Download MRI of respiratory dynamics with 2D steady-state free

Eur Radiol (2009) 19: 391–399
DOI 10.1007/s00330-008-1148-x
MAGN ETIC RE SONA NCE
M. Fabel
B. J. Wintersperger
O. Dietrich
M. Eichinger
C. Fink
M. Puderbach
H.-U. Kauczor
S. O. Schoenberg
J. Biederer
MRI of respiratory dynamics with 2D
steady-state free-precession and 2D gradient
echo sequences at 1.5 and 3 Tesla:
an observer preference study
Received: 6 March 2008
Revised: 14 July 2008
Accepted: 21 July 2008
Published online: 6 September 2008
# European Society of Radiology 2008
C. Fink . S. O. Schoenberg
Department of Clinical Radiology,
University Hospital HeidelbergMannheim,
Mannheim, Germany
M. Fabel (*) . J. Biederer
Department of Diagnostic Radiology,
University Hospital SchleswigHolstein,
Campus Kiel, Arnold-Heller-Str. 9,
24105 Kiel, Germany
e-mail: [email protected]
Tel.: +49-431-5973154
Fax: +49-431-5973151
B. J. Wintersperger . O. Dietrich
Department of Clinical Radiology,
University Hospital Munich,
Munich, Germany
M. Eichinger . M. Puderbach
Department of Oncological
Diagnostics and Therapy,
German Cancer Research Center,
Heidelberg, Germany
H.-U. Kauczor
Department of Diagnostic Radiology,
University Hospital Heidelberg,
Heidelberg, Germany
Abstract To compare the image
quality of dynamic lung MRI with
variations of steady-state freeprecession (SSFP) and gradient echo
(GRE) cine techniques at 1.5 T and
3 T. Ventilated porcine lungs with
simulated lesions inside a chest phantom and four healthy human subjects
were assessed with SSFP (TR/TE=
2.9/1.22 ms; 3 ima/s) and GRE
sequences (TR/TE=2.34/0.96 ms;
8 ima/s) as baseline at 1.5 and 3 T.
Modified SSFPs were performed with
Introduction
In oncological patient care, magnetic resonance imaging
(MRI) has been shown to be particularly sensitive for the
detection of cerebral, abdominal and vertebral metastases
[1–3]. Thus, comprehensive studies for whole-body MRI
have been discussed and developed for screening and
staging of metastatic cancer. Whole-body MRI may only
be clinically established and fully accepted if the protocols
include lung imaging at a reasonable image quality,
especially since many malignancies come along with
nine to ten images/s (parallel imaging
factors 2 and 3). Image quality for
representative structures and artifacts
was ranked by three observers independently. At 1.5 T, standard SSFP
achieved the best image quality with
superior spatial resolution and signal,
but equal temporal resolution to GRE.
SSFP with improved temporal resolution was ranked second best. Further
acceleration (PI factor 3) was of no
benefit, but increased artifacts. At 3 T,
GRE outranged SSFP imaging with
high lesion signal intensity, while
artifacts on SSFP images increased
visibly. At 1.5 T, a modified SSFP
with moderate parallel imaging (PI
factor 2) was considered the best
compromise of temporal and spatial
resolution. At 3 T, GRE sequences
remain the best choice for dynamic
lung MRI.
Keywords Lung . Dynamic MRI .
Parallel imaging . Nodules . Infiltrates
lung metastases [4, 5]. Since MRI of the lung was difficult
for a long time, detection of either lung metastases or a
primary malignancy of the lung, e.g., bronchial carcinoma,
was considered a weak point of this strategy. Present
protocols have filled this gap, and MRI of the lung has
become a powerful tool for research and specific clinical
applications [6, 7].
Within a few years, lung MRI was increasingly
recognized as a radiation-free alternative to CT for specific
clinical problems, e.g., in cases of evaluation for pneumonia
[8–11], in particular for young patients as well as patients
392
with benign lung diseases [12, 13]. As an additional feature,
dynamic lung MRI with real-time SSFP or GRE sequences
is an easy and quick examination and was therefore readily
accepted in clinical routine wherever it became available.
Although computed tomography still provides the
superior spatial resolution and a reasonable soft tissue
contrast for lung imaging, specific clinical problems
warrant the use of dynamic MRI (dMRI) for the assessment
of respiratory mechanics. In particular for patients
presenting with lung cancer, dMRI is a fast and reliable
method to estimate tumor motion and chest wall or
mediastinal adhesion and invasion [6, 7]. This information
can be useful for surgery planning as well as high precision
radiotherapy to protect the surrounding healthy lung tissue
[14].
