| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 All authors: Departments of Radiology and Clinical Neurosciences, University of Calgary, The Seaman Family MR Research Centre, Foothills Medical Centre, 1403-29th St. NW, Calgary, AB, Canada T2N 2T9.
Received July 5, 2005;
accepted after revision August 9, 2005.
Presented in part at the 2004 annual meeting of the International Society
of Magnetic Resonance in Medicine, Kyoto, Japan.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Abdominal MRI protocols include various T1-weighted, T2-weighted, and contrast-enhanced acquisitions that require extended spatial coverage and resolution. Ideally, each acquisition is completed within one breath-hold. At 3.0 T, power deposition (i.e., specific absorption rate [SAR]) concerns can limit achieving these conflicting needs because conventional sequences are based on 6-minute time-average SAR requirements. We optimized abdominal-specific sequences based on an approved short-term 10-second time-average SAR criterion and added a delay time after breath-holding to fulfill the long-term 6-minute time-average power deposition regulation.
RESULTS. Using our strategy, image acquisition time efficiency at 3.0 T was increased approximately twofold compared with conventional abdominal breath-hold pulse sequences for 2D dual-echo gradient-recalled echo, single-shot fast spin-echo, and 3D steady-state free precession sequences. Volunteers experienced a slight sensation of warmth for the single-shot fast spin-echo implementation, the most SAR-intensive sequence.
CONCLUSION. Our optimization strategy is not vendor-specific, is easily implemented for all conventional scanners (provided one can access and modify the pulse sequences directly, or the vendors can make the necessary changes), yields a higher slice-per-unit-time imaging efficiency, and still satisfies all the regulatory power deposition requirements.
Keywords: abdominal imaging • MR technique • MRI
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
25 seconds). Acquisition during a breath-hold minimizes motion artifact
and spatial misregistration of lesions
[1]. Typical abdominal
protocols include T1-weighted, dual-echo gradient-recalled echo scanning (to
observe signal loss due to fatty infiltration
[2] or iron overload
[3] and to assess fat content);
T2-weighted imaging (to characterize lesions such as cysts, hemangiomas, and
metastases [4]); and
multiphasic contrast-enhanced fast 3D gradient-recalled echo acquisitions (to
elucidate static and dynamic enhancement characteristics of tissues and to
visualize vessels [5,
6]). Steady-state free
precession (SSFP) or true fast imaging with steady-state free precession (true
FISP) [7] also can be used,
with or without a contrast agent, to assess lesions and metastases
[8,
9]. T1 weighting typically is achieved using 2D gradient-recalled echo with a TR of 150-200 milliseconds in combination with a large flip angle (50-80°). The two TEs are chosen so that fat and water peaks are opposed-phase and in-phase, respectively. Blood flow must be suppressed to avoid flow-related ghosting artifact in the phase-encoding direction [10]. T2-weighted images are most often acquired with fast spin-echo and single-shot fast spin-echo sequences [11] with or without fat suppression.
High-field 3.0-T MRI can potentially improve abdominal diagnosis by increasing the signal-to-noise ratio (SNR), resolution, and coverage and by minimizing scanning times. Higher field strength results in higher power deposition (i.e., specific absorption rate [SAR]) [12], especially for multislice large-flip-angle acquisitions, fast spin-echo-based techniques, short-TR SSFP, and spatially or chemically saturated scans. To acquire the data in one breath-hold, one may have to compromise either spatial coverage or resolution, avoid saturation altogether (which may lead to flow-related ghosting artifacts or suboptimal tissue-contrast differentiation), or use smaller flip angles (which may reduce SNR or tissue contrast).
Our optimization approach is based on the realization that pulse sequences
must meet two SAR criteria, namely, long-term and short-term power deposition
requirements [13,
14]. The U.S. Food and Drug
Administration (FDA) 1998 guidelines stipulate that if the International
Electrotechnical Commission (IEC) revises its sections on operating SAR
limits, then FDA levels are to be adjusted to conform accordingly
[13]. The IEC 2002 revised
limits for body MRI are SAR 2.0 W/kg for a 6-minute time-average and SAR
within 6.0 W/kg during any given 10-second interval
[14]. The short-term 10-second
time-average represents a threefold increase in allowed power deposition
compared with the 6-minute time-average.
