Gastrointestinal Imaging • Original Research
Azevedo et al.
Abdominal MRI
Downloaded from www.ajronline.org by Indiana Univ-Acq Dept on 08/18/14 from IP address 140.182.64.84. Copyright ARRS. For personal use only; all rights reserved
Gastrointestinal Imaging
Original Research
Free-Breathing 3D T1-Weighted
Gradient-Echo Sequence With
Radial Data Sampling in Abdominal
MRI: Preliminary Observations
Rafael M. Azevedo1
Rafael O. P. de Campos1
Miguel Ramalho1
Vasco Herédia1
Brian M. Dale2
Richard C. Semelka1
Azevedo RM, de Campos ROP, Ramalho M, Herédia V, Dale BM, Semelka RC
Keywords: 3D gradient-echo sequence, abdominal MRI,
free breathing, radial acquisition, uncooperative patient
DOI:10.2214/AJR.10.5881
Received September 26, 2010; accepted after revision
January 18, 2011.
1
Department of Radiology, University of North Carolina at
Chapel Hill, 101 Manning Dr, CB 7510, 2006 Old Clinic
Bldg, Chapel Hill, NC 27599. Address correspondence to
R. C. Semelka ([email protected]).
2
Siemens Healthcare, Cary, NC.
AJR 2011; 197:650–657
0361–803X/11/1973–650
© American Roentgen Ray Society
650 OBJECTIVE. The purposes of this study were to evaluate the feasibility of a free-breathing 3D gradient-recalled echo sequence with radial data sampling (radial 3D GRE) in abdominal MRI compared with a standard 3D GRE volumetric interpolated breath-hold examination (VIBE) sequence for imaging of cooperative patients and to perform a preliminary
assessment in imaging of noncooperative patients.
MATERIALS AND METHODS. Fifty-five consecutively registered patients who underwent unenhanced and contrast-enhanced abdominal MRI with the free-breathing radial
3D GRE technique constituted the study population. Two readers independently and blindly
evaluated the images.
RESULTS. Overall image quality with the contrast-enhanced radial 3D GRE sequence
was lower than but rated at least nearly as good as that with the 3D GRE VIBE sequence (p <
0.0001). Higher scores were recorded for 3D GRE VIBE images with respect to pixel graininess, streaking artifact, and sharpness (p = 0.0009 to p < 0.0001). Except for sharpness of
vessels on unenhanced images, results for the radial 3D GRE sequence did not differ significantly in the comparison of cooperative and noncooperative patients (p = 0.004). For imaging
of noncooperative patients, radial 3D GRE images of children had higher ratings for shading
(unenhanced, p = 0.0004; contrast-enhanced, p < 0.0001) and streaking artifacts on contrastenhanced images (p = 0.0017) than did those of adults. Overall image quality was higher for
pediatric patients. In lesion analysis, use of the 3D GRE VIBE sequence was associated with
significantly greater detectability, confidence, and conspicuity than was use of the radial 3D
GRE sequence (p = 0.00026–0.011).
CONCLUSION. A free-breathing radial 3D GRE sequence is feasible for abdominal
MRI and may find application in imaging of patients who are unable to suspend respiration,
especially children.
T
hree-dimensional gradient-recalled
echo (GRE) imaging with uniform fat suppression is currently
used for multiphase gadoliniumenhanced imaging of the upper abdomen and
is a critical component of abdominal MRI
examinations [1–3]. Artifacts caused by res­
piratory motion are one of the major causes of
image degradation, potentially obscuring important anatomic structures and lesions [4].
Breath-hold T1-weighted sequences are shortduration sequences that allow motion-free acquisition of diagnostic quality images of most
patients. Lee et al. [2] and Krinsky et al. [5]
reported that as many as 7% of patients undergoing hepatic MRI were unable to hold
their breath for 15 seconds. Various techniques have been used to reduce respiratory
artifacts when breath-hold techniques are
not applicable [4, 6–8]. Respiratory triggering and navigation are routinely used for T2weighted imaging and coronary imaging [7–
9]. Less attention has been paid to their use
in mitigating respiratory motion artifacts in
T1-weighted imaging [6, 7, 10]. Furthermore, these techniques require additional
steps in the clinical workflow and can fail in
cases of irregular respiration [4, 11, 12].
The free-breathing T1-weighted 3D GRE
with radial data sampling sequence in thoracic, cardiac, and coronary imaging has
been evaluated in several studies [13–15],
but only one report, to our knowledge, describes its application in abdominal MRI
[16]. In that study, respiratory motion–compensated radial dynamic contrast-enhanced
MRI of the chest and abdomen was performed
for only five patients. The investigators also
AJR:197, September 2011
Downloaded from www.ajronline.org by Indiana Univ-Acq Dept on 08/18/14 from IP address 140.182.64.84. Copyright ARRS. For personal use only; all rights reserved
Abdominal MRI
did not analyze lesion detectability, conspicuity, extent of artifacts, or overall image quality of the sequence. The purposes of our
study were to evaluate the feasibility of freebreathing 3D GRE with radial data sampling
acquisition in abdominal MRI compared
with a standard 3D GRE volumetric interpolated breath-hold examination (VIBE) sequence in imaging of cooperative patients
and to perform a preliminary assessment in
imaging of noncooperative patients.