CT so far does not allow acquiring dynamic studies of
respiratory motion on a routine basis. Low-pitch helical CT
or equivalent 4D CT techniques acquire data from the
shallow-breathing patient at rest. Retrospective gating
corresponding to a selected respiration depth is used to
reconstruct 3D data from multiple respiratory cycles. This
CT technique does not allow examining respirationcorrelated motion of a tumor during variable conditions,
e.g., forced in- and expiration[15].
In this respect, MRI is superior to 4D CT in determining
tumor motion. With the implementation of parallel imaging
on multi-channel MR systems and development of new
reconstruction algorithms, MRI became faster and nowadays allows free-breathing, dynamic examination of the
lung.
With this background, recommended protocols for lung
MRI have been discussed [15, 16]. One approach includes
a short sequence of dynamic imaging at free breathing and
at defined breathing maneuvers.
Clinical usage of dynamic lung MRI at 1.5 T is well
reported for detection of lung pathologies[10, 17, 18].
Therefore, dynamic sequences have been included into
recommendations for standard MRI protocols of the lung
with 1.5-T systems [6]. Initial results for non-dynamic lung
MRI at 3 T are also described [19]. So far, the established
FLASH 2D sequences provide a high temporal resolution
(eight to nine acquisitions per second), but with an
unsatisfying signal and parenchymal contrast.
Steady-state free-precession (SSFP)-based protocols
yield higher signal and contrast, but image quality might
be degraded by artifacts, and the temporal resolution is
inferior to 2D-FLASH [20].
The aim of this study was to compare the image quality
of representative structures and sensitivity of artifacts in
modified SSFP sequences with improved temporal resolution at 1.5 T and 3 T and to evaluate if 3 T offers advantages
over 1.5 T and which of the modified sequences can be
recommended.
Since, to the best of our knowledge, the implementation
of dynamic MRI at 3 T has not yet been evaluated, we
compared the image quality of parallel imaging-accelerated
SSFP techniques with a standard dynamic SSFP sequence
at 1.5 T and 3 T.
Materials and methods
Magnetic resonance imaging
Examinations were performed on two whole-body MR
systems with field strengths of 1.5 T and 3 T (Magnetom
Avanto/Trio, Siemens Medical Solutions, Erlangen, Germany). Both systems were equipped with a similar multichannel receiver system and identical gradient coil strength
[45 mT/m (72 mT/m effective)]. Imaging was performed
using a matrix coil system with 2 rings of 6 elements each
(total 12 elements).
Five different dynamic sequences were evaluated for
lung MRI: An established true fast imaging with steadystate precession (SSFP, Siemens = TrueFISP) sequence
with implemented parallel imaging at a factor of 2 (parallel
imaging factor 2, in the following referred to with the
acronym of the manufacturer as “PAT”) as recommended in
the literature [20] served as the baseline and standard of
reference (three images/s).
Improvement of temporal resolution was achieved by
reducing spatial resolution, using higher sequence bandwidth to lower TR (repetition time) and TE (echo time) and
changes in the settings of parallel imaging algorithms. This
resulted in two different modifications with either a
temporal resolution of nine frames per second (GRAPPA
PAT 2, TSENSE PAT 3) or a temporal resolution of ten
frames per second (GRAPPA PAT 3). Detailed sequence
parameters are specified in Table 1. These sequences were
compared to a dynamic GRE technique with eight frames
per second (fast low-angled shot gradient echo technique;
FLASH). These protocols were transferred to the 3-T
scanner, keeping imaging parameters as similar as possible.
Minor necessary differences demanded by physical constraints (e.g., imaging time) are documented in Table 1.
To overcome the limitations of phantom, for the
experiments with simulated pulmonary lesions (nodules
and infiltrates) in static conditions, four healthy subjects
were examined additionally.
Phantom experiments
Five porcine heart-lung preparations were examined using
a dedicated chest phantom as previously described [21]. A
6.5-mm tracheal tube was positioned into the trachea, and
the lungs were inflated by evacuation of the artificial
pleural space using a negative pressure of -20 to -30 hPa.