In general, vendors calculate minimum sequence times (or, equivalently, the maximum number of slices per unit time) based on gradient timing limitations (i.e., how closely and rapidly gradients can be turned on and off) and the 6-minute time-average SAR, because this also guarantees satisfying the 10-second time-average SAR requirement.
However, because abdominal MR images are ideally acquired within about a 25-second breath-hold, we propose to maximize scanning efficiency by optimizing sequence times based on gradient timing limitations and the 10-second (as opposed to 6-minute) time-average SAR requirement [15, 16]. We then ensure compliance with the 6-minute time-average SAR criterion by intentionally programming a delay into each imaging sequence at the conclusion of the breath-hold. This delay acts as both a cooling period for the patient and a rest between acquisitions.
Our goals are to show the potential time, spatial coverage, and resolution benefits at 3.0 T for abdominal breath-hold MRI, to assess the feasibility of implementation, and to illustrate the potential clinical benefits while posing no added heating risks to the patient.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
grad and 6min, respectively). The
manufacturer's FDA and IEC configuration files report identical peak
whole-body SAR limits at 3.0 T (2.0 W/kg for a 6-minute time-average), and the
short-term power monitors are calibrated to accept three times that value in a
10-second interval. The power monitor detects and records the power sent to the patient (Pout) and the power received within the coil (Prec). The difference between Pout and Prec is assumed to be the net power deposited in the patient—that is, the instantaneous SAR. The short- and long-term power monitors continuously calculate the average SAR in the last 10-second and 6-minute intervals, respectively, and they may or may not display these values on the console. If the monitored short- or long-term SAR values exceed their respective FDA/IEC limits, the sequence either pauses until restarted by the user or stops altogether so as not to overheat or injure the patient.
For 2D gradient-recalled echo acquisition, multiple slices are excited in one TR. The sequence time is defined as the minimum time allowed between consecutive radiofrequency excitations, each of which yields one line of k-space data at a different slice location. Analogously, 2D fast spin-echo is also a multislice acquisition, and its sequence time is defined as the time between consecutive 90° radiofrequency excitations, each of which excites a different slice location. The 2D single-shot fast spin-echo sequence acquires all the lines of k-space for a given slice in one TR, but the single-shot fast spin-echo sequence time is also given by the time between consecutive 90° radiofrequency excitations—namely, the slice-to-slice time or TR. The sequence time for the 3D true FISP acquisition is its TR value—that is, the time between consecutive radiofrequency pulses.
Our SAR-specific efficiency changes in each sequence included calculating
the 10-second time-average for the SAR sequence time (10sec =
1.05 x 6min / 3 [the allowed threefold increase in power
deposition reduces the sequence time by a factor of 3, but we add an extra 5%
margin of safety, hence the 1.05 multiplier]); setting pulse sequence time
constraints accordingly (i.e., internally selecting the greater of
grad and 10sec); programming a required delay
time (delay) after the breath-hold acquisition to satisfy the
6-minute time-average SAR requirement; and displaying both the imaging (i.e.,
breath-hold) time and the total scanning time by modifying the user
display.
Because each time-efficient sequence acquires the breath-hold images using
10-second (as opposed to 6-minute) time-average SAR requirements and because
typical abdominal protocols require many different acquisitions, we could
potentially exceed the 6-minute time-average SAR limitations before the
abdominal protocol is completed. However, by enforcing a delay time after the
imaging portion of each acquisition, we give the patient a rest period between
successive sequences and we ensure that the 6-minute time-average SAR
conditions are met. In effect, the scanner is in use, but no radiofrequency or
gradient pulses are played out. The minimum delay time to satisfy the 6-minute
time-average SAR is calculated as the difference between
Max(grad, 6min) and Max(grad,
10sec) [where Max (A,B) represents the greater of A and B]
times the total number of excitations needed—that is, (number of slices)
x (number of phase encodes) x (number of averages). The total
scanning time for each particular sequence is thus the image data acquisition
time (which equals the breath-hold time) plus the delay time.
Note that the total scanning times are similar for the conventional and time-efficient implementations for identical parameters. The main difference is that the conventional sequences require a breath-hold throughout the entire acquisition; this may either extend beyond 25 seconds or require a reduction in the number of slices or phase encodes. In short, our approach concentrates the imaging (breath-hold) portion in a rapid acquisition (albeit at a higher SAR but still within short-term SAR limits) followed by a delay time to satisfy long-term SAR constraints.
