| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Medicine, Baylor College of Medicine, Houston, TX.
2 Philips Medical Systems, Cleveland, OH.
3 Department of Radiology, Baylor College of Medicine, Houston, TX.
4 Department of Radiology, Texas Heart Institute at St. Luke's Episcopal
Hospital, 6720 Bertner Ave., MC 2-270, Houston, TX 77030.
5 Department of Cardiology, Texas Heart Institute at St. Luke's Episcopal
Hospital, Houston, TX.
Received February 1, 2006;
accepted after revision May 16, 2006.
S. D. Flamm is partially supported by research grants from Philips Medical
Systems, Siemens Medical Solutions, GE Healthcare, Tyco
Healthcare/Mallinckrodt, and Bracco Diagnostics.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. The LV function of 22 healthy volunteers and 15 patients was evaluated using a standard 2D SSFP sequence and a 3D SSFP sequence with SENSE at 1.5 T. Ventricular volume, ejection fraction, and LV mass were calculated with each method, and signal-to-noise ratios (SNRs) and myocardium-to-blood contrast-to-noise ratios (CNRs) were measured. Agreement between the two methods was assessed using Bland-Altman analysis, and results were compared using a paired Student's t test (p < 0.05). The local institutional review board approved the study protocol, and all participants gave signed informed consent. The study complied with the Health Insurance Portability and Accountability Act.
RESULTS. Both techniques produced similar estimates of ejection fraction (mean bias ± SD, -1.2% ± 3.6%) and LV mass (mean bias, ± SD-1.2 ± 10.9 g). No significant differences were found in calculated volumes, ejection fraction, or LV mass between the two methods. Acquisition time was reduced by 82%, to a single breath-hold (18 ± 3 seconds), with the 3D SSFP technique. SNR and CNR were significantly lower with the 3D method than with the standard method.
CONCLUSION. Three-dimensional SSFP imaging with SENSE can reduce acquisition time to a single breath-hold and can provide LV function quantification comparable to that obtained with conventional 2D SSFP imaging.
Keywords: 2D imaging • 3D imaging • cine • cardiovascular imaging • left ventricular function • MRI • sensitivity encoding • SSFP imaging
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Several strategies for accelerating image acquisition have been studied. Various 3D imaging techniques, such as SSFP with short TRs [10], projection reconstruction [11, 12], and spatio-temporal frequency (k - t) broad-use linear acquisition speed-up technique (k-t BLAST) [13], have been used to assess LV function in a single breath-hold. However, the clinical usefulness of some of these techniques is questionable because they require long breath-holds, which may be challenging for patients.
Parallel imaging techniques such as sensitivity encoding (SENSE) use multiple receiver coils to obtain images more rapidly, but one disadvantage of these techniques is a decreased signal-to-noise ratio (SNR) [14, 15]. In a comparative study, 2D SSFP combined with SENSE significantly reduced acquisition time without sacrificing spatial resolution, temporal resolution, and myocardium-to-blood contrast when compared with 2D SSFP without SENSE [16]. The results of Bland-Altman analysis from this study revealed that the two techniques provided comparable assessment of ejection fraction (mean bias ± SD, -0.2% ± 1.4%) and LV mass (mean bias ± SD, 1.7 ± 4.2 g). The acquisition time for the SENSE-assisted 2D SSFP was 140 ± 21 seconds; for the conventional 2D SSFP sequence, it was 278 ± 37 seconds. However, multiple breath-holds were still required with SENSE-assisted 2D SSFP.
Three-dimensional SSFP combined with SENSE may reduce the time required to obtain the image and may permit the use of a single breath-hold for image acquisition. Signal intensity is higher with 3D imaging than with 2D imaging because a larger volume of tissue is excited with each pulse. This increased signal can be traded for even greater acceleration with SENSE, thereby enabling a single breath-hold examination. This study tested the hypothesis that 3D cine SSFP imaging with SENSE can reduce acquisition time to a single breath-hold while allowing quantitative assessment of LV function comparable to that obtained with standard 2D cine SSFP imaging.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MRI Data Acquisition
All imaging was performed with a 1.5-T MR scanner (Achieva, Philips Medical
Systems) with a five-element synergy cardiac coil and vector cardiographic
gating. One of two operators performed all examinations. After initial scout
imaging and a reference acquisition, both the standard 2D cine SSFP sequence
and the SENSE-assisted 3D cine SSFP sequence were used to obtain a series of
short-axis sections to cover the entire left ventricle.