For simulation of lung nodules, agarose gel (Agarose 5 g/l,
BD, Franklin Lakes, NJ) was injected into one lung using a
50-ml syringe. Nodule volumes varied between 0.5 and
2.0 ml. For simulation of infiltrates, 50 ml of diluted
393
Table 1 Sequence parameters for lung MRI at 1.5 T and 3 T. Numbers in parentheses () indicate differing values at 3 T compared to 1.5 T
Sequence
FLASH 2D
origin SSFP
SSFP
GRAPPA PAT2
SSFP
GRAPPA PAT3
SSFP
TSENSE PAT3
TR
TE
TA
BW
Flip
angle
FOV
FOV
phase
Base
resolution
Phase
resolution
Slice
thickness
Time
resolution
(ms) (@3 T) (ms)
s (@3 T) Hz/pixel
(°)
(mm)
(%)
(mm)
(%)
(mm)
acquis./s
2.34
2.9 (2.9)
0.96
1.22
25
20
980
890
5
65
350
350
87.5
100
128
256
75
66
12
12
8
3
2.4 (2.7)
1.04
14 (16)
1,240
62
350
91.7
192
75
10
9
2.4 (2.7)
1.04
12 (14)
1,240
62
350
91.7
192
75
10
10
2.4 (2.7)
1.04
14 (16)
1,240
55
350
91.7
192
75
10
9
agarose gel was instilled into the bronchial system of the
contralateral lung during reinflation. During the MRI, the
lungs were ventilated at 8/min by a diaphragmatic pump;
for details, see [22]. For the purpose of correlation and
documentation of simulated lesions, helical CT of all
phantoms was performed using a tube voltage of 120 kV
and a tube current of 560 mA (64-slice CT; Somatom
Sensation 64, Siemens Medical Solutions, Forchheim,
Germany). Images were acquired with a slice collimation
of 0.6- and 1-mm axial and coronal reconstructions with an
increment of 0.5 mm using soft tissue (B30), and lung
kernels (B50) were performed (Fig. 3).
Subjects
Fig. 1 a-e Dynamic lung MRI of the porcine lung phantom,
simulated infiltrates in the left lower lobe at 1.5 T (upper line) and
3 T (lower line) using five different pulse sequences: a FLASH,
b SSFP, c SSFP GRAPPA PAT2, d SSFP GRAPPA PAT3, e SSFP
TSENSE (for sequence details, see Table 1)
Since it appeared unacceptable for patients with severe
lung disease to undergo the full protocol at both scanners,
we decided to evaluate the sequences in four healthy nonsmoking subjects (median age: 32.5 years; age range: 30–
36 years) according to the guidelines of the local ethics
committee. Informed written consent was obtained prior to
the examinations. Scan direction was head first supine,
arms positioned at the side of the body. No ECG-triggering
or respiratory gating was used. All measurements were
acquired in free breathing and during repeated deep in- and
expiration as instructed by the operator of the MR system.
394
Image analysis
Table 3 Results of ranking by three independently observers
(median/mode) for 3-T lung MRI, phantom
Phantom studies as well as examinations of subjects were
evaluated independently by three radiologists with different experience in reading chest MRI (3, 5 and 9 years MRI
experience) using a double monitor PACS workstation
(Chili Workstation, version 2.3, Heidelberg, Germany).
Data sets were prepared and loaded before by a fourth, nonreading radiologist. The images were evaluated using
parallel cine displays on the monitors and comparable
window settings. The applied workstation allowed for up to
eight parallel cine displays. Image evaluation was started
when the viewer indicated real-time conditions for the cine
display. The cine display could be halted, if necessary. The
observers were blinded to sequence type, field strength,
study time, subject name and phantom number. The series
were displayed simultaneously, but with variable order on
the double screen of the workstation. Criteria for evaluation
were determined as followed:
Phantom 3 T
1. Image quality andside-by-side comparison: Image
ranking and comparison were done for all sequence
variations acquired at both field strengths. Image
quality was ranked in respect to (a) type and extent
of artifacts (e.g., parallel acquisition, field heterogeneity and motion artifacts), (b) the delineation of
anatomical structures (mediastinal structures, diaphragm and chest wall), (c) detection of lesions (lung
nodules and infiltrates) within the chest phantom, and
(d) overall impression of dynamic appearance using a
five-point scale (best image quality = 1, worst = 5). For
sensitivity of artifacts a three-point scale was used (0 =
absence of artifacts; 1 = artifacts, diagnostic; 2 = severe
artifacts, non-diagnostic).
The readers were asked if the image quality was equal or
if they preferred one to the other.