The 2D dual-echo gradient-recalled echo sequence was further modified by providing more choices of sampling rate and resolution in the readout direction (to allow greater flexibility in obtaining the fat and water opposed-phase and in-phase TEs) and permitting variable-rate application of spatial saturation (to minimize breath-hold time and SAR yet still achieve adequate saturation).
The institutional review board reviewed and approved the study protocol to scan the abdomens of 10 healthy volunteers (six men and four women [age range, 20-45 years; weight range, 55-105 kg]). All subjects gave written informed consent before imaging. To minimize motion artifact due to diaphragmatic drift, the images were acquired during a breath-hold at end-expiration. The protocol consisted of 2D axial dual-echo gradient-recalled echo, 2D coronal single-shot fast spin-echo with fat saturation, 2D axial fast spin-echo, 2D axial single-shot fast spin-echo, and 3D axial true FISP sequences. Imaging parameters are given in Table 1.
|
For each sequence, two investigators acquired both the conventional and the time-efficient implementations before moving to the next acquisition. Implementation ordering of conventional first and time-efficient second or time-efficient first and conventional second was randomized for each acquisition set and volunteer. The sequences (i.e., dual-echo gradient-recalled echo, fast spin-echo, single-shot fast spin-echo, and true FISP) were also acquired in random order for the different volunteers, and the sequence and implementation orders were unknown to each volunteer to prevent bias. The number of slices was adjusted so that the imaging (i.e., breath-hold) times were about 20 seconds. Otherwise, the acquisition parameters were identical (except for the 2D dual-echo gradient-recalled echo, for which the TEs, receiver bandwidth, and matrix size were automatically set in the conventional acquisition; see Table 2). After every sequence set, each volunteer was asked to subjectively rate whether one implementation felt warmer and if any scanning sequence felt uncomfortable.
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Dual-Echo Gradient-Recalled Echo
As expected, the liver-spleen contrast at 3.0 T is different from that at
1.5 T because of changes in relaxation times
[17]. Still, the opposed-phase
images (i.e., the second of the two echo times [TE = 5.8 milliseconds] for the
conventional implementation [Fig.
1B] but the first of the two echo times [TE = 1.3 milliseconds] in
our time-efficient sequence [Fig.
1C]) show the characteristic signal loss at fat-tissue interfaces
compared with the inphase images (Figs.
1A and
1D, respectively).
The average SNR and CNR values are given in Table 3. The opposed-phase implementations are not statistically different (p > 0.10), whereas the conventional in-phase gradient-recalled echo SNRs are statistically greater (p < 0.05) than the time-efficient implementation. Table 2 shows the temporal and spatial advantages of the time-efficient versus the conventional 2D dual-echo gradient-recalled echo implementations, along with our modifications to allow flexibility in selecting the desired TEs, the receiver bandwidth, and the rate of spatial saturation. The conventional sequence, with an equivalent breath-holding time (19 seconds), permits only about 50% of the spatial coverage, whereas an identical spatial extent (20 slices) using the conventional sequence requires a long breath-holding time of 38-58 seconds, depending on patient weight.
|
Fast Spin-Echo and Single-Shot Fast Spin-Echo
The average SNR and CNR characteristics of the fast spin-echo and
single-shot fast spin-echo acquisitions are given in
Table 3. Neither the SNR nor
the CNR values between the two implementations were statistically different
(p > 0.20) for the fast spin-echo implementations. However, the
single-shot fast spin-echo SNR differences between the two implementations
were statistically different (p < 0.01), with the time-efficient
implementation showing about a 50% decrease in SNR even though the only
parameter difference was a change in TR (see
Table 4). Representative
conventional and time-efficient single-shot fast spin-echo axial images (Figs.
2A and
2B) showed subtle tissue
contrast differences—that is, hypointensity—in the major organs
(liver, spleen, skeletal muscle, and kidney) for the time-efficient images but
visually similar signal intensity between the two implementations for CSF and
fat.
|
|
|
Tables 4 and 5 give the temporal and spatial advantages of our time-efficient single-shot fast spin-echo (with fat saturation) and fast spin-echo implementations versus the conventional sequences, respectively. For single-shot fast spin-echo scans with the same breath-hold time (20 seconds), Table 4 shows that, like the 2D dual-echo gradient-recalled echo sequence, the spatial extent using the conventional sequence is about 50% of that obtained using our time-efficient implementation. The conventional fast spin-echo sequence (Table 5) shows that for subjects weighing more than 39 kg, only 9-12 slices are possible (instead of 13 with the time-efficient implementation); this represents an 8-30% reduction in spatial coverage.