Materials and Methods
Patients
Institutional review board approval was obtained
for this HIPAA-compliant retrospective study. The
MRI database was retrospectively searched for the
records of patients who between February 23 and
May 23, 2010, had undergone abdominal MRI examinations that included unenhanced and contrastenhanced free-breathing 3D T1-weighted gradientrecalled echo with radial data sampling (radial 3D
GRE) sequences at 1.5 T. The study group consisted of 55 consecutively registered patients (27 male
patients, 28 female patients; mean age, 44.2 ± 24.7
[SD] years; range, 9 months–83 years) who underwent free-breathing radial 3D GRE sequences. Of
these 55 subjects, 39 (18 male patients, 21 female
patients; mean age, 47.9 ± 22.6 [SD] years) underwent a cooperative protocol, and 16 subjects underwent a noncooperative protocol. Eight of the 16 patients were children (six boys, two girls; mean age,
2.6 ± 1.1 years), and eight were adults (three men,
five women; mean age, 56.5 ± 8.2 years). MRI was
used at the discretion of the referring physicians for
all examinations.
The primary indications for imaging included evaluation of liver lesions (n = 3); screening
for possible hepatocellular carcinoma or cirrhosis surveillance (n = 10); evaluation of nonhepatic malignancy, including postsurgical surveillance
(n = 12); evaluation of renal lesions (n = 4); post­
intervention assessment of the liver, including hepatocellular carcinoma evaluation (n = 1); evaluation of abdominal pain (n = 2); surveillance after
liver or kidney transplant (n = 6); postintervention
assessment of nonhepatic malignancy (n = 4); and
evaluation of bladder cancer (n = 6), suspected
biliary disease (n = 1), chronic pancreatitis (n =
1), carcinoid (n = 2), dysfunctional uterine bleeding (n = 1), and neutropenic fever (n = 2).
Twelve of the 39 cooperative patients had a total of 27 focal liver lesions (diameter range, 0.5–
5.1 cm; mean, 1.84 ± 1.21 cm). These lesions were
cysts (n = 6), hemangioma (n = 2), metastatic disease (n = 12), hepatocellular carcinoma (n = 6),
and an undetermined solid liver lesion (n = 1).
Only two of the 16 noncooperative patients had
focal liver lesions, both having metastasis. Characterization of observed liver lesions was not part
of the study design but was performed according
to standard MRI criteria [17].
MRI Technique
MRI of the abdomen was performed with a
1.5-T MRI system (Avanto, Siemens Healthcare)
with a phased-array torso coil according to one
of two protocols: cooperative and noncooperative. The cooperative protocol included transverse
unenhanced and contrast-enhanced T1-weighted
breath-hold sequences with 3D GRE VIBE technique and fat suppression (TR/TE, 4.3/1.7; flip
angle, 10°; slice thickness, 3.5 mm; gap, 0; mean
number of slices, 72 ± 15.2; range, 35–80 slices;
FOV, 360 mm; matrix size, 144 × 320, phase ×
frequency; acquisition time, 19 seconds). All 3D
GRE VIBE studies also included generalized autocalibrating partially parallel acquisitions parallel imaging (acceleration factor, 2).
The noncooperative protocol did not include a
T1-weighted 3D GRE VIBE sequence. Motion-resistant magnetization-prepared rapid gradient echo
was used as an unenhanced (without water excitation) and a contrast-enhanced (with water excitation) T1-weighted sequence [6]. Free-breathing
radial 3D GRE was added to both protocols. The
sequence acquisition was performed in a freebreathing manner without trigger and navigation
techniques. Parallel imaging was not used for the
radial 3D GRE sequence tested. The sequence parameters were TR/TE, 3.83/1.6; flip angle, 10°;
slice thickness, 3.0 mm; gap, 0 mm; mean number of slices, 80 ± 11.56; range, 56–96 slices; FOV,
380 mm; matrix size, 380 × 380; acquisition time,
84 seconds. The radial GRE sequence is a spoiled
steady-state gradient-echo sequence in which the
readout direction is altered each repetition and every acquired line passes through the center of the kspace such that the sampling pattern in the k-space
is not a rectangular grid but is a set of radial spokes.
The radial data are regridded onto a rectangular kspace grid before standard reconstruction [18].
For cooperative subjects contrast-enhanced
standard 3D GRE VIBE sequences were always
performed before the radial 3D GRE sequence.
Subjects were instructed to breathe normally during radial 3D GRE acquisition.
Image Analyses
The sequences under investigation were independently and retrospectively evaluated on a PACS
terminal by two blinded reviewers (each with 5
years of experience in abdominal imaging) to determine overall image quality, extent of artifacts,
and lesion detectability and conspicuity. For each
dataset, both readers scored 10 parameters of im-
age quality. The readings were separated into two
sessions. The unenhanced images were evaluated in
the first session and the contrast-enhanced images in
the second. Because of the late contrast-enhanced
performance of the radial 3D GRE sequence, comparison was made with the 3-minute contrastenhanced interstitial phase dataset of the 3D GRE
VIBE acquisition that were closest to the time of injection. The coordinator of the study randomly uploaded the images on the terminal for individual
evaluation by the reviewers, who were blinded to
the parameters of the sequences and to clinical information about the subject. Before image evaluation,
both radiologists collectively reviewed a training
dataset on five patients and agreed on the interpretations and scores for each parameter evaluated.
These data were not included in the study.
The reviewers graded images for the presence
of artifacts (respiratory ghosting, streaking artifacts, pixel graininess, integrated parallel acquisition technique [iPAT] artifact, and shading) and
sharpness (blurring of liver, pancreas, and vessels) using a 5-point scale (1, severe; 2, moderate;
3, mild; 4, minimal, 5, absent).