2. Signal intensity measurements of nodules and infiltrates:
Measurements were performed for all sequences by an
independent observer. Since parallel imaging was used
for all sequences, the non-uniform distribution of noise
Table 2 Results of ranking by three independently observers
(median/mode) for 1.5-T lung MRI, phantom
SSFP
GRAPPA
PAT3
2/2
4/4
5/5
4/4
2/2
2/2
SSFP
TSENSE
1/1
3/3
3/2
3/3
2/2
2/2
obviated valid SNR measurements. Instead, the lesion
contrast (CNODULES and CINFILTRATES) defined by the
ratio of the signal intensity of the lesion and adjacent
normal lung tissue was determined. For this purpose,
circular regions of interest with an average number of
50 pixels were positioned in the lesions as well as the
adjacent normal lung tissue[23].
Statistical evaluation
For statistical evaluation, the median and mode of the
rankings were calculated. As the ranking of the sequences
resulted in ordinal data (1 = best quality, 5 = worst quality),
the median and mode of the ranking were calculated; mode
function represents the most frequently occurring value in a
list of numbers.
For the comparison of the sequences at different field
strengths, we calculated the percentage for three categories:
1. Equal image quality
2. 1.5 Tesla preferred
3. 3 Tesla preferred.
CT scans in the phantom studies were not further evaluated
in detail, but served as a reference for the identification of
nodules and infiltrates as a ground truth (Fig. 3).
Table 4 Results of ranking by three independently observers
(median/mode) for 1.5-T lung MRI, subjects
Phantom 1.5 T
FLASH SSFP SSFP
2D
GRAPPA
PAT2
Artifacts
0/0
1/1
1/1
Diaphragm
2/2
1/1
3/3
Mediastinum 2/2
1/1
3/3
Chest wall
2/2
1/1
3/3
Infiltrates
2/2
1/0
2/2
Nodules
1/1
1/2
2/2
FLASH SSFP SSFP
2D
GRAPPA
PAT2
Artifacts
0/0
1/1
2/2
Diaphragm
1/1
2/2
5/5
Mediastinum 2/1
1/1
4/4
Chest wall
1/1
2/2
4/4
Infiltrates
1/2
2/2
2/2
Nodules
1/1
2/2
2/2
SSFP
GRAPPA
PAT3
1/1
4/4
5/5
3/4
2/2
2/2
SSFP
TSENSE
1/1
5/5
4/4
5/5
2/2
2/2
Subject 1.5 T
FLASH SSFP SSFP
2D
GRAPPA
PAT2
Artifacts
0/0
0/0
1/1
Diaphragm
2/1
1,5/1 3/3
Mediastinum 3/2
1/1
3/3
Chest wall
2,5/2
1/1
3/3
SSFP
GRAPPA
PAT3
2/2
5/5
5/5
5/5
SSFP
TSENSE
1/1
3/3
4/4
4/4
395
Table 5 Results of ranking by three independently observers
(median/mode) for 3-T lung MRI, subjects
Subject 3 T
FLASH SSFP SSFP
2D
GRAPPA
PAT2
Artifacts
0,5/1
1/1
2/2
Diaphragm
2/1
1/1
3/3
Mediastinum 1/1
2/2
3,5/4
Chest wall
1/1
2/2
4/4
SSFP
GRAPPA
PAT3
2/2
4/4
4/4
3/3
SSFP
TSENSE
2/2
5/5
5/5
5/5
Results
The comparison revealed different preferences for sequences in respect to different criteria at both field
strengths and conditions (i.e., chest phantom vs. healthy
subjects; Tables 2, 3, 4 and 5).
Phantom experiments
1. Image quality and side-by-side comparison:
a. Type and extent of image artifacts
As expected, GRE showed the fewest artifacts at
both field strengths followed by the baseline SSFP
and the TSENSE SSFP (Tables 2, 3, 4 and 5).
Fig. 2 a-e Dynamic lung MRI using porcine lung phantom,
simulated lung nodules in the right upper lower lobe at 1.5 T (upper
line) and 3 T (lower line) using five different pulse sequences:
Modified SSFP sequences with GRAPPA PAT 2
and PAT 3 showed the most artifacts at both field
strengths at all. However, artifacts were most
prominent at 3 T, in up to 56% resulting in nondiagnostic image quality. All readers found more
artifacts in images of the chest phantom compared
to subjects.
b. Delineation of anatomical structures
For the delineation of mediastinal anatomy, the
baseline SSFP sequence was preferred at both field
strengths [median/mode (1/1), Tables 2 and 3]. In
the phantom study the baseline SSFP sequence was
rated best for delineation of the diaphragm at 1.5 T
(1/1) (Table 2); GRE was preferred at 3 T (1/1)
(Table 3). For the separation of the chest wall the
baseline SSFP sequence achieved the best ranking
at 1.5 T (1/1) (Table 2), (3 T: 2/2), (Table 3).
c. Detection of lesions
In detection of the simulated lesions GRE was rated
best for the detection of nodules (1/1) at both field
strengths (Tables 2 and 3). The baseline SSFP
sequence was rated best in visualization of
infiltrates at 1.5 T (1/1) (Table 2), whereas GRE
was preferred at 3 T (1/2) (Table 3).