Imaging parameters for the 2D SSFP sequence were as follows: TR/TE, 2.7-3.1/1.4-1.5; flip angle, 65°; temporal resolution, 25-39 milliseconds; in-plane resolution, 1.9 x 1.9 to 2.6 x 2.7 mm (depending on patient size); mean value, 2.2 ± 0.2 x 2.2 ± 0.2 mm; breath-hold duration, 10-12 heartbeats per section acquired. The slice thickness of each slice was 8 mm with an interslice gap of 2 mm.
Imaging parameters for the SENSE-assisted 3D SSFP sequence were as follows: 2.4-2.6/1.2-1.3; flip angle, 55°; acquired temporal resolution, 76 milliseconds; reconstructed temporal resolution, 40-53 milliseconds; in-plane resolution, 2.1 x 2.2 to 2.9 x 3.2 mm (depending on patient size), mean value, 2.4 ± 0.2 x 2.5 ± 0.2 mm; breath-hold duration, 18-20 heartbeats for complete acquisition of LV volume. A tailored radiofrequency pulse was used to excite a slice volume that was slightly greater than the imaging volume. The spatial encoding along the slice direction encoded a total of 14 slices, and two slices on either side of the imaging volume were discarded on reconstruction. In addition, for the 3D SSFP sequence a SENSE factor of 3.0 was used, and partial Fourier sampling was used to further reduce acquisition time.
We obtained a total of 74 complete measurements of LV volume: 37 with standard 2D SSFP and 37 with SENSE-assisted 3D SSFP. We used the breathhold duration and the number of breath-holds necessary to obtain images of the entire left ventricle to calculate the acquisition time for each sequence.
The SENSE reconstruction requires information about the coil-sensitivity profiles, as described by Pruessman et al. [14]. We estimated the coil sensitivities in the standard manner as previously described by Kacere et al. [16] in their study of SENSE-assisted 2D SSFP. The total acquisition time of the reference examination was 53 seconds.
Data Analysis
All 74 image data sets of LV function were analyzed at a postprocessing
workstation (EasyVision release 5.0, Philips Medical Systems). An observer
with 2 years of experience in cardiac MRI drew endocardial and epicardial
contours on each section of the left ventricle at end diastole and end
systole. Based on these contours, voxel summation was used to calculate
end-diastolic volume (EDV), endsystolic volume (ESV), and LV mass. Stroke
volume (SV) was calculated by using the formula SV = EDV -
ESV, and ejection fraction (EF) was calculated by using the formula
EF = (SV/EDV) x 100. Automated calculation of ejection
fraction and LV mass was not attempted because the results from manual
calculations have been more reliable and reproducible in our laboratory.
Reproducibility data for this laboratory have been described previously in the
assessment of LV function, showing close agreement between the two techniques
and low interobserver variability
[16].
Irregular regions of interest (ROIs) were drawn on the central short-axis section at end diastole. The ROIs included the septal myocardium (124-416 mm2), LV blood (208-489 mm2), skeletal muscle in the shoulder region (93-395 mm2), and air space in the lungs (215-496 mm2), which did not include any major blood vessels. The average and SD of the signal intensity from the ROIs were calculated. Although the conventional measure of noise is the SD of signal intensity in regions outside the body, such measurements may result in artificially low estimates of noise in SENSE reconstruction [14]. To avoid misrepresenting the actual SNR calculations, we used the statistics from air space in the lungs as a surrogate measure for noise. The SNR of tissues was computed as the ratio of mean signal intensity of tissue to the SD of noise. The blood-to-muscle contrast-to-noise ratio (CNR) was computed as the ratio of the difference between the mean signal intensity of LV blood and myocardium to the SD of noise.