Streaking artifact occurs in radial acquisition
when the number of spokes sampled is too small
in relation to the diameter of the imaged object
[19]. Aliasing occurs when any part of the body
extends outside the FOV and a signal produced by
this structure reaches the receiver coil and is folded over into the image. Wraparound artifact at or
near the center of the image as a result of parallel image acquisition is described as iPAT artifact
[20, 21]. Shading was considered present when
lack of uniformity of signal intensity was present
within a slice, occurring most often in the phaseencoding direction.
The reviewers rated the homogeneity of fat attenuation using a 4-point scale, higher scores representing better fat suppression (1, no fat attenuation;
2, central-only inhomogeneous fat attenuation; 3,
central-only homogeneous fat attenuation; 4, central-peripheral homogeneous fat-attenuation). The
reviewers assessed overall image quality using the
following considerations: image sharpness, homogeneity of signal intensity, liver signal intensity,
kidney corticomedullary differentiation (on unenhanced images), and severity of artifacts. They
graded these factors on a 5-point scale (1, poor; 2,
fair; 3, good; 4, very good; 5, excellent).
The reviewers also evaluated focal liver lesions
on the images obtained with the sequences studied. Only lesions larger than 5 mm were evaluated. For each lesion identified, subjective conspicuity was assigned as follows: 1, poor; 2, fair; 3,
good, 4, excellent. The reviewers also scored their
confidence in detection of each lesion as follows:
1, suspicious, 2; possible, 3; probable, 4; definite.
AJR:197, September 2011651
Downloaded from www.ajronline.org by Indiana Univ-Acq Dept on 08/18/14 from IP address 140.182.64.84. Copyright ARRS. For personal use only; all rights reserved
Azevedo et al.
If the reader did not detect a lesion on a certain
image, grade 0 was designated during the statistical
analysis of conspicuity and confidence level. The total number of lesions was determined in consensus
by the coordinator and a fourth radiologist (7 years
of experience in abdominal imaging) who evaluated all images from each examination, including contrast-enhanced and T2-weighted images. During statistical analysis only lesions with confidence levels
graded 3 or 4 were defined as confidently detected.
Statistical Analyses
Interobserver reproducibility for the qualitative data was assessed with kappa statistics. A
kappa value less than 0.2 indicated slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, good agreement; and
greater than 0.0.81, excellent agreement [22]. For
cooperative subjects, the statistical significance
of data from the comparative analysis of 3D GRE
VIBE and radial 3D GRE, including liver lesion
analysis (conspicuity and confidence level), was
assessed with the Wilcoxon signed rank test. The
nonparametric McNemar test was used to evaluate the statistical significance of lesion detection.
Bonferroni correction for multiple comparisons
(10 comparisons) was applied, and p < 0.005 was
regarded as statistically significant. The comparative analysis of radial 3D GRE between cooperative and noncooperative subjects was performed
with the Mann-Whitney U test. Bonferroni correction for multiple comparisons (nine comparisons) was applied, and p < 0.0056 was regarded as
statistically significant. Statistical analyses were
performed with MedCalc for Microsoft Windows
(version 11.3.0.0, MedCalc Software).
Results
Kappa values for agreement between the
two reviewers for the independent qualitative data analyses ranged from 0.68 to 0.92.
The kappa value was 1.0 for streaking artifacts on 3D GRE VIBE images and for iPAT
artifacts on radial 3D GRE images. In no case
was there a significant difference in rating between readers. Averaged data are shown to
simplify and improve clarity of display. The
average scores of all the parameters qualitatively evaluated, including artifacts, fat attenuation, and overall image quality, in 3D GRE
VIBE and radial 3D GRE acquisitions in cooperative subjects are displayed in Table 1.
Concerning sharpness, radial 3D GRE had
significantly lower mean ratings than 3D GRE
VIBE for unenhanced pancreas (p = 0.0009)
and in all contrast-enhanced comparisons (liver,
p = 0.0010; pancreas, p < 0.0001; vessels, p <
0.0001). However, radial 3D GRE sharpness
was always rated higher than 3 (mild to minimal blurring). Streaking artifacts were absent
from the 3D GRE VIBE images of all subjects,
and radial 3D GRE had mean ratings of 3.32
and 3.24 on unenhanced and contrast-enhanced
images (mild to minimal) (p < 0.0001). On unenhanced images, overall image quality was adequate with radial 3D GRE. The mean rating
was 3.45 (good to very good) with no significant difference compared with 3D GRE VIBE,
which had a mean rating of 3.56 (p = 0.3680).
On contrast-enhanced images, overall image
quality was significantly inferior with radial 3D
GRE than with 3D GRE VIBE (p < 0.0001)
with a mean rating of 2.94 (nearly good).
In four cooperative subjects, the 3D GRE
VIBE images had motion artifacts rated 3 or
2 (mild or moderate), and the free-breathing radial 3D GRE images were rated 4 or 5
(minimal or absent) with higher overall image quality (Fig. 1). No iPAT artifacts were
found on radial 3D GRE images of any subject because parallel imaging was not used
in this sequence, but 3D GRE VIBE images
had mean ratings of 4.00 and 4.06 (minimal)
on the unenhanced and contrast-enhanced
images (p < 0.0001 in both comparisons).
Regarding fat attenuation, radial 3D GRE
images had significantly higher (p < 0.0001)
scores than did 3D GRE VIBE images with
averages of 3.60 for unenhanced and 3.68
for contrast-enhanced images (central homogeneous, periphery slightly heterogeneous).
There were no significant differences in the
scores of shading and motion artifacts between
3D GRE VIBE and radial 3D GRE images.