The simulated lung abnormalities are demonstrated
in Figs. 1 and 2, their correlation to the CT scan in
Fig. 3.
2. Signal intensity measurements of nodules and infiltrates:
a FLASH, b SSFP, c SSFP GRAPPA PAT2, d SSFP GRAPPA
PAT3, e SSFP TSENSE) (for sequence details, see Table 1).
396
Fig. 3 a-b Correlation of lesions in CT and MR scans: a infiltrates (baseline SSFP, 3 T), b nodules (GRE, 3 T)
Comparison of signal intensities at 1.5 and 3.0 T
At 1.5 T, the lesion-to-background contrast in nodules
ranged from 2.4 (2.04; 2.18) on the SSFP images with R=3
to 4.7 (3.16; 6.23) on the baseline SSFP sequence. The
GRE sequence showed an intermediate lesion-to-background contrast of nodules 3.7 (2.52; 5.02) (Table 6).
For the infiltrates, the lesion-to-background contrast was
lowest on the SSFP images with TSENSE 2.1 (1.39; 2.83) and
highest on the baseline SSFP sequence 4.3 (3.24; 5.29) with
intermediate results for the GRE sequence, 2.21 (1.88; 2.54)
(Table 6).
At 3 T, the signal intensity measurements showed an
increase in the lesion-to-background contrast of simulated
lung nodules compared to 1.5 T for all sequences. This effect
was most prominent on GRE images with an increase of 31%
from 1.5 T to 3 T and least prominent on the modified SSFP
sequence with a parallel factor of 3. The contrast of infiltrates
on GRE increased by 21% from 1.5 T to 3 T (Table 6). In the
side-by-side comparison, the GRE sequence was rated
superior for 3 T in nearly 70%, whereas all other sequences
were ranked superior at 1.5 T between 61 and 94% (Table 7).
Healthy subjects
1. Image quality and side-by-side comparison:
a. Type and extent of image artifacts
GRE images showed the fewest artifacts at both field
strengths (1.5 T: 0/0, 3 T: 0.5/1) (Tables 4 and 5) followed by
the baseline SSFP and the TSENSE. Modified SSFP
sequences with GRAPPA PAT2 and PAT3 showed the most
artifacts at both field strengths at all (Tables 4 and 5) (Fig. 4).
Fig. 4 a-e Dynamic lung MRI of a healthy subject at 1.5 T (upper line) and 3 T (lower line) using five different pulse sequences: a FLASH,
b SSFP, c SSFP GRAPPA PAT2, d SSFP GRAPPA PAT3, e SSFP TSENSE (for sequence details, see Table 1)
397
Table 6 Lesion-to-background contrast (nodules and infiltrates at 1.5 T and 3 T, mean, standard deviation, 95% confidence interval)
Contrast nodule
Sequence
FLASH 2D
Origin SSFP
SSFP GRAPPA PAT2
SSFP GRAPPA PAT3
SSFP TSENSE
Contrast infiltrates
Sequence
FLASH 2D
Origin SSFP
SSFP GRAPPA PAT2
SSFP GRAPPA PAT3
SSFP TSENSE
1.5 T
3T
Mean
SD
95% CI
Mean
SD
95% CI
SNR increase(%)
3.77
4.70
2.96
2.44
2.56
1.5 T
Mean
1.96
2.41
0.12
0.63
0.60
[2.52;5.02]
[3.16;6.23]
[2.88;3.03]
[2.04;2.18]
[2.84;2.94]
0.92
1.40
1.74
0.29
0.65
[4.90;6.10]
[4.49;6.26]
[2.99;5.20]
[2.35;2.71]
[2.41;3.23]
31
13
28
4
9
Std
95%CI
5.48
5.38
4.10
2.53
2.82
3T
Mean
SD
95% CI
SNR increase(%)
2.21
4.27
2.64
2.26
2.11
0.51
1.62
1.42
0.35
1.13
[1.88:2.54]
[3.24;5.29]
[1.74;3.54]
[2.03;2.48]
[1.39;2.83]
2.79
4.52
2.37
2.79
1.98
1.06
3.15
1.32
1.52
0.28
[2.06;3.40]
[2.52;6.51]
[1.52;3.21]
[1.82;3.76]
[1.80;2.15]
21
6
-11
19
-6
Artifacts were most prominent at 3 T, in up to 60%
resulting in non-diagnostic image quality (Fig. 4, lower
line, especially 4d and 4e). Compared to the baseline SSFP
and GRE, SSFP with a parallel imaging factor 2 was the
preferred modification to the baseline SSFP sequence at
both field strengths. GRAPPA PAT 3 and TSENSE PAT3
did not show any substantial benefit.