|
True FISP
Table 3 gives the SNR and
CNR measurements for the conventional and time-efficient true FISP
implementations. No statistical difference between the two implementations was
found (p > 0.25), even though the TRs differed
(Table 6). As shown for the 2D
dualecho gradient-recalled echo and single-shot fast spin-echo findings
(Tables 2 and
4, respectively),
Table 6 shows a twofold
increase in spatial or temporal efficiency using our optimized paradigm.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Our time-efficient optimization strategy satisfies all regulatory SAR requirements (both short-term and long-term), is easily implemented for all breath-hold sequences (provided one can access and modify the pulse sequences directly, or the vendors can make the appropriate changes), is not vendor-specific, and poses no added risk to the patient except a slightly increased sensation of warmth for the single-shot fast spin-echo sequence, the most SAR-intensive acquisition. The methodology was successfully implemented for 2D dual-echo gradient-recalled echo, fast spinecho, single-shot fast spin-echo, and 3D true FISP acquisitions in our abdominal protocol. The 3D gradient-recalled echo sequence was also modified accordingly, but because of the small flip angle required (10-15°), power deposition at 3.0 T was not an issue.
Dual-Echo Gradient-Recalled Echo
An important characteristic of dual-echo gradient-recalled echo is the
simultaneous acquisition of both the opposed-phase and the in-phase echoes.
This guarantees that they are inherently coregistered and avoids any
possibility of slice misregistration.
Because dual-echo gradient-recalled echo is used to assess the fat content in lesions, the opposed-phase TE preferably should be less than the in-phase TE to ensure that signal loss in the opposed-phase image relative to the inphase image is directly due to the fat and water peaks being 180° out of phase with one another instead of T2* dephasing effects [18]. This condition was not met with our vendor-provided conventional version (Table 2) and is the reason we added flexibility to our time-efficient 2D dual-echo gradient-recalled echo sequence in addition to the SAR optimization.
To simultaneously acquire the minimum-time fat and water opposed-phase and
in-phase echoes (first-echo TE 
1.2 milliseconds and second-echo TE = 2.3
milliseconds, respectively), we increased the receiver bandwidth and decreased
the number of readout points (Table
2). The SNR in MR images is known to scale as the voxel (volume
element) size and the square root of the data acquisition time (i.e., the
total time that the analog-to-digital receiver is on)
[19]. The time-efficient SNR
(at the same TE) theoretically should be about 0.8 times that of the
conventional SNR, all else being equal. If we scale the noise values of the
conventionally acquired in-phase ROIs by 0.8 (the theoretic SNR factor), then
the SNR measurements of the in-phase images are no longer statistically
different (p > 0.20). By comparison, the opposed-phase SNR
measurements between the two implementations are not statistically different
(p > 0.10), even though the TEs and acquisition parameters differ
(Tables 2 and
3). Although the time-efficient
SNR is expected to decrease by 20% compared with the conventional acquisition
because of the higher receiver band-width and lower readout resolution, the
reduced TE (1.3 vs 5.8 milliseconds) produces much less T2*
dephasing (Figs. 1A,
1B,
1C, and
1D) and compensates for this
loss in SNR.
Fast Spin-Echo and Single-Shot Fast Spin-Echo
Because neither echo spacing nor echo-train length was modified, one would
expect to measure identical signal and contrast characteristics between the
conventional and time-efficient fast spin-echo and single-shot fast spin-echo
implementations. This was indeed the case for multishot fast spin-echo, for
which the SNR and CNR values were not statistically different (p >
0.20).
For single-shot fast spin-echo, however, the SNR differences were statistically significant for all tissues (p < 0.01) and, consequently, so were all the CNR measurements. This was surprising because the contrast characteristics are visually similar between the two implementations (Figs. 2A and 2B), even though the SNR of the time-efficient single-shot fast spin-echo implementation is about 65% that of the conventional single-shot fast spin-echo, and the only parameter difference was the TR.
One investigator performed further experiments on a quality assurance phantom (filled with doped water) using the time-efficient single-shot fast spin-echo sequence. By varying the TR range from 500 to 5,000 milliseconds in steps of 500 milliseconds, SNR increased as TR became larger. Although the cause is not fully known, the authors hypothesize that stimulated echoes, long time-constant eddy currents, imperfect slice profiles, or any combination thereof may be responsible for this TR-dependent signal intensity change.