Table 2 shows the results of qualitative
analysis, including artifacts, fat attenuation,
and overall image quality, of the use of the
radial 3D GRE sequence for imaging of cooperative and noncooperative patients. There
were no significant differences between radial 3D GRE images of cooperative and those
of noncooperative subjects (both adult and
pediatric patients) with either unenhanced or
contrast-enhanced technique. The only exception was the analysis of sharpness of vessels on unenhanced images, in which images
of noncooperative subjects received a higher mean rating (3.9) than those of cooperative subjects (3.4) (p = 0.004). These ratings,
TABLE 1:Results of Qualitative Analysis of Two Gradient-Echo Sequences for Imaging of Cooperative Patients
Unenhanced
Quality Parameter
Contrast-Enhanced
3D GRE VIBE
Radial 3D GREa
p
4.74±0.54
4.91 ± 0.33
0.0460
Respiratory motion
Shading
4.19±0.06
3.92 ± 0.92
0.0124
Pixel graininess
4.33±0.47
3.41 ± 0.59
< 0.0001b
Overall image quality
3.56±0.66
3.45 ± 0.96
3D GRE VIBE
Radial 3D GREa
p
4.76± 0.56
0.6570
4.78± 0.59
4.28± 0.51
4.18± 0.62
0.2920
4.37± 0.49
3.30± 0.56
< 0.0001b
0.3680
3.78± 0.96
2.94± 0.88
< 0.0001b
4.06± 0.61
5.00± 0.00
< 0.0001b
Integrated parallel acquisition technique
4.00±0.06
5.00 ± 0.00
< 0.0001b
Streaking artifacts
5.00±0.00
3.32 ± 0.65
< 0.0001b
5.00± 0.00
3.24± 0.74
< 0.0001b
3.08± 0.48
3.68± 0.47
< 0.0001b
3.06±0.34
3.60 ± 0.54
< 0.0001b
Liver
3.88±0.51
3.78 ± 0.83
0.4000
3.87± 0.63
3.41± 0.84
0.0010b
Pancreas
3.90±0.55
3.53 ± 0.84
0.0009b
3.67± 0.66
3.10± 0.88
< 0.0001b
Vessels
3.67±0.53
3.40 ± 0.84
0.0157
4.05± 0.75
3.40± 0.87
< 0.0001b
Fat attenuation
Sharpness
Note—Values are mean ± SD. Wilcoxon signed rank test with Bonferroni correction was used for statistical analysis. GRE = gradient-recalled echo, VIBE = volumetric
interpolated breath-hold examination.
aFree-breathing 3D T1-weighted gradient-echo sequence using radial data sampling.
bSignificant at p < 0.005.
652
AJR:197, September 2011
Downloaded from www.ajronline.org by Indiana Univ-Acq Dept on 08/18/14 from IP address 140.182.64.84. Copyright ARRS. For personal use only; all rights reserved
Abdominal MRI
A
B
C
D
Fig. 1—29-year-old woman with history of soft-tissue sarcoma.
A–D, Transverse T1-weighted fat-suppressed MR images acquired in unenhanced (A and B) and interstitial contrast-enhanced (C and D) phases. Standard breath-hold
3D gradient-recalled echo (GRE) volumetric interpolated breath-hold examination (VIBE) sequence was used for A and C and free-breathing 3D GRE with radial data
sampling sequence for B and D. Despite cooperative protocol, unenhanced 3D GRE VIBE image shows mild motion artifacts (white arrows, A) and integrated parallel
acquisition technique artifact (black arrows, A), and contrast-enhanced image shows minimal motion artifacts (arrows, C). No motion artifacts are present on freebreathing radial 3D GRE images (B and D), but minimal streaking artifacts (arrows, B and D) and pixel graininess are evident. Cyst is evident in left kidney.
however, represent good image sharpness for
both sequences (mild to minimal blurring
of vessels). Table 2 also shows the results of
qualitative analysis of radial 3D GRE images
of cooperative patients compared only with
those of noncooperative adult patients (i.e.,
pediatric subjects excluded). On both unenhanced and contrast-enhanced images, there
were no statistically significant differences
in any comparison.
In the comparison of the results of qualitative analysis of radial 3D GRE images of
adult and pediatric noncooperative patients
(Table 3), the images of noncooperative patients received significantly higher ratings
for shading on both unenhanced (p = 0.0004)
and contrast-enhanced images (p < 0.0001)
and for streaking artifacts on contrast-enhanced images (p = 0.0017). There was no
significant difference between pediatric and
adult noncooperative patients with respect
to respiratory motion, pixel graininess, and
overall image quality (Fig. 2).
Table 4 shows the detectability, confidence, and conspicuity results for focal liver lesions on unenhanced images. No significant difference in lesion analysis was
observed between readers, so averaged data
are displayed. The 3D GRE VIBE sequence
was associated with significantly higher detectability, confidence, and conspicuity than
the radial 3D GRE sequence in all comparisons (p = 0.0003–0.011). Approximately 55%
(15/27) of all liver lesions measured 1.5 cm
or less in diameter. In all cases (five lesions)
in which reviewers were confident about detection of the lesion on 3D GRE VIBE images but not radial 3D GRE images, the lesions
were smaller than 1.0 cm. Detectability, confidence, and conspicuity were not compared
on contrast-enhanced images because the radial 3D GRE sequence was performed in the
interstitial phase, which is poor for detection
of liver lesions [17].
Discussion
The results of our study show that a 3D
GRE sequence with radial acquisition in a
free-breathing manner is feasible for abdominal MRI studies. Interestingly, we observed
that there were no significant differences between breath-hold 3D GRE and free-breathing radial 3D GRE with respect to respiratory
motion in cooperative subjects. Radial 3D
GRE images consistently showed minimal
motion artifacts (mean ratings of 4.91 on unenhanced and 4.76 on contrast-enhanced images) despite acquisition during respiration.