b. Delineation of anatomical structures
For the delineation of mediastinal anatomy in subject
studies, the baseline SSFP was ranked best at 1.5 T (1/1)
(Table 4), whereas GRE was rated best at 3 T (1/1)
(Table 5). The baseline SSFP sequence achieved the best
ranking for delineation of the diaphragm at both field
strengths (1.5 T: 1.5/1; 3 T: 1/1) (Tables 4 and 5). For the
separation of the chest wall, the baseline SSFP sequence
achieved the best ranking at 1.5 T in the healthy subjects
(1/1) (3 T: 2/2) (Tables 4 and 5).
Discussion
The comparison of the image quality and sensitivity of
artifacts in modified SSFP sequences with improved
temporal resolution in this study showed feasibilities of
dynamic lung MRI at 3 T but implies further challenges.
As known, (non-dynamic) lung MRI is already difficult
at 1.5 T due to the low proton density and the high
magnetic susceptibility of lung tissue. Multiple air-tissue
interfaces result in large local magnetic gradients that cause
intravoxel phase dispersion [24, 25]. This makes lung MRI
even more challenging at 3 T. Nevertheless, recent studies
have shown the feasibility of MRI at 3 T to detect diffuse
lung disease [19, 26] and the increasing availability of 3-T
MR systems in clinical routine (in some cases as alternative
to 1.5 T) necessitates optimization of protocols for
(dynamic) lung MRI (Table 8).
In a previous study, five different pulse sequences were
implemented at 1.5- and 3-T systems, and image quality
was compared [27]. To comply with this development, the
dynamic sequences evaluated in the present study as they
were optimized for a common 1.5-T system were also
transferred to a 3-T system. In different studies regarding
cardiac cine MRI signal as well as the increase of artifacts
at SSFP sequences are described [28, 29].
To the best of our knowledge, dynamic lung MRI
sequences have not been evaluated at 3 T so far.
Table 7 Comparison of ranking, subjects and percentages
Table 8 Comparison of ranking, phantom studies and percentages
Sequence
Equal
1.5 T superior
3 T superior
Sequence
Equal
1.5 T superior
3 T superior
FLASH 2D
SSFP
SSFP GRAPPA PAT2
SSFP GRAPPA PAT3
SSFP TSENSE
30.5%
27.7%
8.3%
19.4
5.5%
0%
61.1%
75.0%
63.8%
94.4%
69.4%
11.1%
16.6%
16.6%
0%
FLASH 2D
SSFP
SSFP GRAPPA PAT2
SSFP GRAPPA PAT3
SSFP TSENSE
33.3%
20.0%
20.0%
17.7%
42.2%
6.6%
66.6%
66.6%
75.5%
37.7%
60.0%
13.3%
13.3%
6.6%
20.0%
398
As a baseline for the study, we used two dynamic lung
MRI sequences proposed for 1.5-T systems (FLASH,
SSFP). In addition to this, three modifications of the SSFP
sequence using parallel acquisition techniques with PI
factors of 2 and 3 at 1.5 and 3 T (GRAPPA and TSENSE)
were included.
The comparison of the two field strengths was facilitated
by the fact that both scanners were equipped with similar
gradient systems. The imaging parameters were kept as
similar as possible. However, SAR limits and lengthening
of T1 forced adapting the imaging parameters slightly
(Table 1). In particular, it was difficult to maintain the same
temporal resolution after transferring the sequence protocols. This implies that, already from the sequence setup,
imaging at higher field strength appeared not to be
beneficial. As expected, the baseline SSFP achieved the
best image quality at 1.5 T.
SSFP with GRAPPA PAT 2 achieved a significantly
improved temporal resolution of nine acquisitions per second
and was still ranked second best by overall image quality.
Modifications with PAT3 (GRAPPA, TSENSE) were more
prone to artifacts and were of no additional benefit.