Statistics
All data are presented as mean ± SD. We used the Bland-Altman method
[17] to assess the agreement
between the standard 2D SSFP and the SENSE-assisted 3D SSFP measurements of LV
function (EDV, ESV, ejection fraction, and LV mass). The paired Student's
t test was used to assess statistical significance. A p
value of less than 0.05 was considered statistically significant.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
|
|
|
Both blood SNR and myocardium-to-blood CNR were significantly lower with SENSE-assisted 3D SSFP than with 2D SSFP acquisitions (Table 2).
|
Acquisition time was 100 seconds ± 8 seconds with the 2D SSFP sequence and 18 seconds ± 3 seconds with the SENSE-assisted 3D SSFP sequence (p < 0.001), indicating an 82% reduction in acquisition time with SENSE-assisted 3D SSFP.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Because of its high SNR and superior blood-to-muscle contrast, SSFP imaging is the method of choice for the quantitative assessment of LV function with cine MRI [6-8]. However, to cover the LV volume, conventional techniques require 10 to 12 breath-holds, which may lead to patient restlessness, examination complexity, and inaccuracy caused by slice misregistration between breath-holds. Thus, a technique that yields comparable assessment in a single breath-hold is useful.
By combining 3D SSFP imaging with SENSE, we have been able to assess LV function in a single breath-hold. It has been shown previously that SENSE-assisted 2D SSFP is accurate and takes half the time to obtain images compared with conventional 2D SSFP; however, multiple breath-holds were still required [16]. SENSE uses sensitivity data from multiple receiver coils to undersample k-space and accelerate acquisition, but usually with an inherent loss in SNR [14]. This SNR loss limits the extent to which SENSE acceleration factors may be used before image degradation becomes extreme. Obtaining data in a volumetric fashion with 3D SSFP intrinsically improves SNR, and this increased signal can be readily traded for higher SENSE acceleration factors. In this study, we combined a volumetric SSFP acquisition with a SENSE factor of 3.0, in conjunction with partial Fourier sampling, to reduce the acquisition time to a single breath-hold of 13 to 25 seconds (depending on heart rate). This approach reduced acquisition time to approximately one-sixth that of the 2D technique. In clinical practice, patients require time to recover between breath-holds (10-14 seconds per slice); therefore, the effective reduction in scanning time with the single breath-hold 3D technique is even greater. However, the reference scan (53 seconds) used for SENSE reconstruction may partly offset this reduction in acquisition time.
From a workflow perspective, all data were acquired on a commercially available MR scanner that had an integrated SENSE acquisition and reconstruction, without the need for any modifications. The reference scan was the only additional requirement for the clinical application of SENSE. The reconstruction for the SENSE-accelerated 3D SSFP sequence was online and was automatically completed within a few seconds after data acquisition. Interestingly, because the raw data for the SENSE acquisition is undersampled, the computational cost of a SENSE reconstruction is lower than that of a conventional fully sampled Fourier transformation reconstruction. A commercially available five-element phased-array surface coil was used for data acquisition and required no special considerations for coil positioning beyond that of routine cardiac examinations. It is conceivable that a surface coil with a greater number of coil elements might provide greater benefit for parallel imaging, although at present none is available in this laboratory.
Spatial and temporal resolutions were slightly lower with the SENSE-assisted 3D SSFP method than with the 2D method because of some blurring caused by increased pixel size and cardiac motion. In addition, some residual unfolding artifacts related to SENSE were noted near the basal slices. However, the quantitative evaluation of LV function was not affected by reduced image quality or the presence of artifacts.
The SNR was expected to be higher with the 3D SSFP technique than with the 2D technique because of the larger voxel size (voxel volume, 62.5 mm3 for 3D SSFP vs 48.4 mm3 for the 2D SSFP technique) and volumetric coverage. To gain acquisition speed, we traded off some of the increase in SNR with the 3D SSFP technique by the addition of SENSE (SNR reduced by a factor of 3), the use of a higher bandwidth (2,235 Hz/pixel for 3D SSFP vs 1,250 Hz/pixel for the 2D SSFP), and additional noise secondary to the coil geometry factor [14]. Despite adding these factors, which reduce SNR, the 3D SSFP technique should still be expected to have an advantage in SNR over the 2D SSFP technique. However, we found that LV blood SNR was 36% lower and CNR was 40% lower with the SENSE-assisted 3D SSFP acquisition compared with 2D SSFP.
We believe that the blood SNR of the 3D SSFP technique was substantially reduced because of the saturation of blood magnetization with volumetric excitation and because of the continuous mixing of LV blood with various spin histories throughout the cardiac cycle within the excited volume. This reduction in blood SNR contributed to the reduction in the blood-to-myocardium CNR. Surprisingly, a 20% reduction (p = 0.007) occurred in the myocardial SNR with the 3D SSFP technique. The myocardial SNR reduction in the patient subgroup was not statistically significant. However, the SNR of the skeletal muscle was 20% higher (p = 0.001) for the 3D SSFP technique, which agrees with theoretic predictions. We believe the myocardial SNR, as opposed to the muscle SNR, may be lower in the 3D technique because the blood volume fraction is greater within the myocardium than in resting skeletal muscle. These results suggest that the loss of blood SNR is primarily caused by flow-related saturation associated with volumetric acquisition; therefore, the loss of blood SNR caused by saturation might be minimized if the 3D SSFP images were obtained after contrast agent administration. Nevertheless, the close agreement between the results obtained with 2D SSFP and 3D SSFP with SENSE in this study suggests that the loss of SNR and CNR is tolerable and does not alter the evaluation of LV function.
To our knowledge, no studies have been published comparing 2D SSFP and SENSE-assisted 3D SSFP in the quantitative evaluation of LV function. Preliminary data have shown that the two methods produce similar estimations of ejection fraction, but significantly different measurements of LV mass (Kassner A, et al., presented at the 2003 annual meeting of the International Society for Magnetic Resonance in Medicine). However, the population for this study comprised only eight healthy volunteers, and acquisition times and estimates of SNR and CNR were not reported.
Several studies have assessed the use of 3D cine MRI for evaluating LV function in a single breath-hold; however, most have been feasibility studies of 10 or fewer individuals. Although 3D SSFP with a short TR (2.4-2.8) provides high spatial resolution, SNR, and CNR, the prolonged breath-hold (20-46 heartbeats) required for this technique may be poorly tolerated by patients [10]. Three-dimensional SSFP with projection reconstruction has been compared with standard 2D SSFP imaging and found to offer high spatial resolution and comparable LV function analysis [12]; however, the 3D SSFP technique was associated with radial streak artifacts and significant decreases in SNR and CNR. In a small study of the rapid assessment of LV function, k-t BLAST combined with 3D cine MRI provided high spatial and temporal resolution [13]. Because this technique requires a brief training scan and an undersampled high-resolution scan, reconstruction errors can be introduced if these two scans are performed in separate breath-holds [13]. Despite this limitation, k-t BLAST permits rapid examination and can produce high-quality images in a single, short breath-hold, particularly if combined with partial Fourier sampling [13].
A limitation of the present study was the lack of a reference standard to validate the measurements of ejection fraction and LV mass with the 2D SSFP and with the SENSE-assisted 3D SSFP acquisitions. Cardiac MRI is highly accurate in estimating LV function [4, 5], and, in current clinical practice, 2D SSFP is the sequence of choice for the assessment of LV function because of its superior temporal resolution, image quality, and blood-to-myocardium CNR when compared with standard gradient-echo sequences [6-8]. However, in most of these studies, ventricular volumes measured with SSFP were significantly higher than those measured with gradient-echo sequences, and it is unclear which sequence is more accurate [7, 18]. Nevertheless, SSFP has been shown to be highly accurate in animal models [19, 20], a phantom [21], and against other controls [22], although its position as the reference standard remains controversial. Larger studies involving animal models or phantom hearts are necessary to make such an assertion.
In summary, this study has shown that 3D SSFP imaging with SENSE can reduce acquisition time to a single breath-hold and produce quantitative measurements of LV function comparable to those obtained with conventional 2D SSFP imaging. This shorter acquisition time can reduce patient examination time and make cine MRI assessment of LV function more tolerable for patients.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  R. J. Gibbons, P. A. Araoz, and E. E. Williamson The Year in Cardiac Imaging J. Am. Coll. Cardiol., September 4, 2007; 50(10): 988 - 1003. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