Corroborating this finding, when radial 3D
GRE images of noncooperative subjects were
evaluated, the motion artifacts remained minimal (mean ratings of 4.75 on unenhanced and
4.63 on contrast-enhanced images), comparable
AJR:197, September 2011653
0.273
0.022
3.62 ± 0.7
3.78 ± 0.8
3.40 ± 0.9
0.349
3.62 ± 0.9
3.93 ± 0.9
3.40 ± 0.5
Vessels
Note—Values are mean ± SD. Mann-Whitney test with Bonferroni correction was used for statistical analysis.
a Adult and pediatric patients.
bResults were considered statistically significant at p < 0.0056.
0.092
0.004b
0.085
0.102
0.069
3.43 ± 0.5
3.75 ± 0.4
3.66 ± 0.6
3.38 ± 0.7
3.10 ± 0.9
3.41 ± 0.8
0.451
0.229
0.899
0.625
3.62 ± 0.5
3.25 ± 0.6
Pancreas
3.84 ± 0.8
3.78 ± 0.8
3.53 ± 0.8
Liver
3.56 ± 0.8
0.123
0.005
3.87 ± 0.3
3.94 ± 0.2
3.68 ± 0.5
0.339
0.205
3.75 ± 0.4
3.60 ± 0.5
Fat attenuation
3.75 ± 0.4
3.32 ± 0.6
Sharpness
0.512
0.216
0.216
3.00 ± 0.6
3.50 ± 0.9
3.24 ± 0.7
0.159
0.386
3.06 ± 0.7
0.607
Streaking artifacts
3.53 ± 0.9
0.013
0.441
3.25 ± 0.4
3.06 ± 0.7
3.41 ± 0.8
3.47 ± 0.7
3.30 ± 0.6
2.94 ± 0.9
0.347
0.205
0.786
0.492
3.19 ± 0.6
3.19 ± 0.6
3.41 ± 0.7
Overall image quality
3.56 ± 0.9
3.41 ± 0.6
3.45 ± 0.9
Pixel graininess
0.041
0.006
0.239
0.102
4.50 ± 0.6
3.69 ± 0.6
4.31 ± 0.8
4.63 ± 0.6
4.76 ± 0.6
4.18 ± 0.6
0.560
0.056
0.045
0.051
3.88 ± 0.6
4.75 ± 0.4
4.75 ± 0.5
Shading
4.31 ± 0.7
4.91 ± 0.3
3.92 ± 0.9
Respiratory motion
All Cooperative
Subjects (n = 39)
Quality Parameter
All
Noncooperative
Subjects (n = 16)a
Noncooperative
Adults (n = 8)
All
Noncooperative vs
Cooperative vs
Noncooperative
All Cooperative
Noncooperative
Adults
Subjects (n = 39)
Adult
Noncooperative
Subjects (n = 16)
Noncooperative
Adults (n = 8)
All
Noncooperative vs
Cooperative vs
Noncooperative
Noncooperative
Adults
p
Contrast-Enhanced
p
Unenhanced
Downloaded from www.ajronline.org by Indiana Univ-Acq Dept on 08/18/14 from IP address 140.182.64.84. Copyright ARRS. For personal use only; all rights reserved
TABLE 2:Results of Qualitative Analysis of Radial Free-Breathing 3D T1-Weighted Gradient-Echo Sequence Using Radial Data Sampling: Cooperative and
Noncooperative Patients
654
Azevedo et al.
to findings in cooperative subjects. In addition, radial 3D GRE images
had mean ratings higher than 3 (mild) in the analysis of all artifacts in
both cooperative and noncooperative patients. The overall quality of
unenhanced images obtained with free-breathing radial 3D GRE was
higher than 3 (good) with no significant difference compared with
breath-hold 3D GRE VIBE. The latter sequence is regarded as the
most effective approach to obtaining motion-free high-quality images
of patients able to cooperate with breath-holding [3]; hence, this finding was somewhat unexpected.
The overall quality of free-breathing radial 3D GRE contrast-enhanced images was significantly lower than that of breath-hold 3D
GRE images (p < 0.0001), but the mean rating was 2.94 (nearly good).
In four cooperative subjects, the 3D GRE images had motion artifacts
rated 3 or 2 (mild or moderate), and the free-breathing radial 3D GRE
images had artifacts rated 4 or 5 (minimal or absent) with improvement
of overall image quality. Although no statistical conclusions can be
drawn regarding this difference, owing to the small number of subjects
who were cooperative but had difficulty holding their breath for the entire acquisition, this result may suggest that in the care of patients who
cannot cooperate with breath-hold instructions, radial 3D GRE may be
an alternative. As further evidence, among noncooperative subjects,
radial 3D GRE images had mean overall quality ratings of 3.56 for unenhanced and 3.41 for contrast-enhanced images (good to very good).
These findings may have important clinical implications because important data can be missed in the use of breath-hold sequences for imaging of patients who cannot adequately hold their breath [4, 6].
To our knowledge, this study is the first to comprehensively evaluate a free-breathing T1-weighted 3D GRE sequence with radial data
sampling in abdominal MRI. Our findings show the technique may be
a feasible alternative for imaging of noncooperative patients. The only
other report [16], to our knowledge, in which the applicability of T1weighted 3D GRE with radial data sampling in abdominal MRI was
evaluated included only five cooperative subjects and was performed
with a form of respiratory-motion compensation.
In the analysis of radial 3D GRE for imaging of noncooperative
subjects in which adult and pediatric subjects were compared, images of children received scores the same as or higher than the images
of adults in most of the parameters evaluated. Statistical significance
was found in the analysis of shading artifact on both unenhanced (p =
0.0004) and contrast-enhanced (p < 0.0001) images and in the analysis of streaking artifacts on contrast-enhanced images (p = 0.0017).
The images of noncooperative adults did not receive significantly
better scores in any of the evaluated parameters. The overall quality of radial 3D GRE images of children (unenhanced images, 3.94;
contrast-enhanced images, 3.75) was not significantly better than the
quality of images of adults (unenhanced images, 3.19, p = 0.0091; contrast-enhanced images, 3.06, p = 0.0237). However, the significantly
less severe shading and streaking artifacts in pediatric patients compared with adults could conceivably lead to better overall image quality. Because sedation and general anesthesia are frequently necessary,
especially for pediatric patients, standard breath-hold T1-weighted sequences are unsatisfactory, and free-breathing radial 3D GRE may be
an adequate alternative. Vanderby et al. [23] reported that use of sedation techniques increases total study time and the costs of pediatric
MRI. Use of radial 3D GRE may help ameliorate these circumstances
because it does not require substantial cooperation.
Causes of image degradation on radial 3D GRE images were the
presence of streaking artifacts, pixel graininess, and differences in
AJR:197, September 2011
Abdominal MRI
Downloaded from www.ajronline.org by Indiana Univ-Acq Dept on 08/18/14 from IP address 140.182.64.84. Copyright ARRS. For personal use only; all rights reserved
TABLE 3:Results of Qualitative Analysis of Radial Free-Breathing 3D
T1-Weighted Gradient-Echo Sequence Using Radial Data Sampling
for Imaging of Noncooperative Patients: Comparison of Adults
and Children
Unenhanced
Contrast-Enhanced
Adults
(n = 8)
Children
(n = 8)
p
Adults
(n = 8)
Children
(n = 8)
p
Respiratory motion
4.75 ± 0.45
4.75 ± 0.58
0.7655
4.50 ± 0.63
4.75 ± 0.45
0.2528
Shading
3.88 ± 0.62
4.75 ± 0.45
0.0004a
3.69 ± 0.60
4.94 ± 0.25
< 0.0001a
Pixel graininess
3.19 ± 0.65
3.62 ± 0.72
0.1293
3.25 ± 0.45
3.69 ± 0.79
0.1039
Overall image quality
3.19 ± 0.65
3.94 ± 0.93
0.0091
3.06 ± 0.68
3.75 ± 0.86
0.0237
Streaking artifacts
3.06 ± 0.68
4.00 ± 0.89
0.0062
3.00 ± 0.63
4.00 ± 0.82
0.0017a
Fat attenuation
3.75 ± 0.45
3.75 ± 0.45
1
3.87 ± 0.34
4.00 ± 0.00
0.2332
Liver
3.62 ± 0.50
4.06 ± 0.93
0.0334
3.75 ± 0.44
3.56 ± 0.51
0.2884
Pancreas
3.25 ± 0.58
3.88 ± 0.96
0.0461
3.43 ± 0.51
3.31 ± 0.79
0.8221
Vessels
3.62 ± 0.88
4.25 ± 1.00
0.0425
3.62 ± 0.72
3.94 ± 0.77
0.1957
Quality Parameter
Sharpness
Note—Values are mean ± SD. Mann-Whitney test with Bonferroni correction was used for statistical
analysis.
aResults were considered statistically significant at p < 0.0056.
image sharpness of the evaluated tissues.
Streaking artifacts were found only on radial 3D GRE images, both unenhanced and
contrast-enhanced. These artifacts may be
more pronounced when the arms of patients
are positioned along their sides, which is the
standard positioning, and occurred in all subjects in our study. These artifacts were more
marked in adults, probably because of the
larger volume of the abdomen and the presence of the arms inside the FOV. Pediatric
patients have smaller abdominal volume, and
the arms were usually outside the FOV. Positioning the subject’s arms above the head
can reduce these artifacts, but this solution
may not be good because of patient discomfort. The other alternative to reducing these
artifacts is to increase the number of samples (spokes), which would increase acquisition time, also undesirable. Future research
incorporating parallel imaging technique
in radial acquisition should reduce imaging
time and may also reduce streaking artifacts.
Radial 3D GRE images had significantly lower ratings of pixel graininess, which is
explained by the higher matrix and the thinner slice thickness used in this sequence
compared with 3D GRE VIBE [14, 19]. The
ration­ale for using a higher matrix in the radial 3D GRE sequence was to maximize the
definition of abdominal structures and possibly of lesion conspicuity. Because radial 3D
GRE is a free-breathing sequence, the higher
matrix was easily added despite the prolonged
acquisition time. Comparable matrix resolution in the breath-hold 3D GRE sequence
would have increased the sequence duration
beyond reasonable breath-hold capability.
Both unenhanced and contrast-enhanced
radial 3D GRE images had lower scores with
respect to sharpness. A significant difference
in sharpness of the pancreas (p = 0.0009) was
found on unenhanced images and in sharpness of the liver (p = 0.0010), pancreas (p <
0.0001), and vessels p < 0.0001) on contrastenhanced images. The signal-to-noise ratio on
the contrast-enhanced radial 3D GRE images,
compared with that on the 3D GRE images,
might have been reduced somewhat owing to
clearance of the contrast material during the
interval between the two data acquisitions.
This factor may be responsible for differences in tissue contrast characteristics; for example, tissue and vessel conspicuity may be
less evident owing to further passage of contrast material out of the intravascular space
between the interstitial phase 3D GRE and
the more delayed radial 3D GRE acquisition.
The time between the 3D GRE and the end
of the radial 3D GRE acquisition was 4–6
minutes. The order of acquisition, however,
was not expected to affect comparison of the
unenhanced MR images.
The reviewers saw iPAT artifacts only on
3D GRE images with average scores of minimal to absent. The absence of iPAT artifacts
A
B
Fig. 2—2-year-old boy with bilateral Wilms tumor.
A and B, Transverse T1-weighted fat-suppressed MR images obtained with free-breathing 3D gradient-recalled echo with radial data sampling technique. Both
unenhanced (A) and interstitial phase contrast-enhanced (B) images obtained in free-breathing acquisition are of high quality. There are no motion artifacts; only mild
pixel graininess is evident.
AJR:197, September 2011655
Azevedo et al.
TABLE 4:Lesion Detectability and Conspicuity and Confidence in
Interpretation on Unenhanced Images
3D GRE VIBE
Radial 3D GREa
p
Detectability
42/54 (77.8)
32/54 (59.3)
0.0020b
Confidence
2.81 ± 1.37
2.5 ± 1.50
0.011b
Conspicuity
2.78 ± 1.39
2.43 ± 1.19
0.0003b
Downloaded from www.ajronline.org by Indiana Univ-Acq Dept on 08/18/14 from IP address 140.182.64.84. Copyright ARRS. For personal use only; all rights reserved
Characteristic
Note—Detectability is the number with percentage in parentheses. Value is based on lesions graded 4 or 5 by
subjective confidence. The total number of lesions depicted in 3D GRE VIBE and radial 3D GRE images
represents the sum of two reviewers. The reference number of lesions (n = 54) was determined by evaluation
of all images in each examination by two radiologists other than the reviewers. Confidence and conspicuity are
mean ± SD. GRE = gradient-recalled echo, VIBE = volumetric interpolated breath-hold examination.
aFree-breathing 3D T1-weighted gradient-echo sequence using radial data sampling.
bResults were considered statistically significant at p < 0.05.
on radial 3D GRE images is explained by the
fact that these artifacts are related to parallel
imaging, which was not used in the radial 3D
GRE sequence tested.
Radial 3D GRE images had better scores
for homogeneity of fat attenuation on both
unenhanced and contrast-enhanced images.
Although greater fat suppression was considered an advantage of use of radial 3D GRE
in general, it did not have a major effect on
differences in overall image quality between
sequences. It is not clear why fat suppression
was superior on the radial 3D GRE images
compared with the conventional 3D GRE
images. The explanation may be related to
chemical shift. In a standard rectilinear acquisition, fat is shifted only in the readout direction. In a radial k-space, each line experiences fat shift in a different direction. The
result is blurring of the fat, which can make
suppression more effective.
The overall image quality with radial 3D
GRE was higher on unenhanced images than
that on contrast-enhanced images. It is not
clear why image quality was inferior on contrast-enhanced images. Possible explanations
include higher signal intensity from the contrast-amplified artifacts of breathing motion
and image heterogeneity due to radial acquisition. Undersampling of higher-order phaseencoding steps might have emphasized lesser edge definitions on the radial acquisition
sequences. Differences in image sharpness
and pixel graininess on radial 3D GRE images may account for the lower detectability, confidence in detection, and conspicuity
of liver lesions on unenhanced images compared with the quality on conventional 3D
GRE images. However, we observed that all
lesions not detected on radial 3D GRE images were smaller than 1.0 cm.
This study had limitations. The first was
the retrospective design. Because radial 3D
GRE is not established in current practice,
656
we had a relatively small number of noncooperative subjects, which might have hampered
the statistical analysis. In addition, our study
group was heterogeneous, including patients
with a range of disease entities. However, results in a heterogeneous population may provide an overview of the performance of the
sequence. Moreover, use of a heterogenous
sample may imply generalizability of the results and may strengthen the potential clinical applicability. The results also provide a
framework for identifying particular patient
groups promising for future research, such
as pediatric patients. An additional limitation was that the noncooperative protocols
did not include 3D GRE in both unenhanced
and contrast-enhanced acquisition because it
requires the patient to perform a breath-hold.
Therefore, we could not compare the two sequences in this patient group. Another limitation was that the contrast-enhanced radial
3D GRE acquisition had a 5- to 6-minute delay. As a result the only reasonable comparison was with 3-minute contrast-enhanced interstitial phase 3D GRE VIBE images. We
therefore did not evaluate lesion detectability
and conspicuity after contrast enhancement.
As a result of our study, we now routinely acquire radial 3D GRE images of pediatric patients. We have stopped performing the
sequence for adult patients who can suspend
respiration but have added it to our noncooperative protocol. An important next step in
the evolution of radial 3D GRE is to modify
the sequence with the incorporation of parallel imaging to accelerate image acquisition so that acquisition can be performed in a
short enough time to allow dynamic gadolinium-enhanced imaging, especially with acquisition in the hepatic arterial phase.
Conclusion
Our preliminary results show that radial 3D GRE acquisition in a free-breathing
manner is feasible for abdominal MRI. Our
results also suggest that this sequence may
be applicable to imaging of patients who are
unable to suspend respiration.
Acknowledgments
We thank Tobias Block and Christian Geppert, Siemens Healthcare, the authors of the
radial 3D GRE sequence.
References
1.Sharma P, Martin DR, Dale BM, Semelka RC. Diagnostic approach to protocoling and interpreting
MR studies of the abdomen and pelvis. In: Semelka RC, ed. Abdominal-pelvic MRI. Hoboken,
NJ: Wiley-Blackwell, 2010:1–43
2.Lee VS, Lavelle MT, Rofsky NM, et al. Hepatic
MR imaging with a dynamic contrast-enhanced
isotropic volumetric interpolated breath-hold examination: feasibility, reproducibility, and technical quality. Radiology 2000; 215:365–372
3.Willatt JM, Hussain HK, Adusumilli S, Marrero
JA. MR imaging of hepatocellular carcinoma in
the cirrhotic liver: challenges and controversies.
Radiology 2008; 247:311–330
4.Noterdaeme O, Gleeson F, Phillips RR, Brady M.
Quantification of missing and overlapping data in
multiple breath hold abdominal imaging. Eur J
Radiol 2007; 64:273–278
5.Krinsky GA, Lee VS, Theise ND, et al. Hepatocellular carcinoma and dysplastic nodules in patients with cirrhosis: prospective diagnosis with
MR imaging and explantation correlation. Radiology 2001; 219:445–454
6.Altun E, Semelka RC, Dale BM, Elias J Jr. Water
excitation MPRAGE: an alternative sequence for
postcontrast imaging of the abdomen in noncooperative patients at 1.5 Tesla and 3.0 Tesla MRI. J
Magn Reson Imaging 2008; 27:1146–1154
7.Vasanawala SS, Iwadate Y, Church DG, Herfkens
RJ, Brau AC. Navigated abdominal T1-W MRI
permits free-breathing image acquisition with
less motion artifact. Pediatr Radiol 2010; 40:
340–344
8.Augui J, Vignaux O, Argaud C, Coste J, Gouya H,
Legmann P. Liver: T2-weighted MR imaging with
breath-hold fast-recovery optimized fast spin-echo
compared with breath-hold half-Fourier and nonbreath-hold respiratory-triggered fast spin-echo
pulse sequences. Radiology 2002; 223:853–859
9.Hirokawa Y, Isoda H, Maetani YS, Arizono S, Shimada K, Togashi K. MRI artifact reduction and
quality improvement in the upper abdomen with
PROPELLER and prospective acquisition correction (PACE) technique. AJR 2008; 191:1154–1158
10.Young PM, Brau AC, Iwadate Y, et al. Respiratory navigated free breathing 3D spoiled gradient-recalled echo sequence for contrast-enhanced
AJR:197, September 2011
Downloaded from www.ajronline.org by Indiana Univ-Acq Dept on 08/18/14 from IP address 140.182.64.84. Copyright ARRS. For personal use only; all rights reserved
Abdominal MRI
examination of the liver: diagnostic utility and comparison with free breathing and breath-hold conventional examinations. AJR 2010; 195:687–691
11.Morita S, Ueno E, Suzuki K, et al. Navigatortriggered prospective acquisition correction
(PACE) technique vs. conventional respiratorytriggered technique for free-breathing 3D MRCP:
an initial prospective comparative study using
healthy volunteers. J Magn Reson Imaging 2008;
28:673–677
12.McConnell MV, Khasgiwala VC, Savord BJ, et al.
Comparison of respiratory suppression methods
and navigator locations for MR coronary angiography. AJR 1997; 168:1369–1375
13.Amano Y, Takahama K, Kumita S. Noncontrastenhanced three-dimensional magnetic resonance
aortography of the thorax at 3.0 T using respiratorycompensated T1-weighted k-space segmented gradient-echo imaging with radial data sampling: pre-
liminary study. Invest Radiol 2009; 44:548–552
14.Shankaranarayanan A, Simonetti OP, Laub G,
Lewin JS, Duerk JL. Segmented k-space and realtime cardiac cine MR imaging with radial trajectories. Radiology 2001; 221:827–836
15.Peters DC, Ennis DB, Rohatgi P, Syed MA,
McVeigh ER, Arai AE. 3D breath-held cardiac
function with projection reconstruction in steady
state free precession validated using 2D cine
MRI. J Magn Reson Imaging 2004; 20:411–416
16.Lin W, Guo J, Rosen MA, Song HK. Respiratory
motion-compensated radial dynamic contrast-enhanced (DCE)-MRI of chest and abdominal lesions. Magn Reson Med 2008; 60:1135–1146
17.Braga L, Armao D, Semelka RC. MRI of the liver.
In: Semelka RC, ed. Abdominal-pelvic MRI.
Hoboken, NJ: Wiley-Blackwell, 2010:45–200
18.Dale B, Wendt M, Duerk JL. A rapid look-up table
method for reconstructing MR images from arbi-
trary K-space trajectories. IEEE Trans Med Imaging 2001; 20:207–217
19.Block KT, Uecker M, Frahm J. Undersampled radial MRI with multiple coils: iterative image reconstruction using a total variation constraint.
Magn Reson Med 2007; 57:1086–1098
20.Zhuo J, Gullapalli RP. AAPM/RSNA physics tutorial for residents: MR artifacts, safety, and quality control. RadioGraphics 2006; 26:275–297
21.Glockner JF, Hu HH, Stanley DW, Angelos L,
King K. Parallel MR imaging: a user’s guide. RadioGraphics 2005; 25:1279–1297
22.Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics
1977; 33:159–174
23.Vanderby SA, Babyn PS, Carter MW, Jewell SM,
McKeever PD. Effect of anesthesia and sedation
on pediatric MR imaging patient flow. Radiology
2010; 256:229–237
F O R YO U R I N F O R M AT I O N
Unique customized medical search engine service from ARRS! ARRS GoldMiner ® is a keyword- and
concept-driven search engine that provides instant access to radiologic images published in peer-reviewed
journals. For more information, visit http://goldminer.arrs.org.
AJR:197, September 2011657