At 3 T, contrast and signal of GRE improved, while
SSFP image quality suffered from artifacts. Furthermore,
both modified SSFP sequences with a parallel factor of 3
showed severe artifacts, partially resulting in non-diagnostic images. By comparison, GRE showed the fewest
artifacts at 3 T. Except for GRE, image quality at 1.5 T was
rated superior to 3 T for all sequences.
The evaluation of the phantom study revealed some
general differences between phantom and subject data. In
particular, more artifacts were seen in the phantom at 1.5 T.
This may be explained by the structure of the phantom
chest wall and its flanges, which include interfaces of
different materials. Another factor was the in- and outflow
of water into the diaphragmatic bubble inducing heavy
flow artifacts. At 3 T the presence of artifacts on phantom
studies was comparable with that found in subjects.
Nevertheless, even with this given bias, the phantom
experiments revealed useful information on how lung
lesions were displayed. At 3 T, GRE was ranked best for
conspicuity of simulated lung nodules in the phantom
experiments. For delineation of the anatomical structures
GRE as well as baseline SSFP achieved the best ratings at
both field strengths, perhaps because of the increase of the
signal-to-background ratio as well as the slightly higher
spatial resolution of baseline SSFP. The SI measurements
were consistent with the imaging parameters. In addition,
the modified SSFP sequences offered a smoother fluid
view of the respiration process, obviously due to the
increased images per second as estimated by all readers.
A first finding in the healthy subjects was an apparently
higher signal intensity increase after transferring the
protocol from 1.5 to 3 T. Other sequences showed this
effect to a lesser extent. Due to the small number of
subjects with variable parameters such as respiratory and
cardiac motion, SI measurements were performed in the
phantom study.
Compared to imaging at 1.5 T, lesion-to-lung-tissue
signal ratios at 3 T indicated a better lesion detection
ability. This effect was more prominent for GRE than for
the other sequences. The GRE sequence showed a
significant increase of lesion signal intensity at 3 T
(31%) with the fewest image artifacts.
In particular SSFP with parallel imaging factors of 3
(GRAPPA PAT 3 and TSENSE PAT3) showed an even
lower signal of the lesions at 3 T compared to 1.5 T.
However, the difference between 1.5 and 3 T was only
significant for the GRE sequence (p<0.001).
In conclusion, the modified SSFP offered an equal
temporal resolution with superior spatial resolution and
signal compared to GRE at 1.5 T. However, image quality
was lower compared to the original SSFP sequence.
GRAPPA 3 and TSENSE PAT3 were of no additional
benefit. All in all, modified SSFP using GRAPPA PAT2
may be a good compromise in respect to temporal and
spatial resolution at 1.5 T. At 3 T, contrast and signal of 2D
GRE improved, while SSFP image quality suffered from
severe artifacts. Under the assumption that the sequence
protocols might be further optimized for 3 T, dynamic lung
MRI for clinical purposes appears to be feasible at 3 T,
preferably with GRE as the method of choice.
References
1. Antoch G, Vogt FM, Freudenberg LS et
al (2003) Whole-body dual-modality
PET/CT and whole-body MRI for
tumor staging in oncology. JAMA
290:3199–3206
2. Antoch G, Vogt FM, Bockisch A,
Ruehm SG (2004) [Whole-body tumor
staging: MRI or FDG-PET/CT?].
Radiologe 44:882–888
3. Schmidt GP, Haug AR, Schoenberg
SO, Reiser MF (2006) Whole-body
MRI and PET-CT in the management
of cancer patients. Eur Radiol 16:1216–
1225
4. Ghanem N, Altehoefer C, Kelly T et al
(2006) Whole-body MRI in comparison to skeletal scintigraphy in detection
of skeletal metastases in patients with
solid tumors. In Vivo 20:173–182
5. Schafer JF, Fischmann A, Lichy M et al
(2004) [Oncologic screening with
whole-body MRI: possibilities and
limitations]. Radiologe 44:854–863
6. Biederer J, Puderbach M, Hintze C. A
Practical Approach to Lung MRI at
1.5T. Magnetom Flash. 02/2006 ed,
2007
399
7. Hintze C, Biederer J, Wenz HW,
Eberhardt R, Kauczor HU (2006) [MRI
in staging of lung cancer.]. Radiologe
46:251–259
8. Eibel R, Herzog P, Dietrich O et al
(2006) Pulmonary abnormalities in
immunocompromised patients: comparative detection with parallel acquisition MR imaging and thin-section
helical CT. Radiology 241:880–891
9. Eibel R, Herzog P, Dietrich O et al
(2006) [Magnetic resonance imaging in
the evaluation of pneumonia].
Radiologe 46:267–4
10. Leutner C, Gieseke J, Lutterbey G et al
(1999) [MRT versus CT in the diagnosis of pneumonias: an evaluation of a
T2-weighted ultrafast turbo-spin-echo
sequence (UTSE)]. Rofo 170:449–456
11. Leutner CC, Gieseke J, Lutterbey G et
al (2000) MR imaging of pneumonia in
immunocompromised patients: comparison with helical CT. AJR Am J
Roentgenol 175:391–397
12. Eichinger M, Puderbach M, Fink C et
al (2006) Contrast-enhanced 3D MRI
of lung perfusion in children with cystic
fibrosis–initial results. Eur Radiol
16:2147–2152
13. Puderbach M, Eichinger M,
Haeselbarth J et al (2007) Assessment
of morphological MRI for pulmonary
changes in cystic fibrosis (CF) patients:
comparison to thin-section CT and
chest x-ray. Invest Radiol 42:715–725
14. Plathow C, Hof H, Kuhn S et al (2006)
Therapy monitoring using dynamic
MRI: analysis of lung motion and
intrathoracic tumor mobility before and
after radiotherapy. Eur Radiol 16:1942–
1950
15. Dinkel J, Welzel T, Bolte H, et al.
(2007) Four-dimensional multislice helical CT of the lung: Qualitative comparison of retrospectively gated and
static images in an ex-vivo system.
Radiother Oncol
16. Puderbach M, Hintze C, Ley S,
Eichinger M, Kauczor HU, Biederer J
(2007) MR imaging of the chest: A
practical approach at 1.5T. Eur J Radiol
17. Kluge A, Gerriets T, Muller C et al
(2005) [Thoracic real-time MRI: experience from 2200 examinations in acute
and ill-defined thoracic diseases]. Rofo
177:1513–1521
18. Schafer JF, Vollmar J, Schick F et al
(2005) [Detection of pulmonary nodules with breath-hold magnetic resonance imaging in comparison with
computed tomography]. Rofo 177:41–
49
19. Lutterbey G, Gieseke J, von FM,
Morakkabati N, Schild H (2005) Lung
MRI at 3.0 T: a comparison of helical
CT and high-field MRI in the detection
of diffuse lung disease. Eur Radiol
15:324–328
20. Biederer J (2005) [Magnetic resonance
imaging: technical aspects and recent
developments]. Med Klin (Munich)
100:62–72
21. Biederer J, Heller M (2003) Artificial
thorax for MR imaging studies in
porcine heart-lung preparations.
Radiology 226:250–255
22. Biederer J, Plathow C, Schoebinger M
et al (2006) Reproducible simulation of
respiratory motion in porcine lung
explants. Rofo 178:1067–1072
23. Dietrich O, Raya JG, Reeder SB, Reiser
MF, Schoenberg SO (2007) Measurement of signal-to-noise ratios in MR
images: influence of multichannel coils,
parallel imaging, and reconstruction
filters. J Magn Reson Imaging 26:375–
385
24. Hatabu H, Gaa J, Tadamura E et al
(1999) MR imaging of pulmonary
parenchyma with a half-Fourier singleshot turbo spin-echo (HASTE) sequence. Eur J Radiol 29:152–159
25. Hatabu H, Chen Q, Stock KW, Gefter
WB, Itoh H (1999) Fast magnetic
resonance imaging of the lung. Eur J
Radiol 29:114–132
26. Lutterbey G, Grohe C, Gieseke J et al
(2007) Initial experience with lungMRI at 3.0 T: Comparison with CT and
clinical data in the evaluation of interstitial lung disease activity. Eur J Radiol
61:256–261
27. Fink C, Puderbach M, Biederer J et al
(2007) Lung MRI at 1.5 and 3 Tesla:
observer preference study and lesion
contrast using five different pulse
sequences. Invest Radiol 42:377–383
28. Gutberlet M, Schwinge K, Freyhardt P
et al (2005) Influence of high magnetic
field strengths and parallel acquisition
strategies on image quality in cardiac
2D CINE magnetic resonance imaging:
comparison of 1.5 T vs. 3.0 T. Eur
Radiol 15:1586–1597
29. Tyler DJ, Hudsmith LE, Petersen SE et
al (2006) Cardiac cine MR-imaging at
3T: FLASH vs SSFP. J Cardiovasc
Magn Reson 8:709–715