Although single-shot fast spin-echo acquires an entire slice (i.e., one image) in less than 1 second, full coverage using the conventional implementation extends beyond about 40 seconds. Single-shot fast spin-echo images could be acquired with the patient breathing freely, but one might observe minor motion artifacts in each image, and diaphragmatic motion between images could lead to spatial misregistration, underestimation of lesion size, or missing a lesion altogether [20].
True FISP
The 3D true FISP sequence yields the highest image quality for a short TR
and large flip angle, conditions that make it a SAR-intensive acquisition.
Although the TRs differed between the conventional and time-efficient true
FISP implementations (7.2 vs 2.5 milliseconds, respectively), the SNR and CNR
measurements (Table 3) were not
statistically different (p > 0.25). This is reasonable when one
considers the signal equation for true FISP
[21]; the 3.0-T T1 and T2
values for liver, kidney, and spleen
[17]; and the fact that the
sequence parameters were otherwise identical. The time-efficient true FISP
theoretic SNR is expected to decrease by less than about 5%.
An added benefit of our time-efficient implementation is that the shorter TR leads to better image quality by reducing the banding artifact that arises from off-resonance and susceptibility effects [7] and by suppressing flow-related ghosting artifacts [22] (Figs. 3A and 3B).
|
|
Study Limitations
We recognize that because only 10 volunteers (all of them healthy) were
scanned, the SNR and CNR measurements have large SDs because the statistical
power was low. However, the SNR trends and observations were consistent on the
basis of well-known theoretic expectations (except for the single-shot fast
spin-echo findings).
Our primary goal was to assess the technical feasibility and quantitative SNR and CNR differences between conventional and time-efficient implementations. A qualitative and quantitative image quality analysis ideally should be performed for patients with specific diseases (e.g., cirrhosis or liver metastases), and the diagnostic image quality of each acquisition should be assessed and rated independently and blindly by two or more body MR radiologists. However, for the purposes of this study, we believe that scanning healthy volunteers obviated this image quality analysis, although one investigator/radiologist did visually compare the images between the two implementations and deemed all the time-efficient sequences diagnostically acceptable.
The volunteers' rating of sequence "heating" was subjective—all stated that they experienced some feeling of extra warmth for the single-shot fast spin-echo implementation, and two of the 10 volunteers said they enjoyed the sensation. This study would have benefited from the use of a surface thermometer to support the impressions of the study subjects. However, because sequences are constrained by SAR, we opted to rely on the manufacturer's calibrated short-term power monitors to ensure that these short-term and long-term limits were not exceeded.
To our knowledge, this is the first attempt at optimizing 3.0-T abdominal-specific MRI pulse sequences based on 10-second SAR limitations. Earlier work by Shellock et al. [23] at 1.5 T showed that patients experienced statistically significant (p < 0.05) physiologic responses (e.g., increased skin temperature, heart rate, blood pressure, oxygen saturation, and cutaneous blood flow) to MR acquisitions using a whole-body-averaged SAR of 6.0 W/kg, and they concluded that persons with normal thermoregulatory function can easily tolerate these changes. Their experiments were conducted in patients scanned continuously for 15 minutes. By comparison, we used short-term whole-body SAR of 6.0 W/kg for < 25 seconds at a time, so we anticipate even fewer physiologic effects.
In summary, our time-efficient optimizations offer a more comprehensive and efficient approach to breath-hold abdominal MRI at 3.0 T. Modified 2D dual-echo gradient-recalled echo, 2D single-shot fast spin-echo, and 3D true FISP showed about a twofold increase in spatial or temporal efficiency compared with the conventional implementation. We effectively reduce the total number of breath-holds required, which can directly minimize patient fatigue and may indirectly provide greater patient comfort, less motion artifact, and higher image quality. More important, our methodology satisfies all regulatory SAR criteria without adding heating risks to the patient.
Acknowledgments
We thank the volunteers for their help and the Canada Foundation for
Innovation, which enabled acquisition of the 3.0-T MR scanner.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  R. B. Stafford, M. Sabati, M. J. Haakstad, H. Mahallati, and R. Frayne Unenhanced MR Angiography of the Renal Arteries with Balanced Steady-State Free Precession Dixon Method Am. J. Roentgenol., July 1, 2008; 191(1): 243 - 246. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
