting of the Society of Magnetic Resonance. Berkeley, Calif: Society of Magnetic Resonance, 1994; 495.
  30. Oesterle C, Markl M, Strecker R, Kraemer FM, Hennig J. Spiral reconstruction by regridding to a large rectilinear matrix: a practical solution for routine systems. J Magn Reson Imaging 1999; 10:84-92.
  31. Scheffler K, Hennig J. Reduced circular field-of-view imaging. Magn Reson Med 1998; 40:474-480.
  32. Weiss S, Rasche V. Projection-reconstruction reduced FOV imaging. Magn Reson Imaging 1999; 17:517-525.
  33. Lauzon ML, Rutt BK. Polar sampling in k-space. Magn Reson Med 1998; 40:769-782.
  34. Lauzon ML, Rutt BK. Effects of polar sampling in k-space. Magn Reson Med 1996; 36:940-949.



This article has been cited by other articles:


Home page
 Am. J. Neuroradiol.Home page
S.K. Yrjana, H. Tuominen, A. Karttunen, N. Lahdesluoma, E. Heikkinen, and J. Koivukangas
Low-Field MR Imaging of Meningiomas Including Dynamic Contrast Enhancement Study: Evaluation of Surgical and Histopathologic Characteristics
AJNR Am. J. Neuroradiol., November 1, 2006; 27(10): 2128 - 2134.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
J. P. Finn, K. Nael, V. Deshpande, O. Ratib, and G. Laub
Cardiac MR Imaging: State of the Technology.
Radiology, November 1, 2006; 241(2): 338 - 354.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
G. A. Krombach, J. G. Pfeffer, S. Kinzel, M. Katoh, R. W. Gunther, and A. Buecker
MR-guided Percutaneous Intramyocardial Injection with an MR-compatible Catheter: Feasibility and Changes in T1 Values after Injection of Extracellular Contrast Medium in Pigs
Radiology, May 1, 2005; 235(2): 487 - 494.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
R. R. Edelman
Contrast-enhanced MR Imaging of the Heart: Overview of the Literature
Radiology, September 1, 2004; 232(3): 653 - 668.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
H. P. Kuhl, E. Spuentrup, A. Wall, A. Franke, J. Schroder, N. Heussen, P. Hanrath, R. W. Gunther, and A. Buecker
Assessment of Myocardial Function with Interactive Non-Breath-hold Real-time MR Imaging: Comparison with Echocardiography and Breath-hold Cine MR Imaging
Radiology, April 1, 2004; 231(1): 198 - 207.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
E. Spuentrup, J. Schroeder, A. H. Mahnken, T. Schaeffter, R. M. Botnar, H. P. Kuhl, P. Hanrath, R. W. Gunther, and A. Buecker
Quantitative Assessment of Left Ventricular Function with Interactive Real-Time Spiral and Radial MR Imaging
Radiology, June 1, 2003; 227(3): 870 - 876.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Roentgenol.Home page
B. Giorgi, S. Dymarkowski, F. Maes, M. Kouwenhoven, and J. Bogaert
Improved Visualization of Coronary Arteries Using a New Three-Dimensional Submillimeter MR Coronary Angiography Sequence with Balanced Gradients
Am. J. Roentgenol., October 1, 2002; 179(4): 901 - 910.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Quicktime Movies
Right arrow All Versions of this Article:
2213010455v1
221/3/827    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Shankaranarayanan, A.
Right arrow Articles by Duerk, J. L.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Shankaranarayanan, A.
Right arrow Articles by Duerk, J. L.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE