Robert Thomas
C. Smith, MD R. McCauley,
Fast ofthe
In this
Spin-Echo Female I#{149} Use
Part
Caroline Reinhold, MD #{149}Ruben Kier,
#{149}
ofa
prospective
Whole-Volume
study,
magnetic
Robert C. Lange, Shirley McCarthy,
terms:
sequences
resonance
Magnetic Pelvic
#{149}
S
resonance (MR), organs, MR. 85.1214
pulse
1992;
184:665-669
sary for structure
T2-weighted
pulse
visualization and detection
of gynecobogic of most pelvic
abnormalities. The acquisition time for T2-weighted images with conventional sequences is approximately 9 minutes. These images frequently contain phase ghost artifacts and blurring caused by both voluntary and involuntary physiologic motion (including respiratory motion, bowel peristalsis,
and
These artifacts not eliminated, tory
motion
vascular
pulsation).
can be reduced, but by means of respiracompensation
tech-
niques, administration of glucagon, and presaturation of inflowing blood. The long examination times and cost currently prevent MR imaging from becoming the initial imaging modality for most pelvic abnormalities. Conventional spin-echo (CSE) pulse sequences acquire echoes at fixed times after each section-selective 90#{176} excitation pulse. When MR imaging is performed with long repetition times (TRs), multiple echoes two)
are
usually
These echoes have the encoding (ie, the phase changed
between
acquired.
same phase encoding is measurements
of the separate echoes). This allows reconstruction of multiple images per section location, each with a different echo time (TE). For a two-echo, long-TR sequence, this yields intermediate and T2-weighted images at each location. The phase-encoding gradient of a CSE
I From the Department of Diagnostic ing, Yale University School of Medicine, dar St, New Haven, CT 06510. Received ber 18, 1991; revision requested January revision received January 30; accepted 24. Address reprint requests to R.C.S. C RSNA, 1992
(SE)
PIN-ECHO
sequences are essential for magnetic resonance (MR) imaging of the body, particularly for MR imaging of the female pelvis. T2 contrast is neces-
not Radiology
MD
Coil’
(typically Index
PhD PhD,
#{149} #{149}
MR Imaging Pelvis
axial and (MR) images were obtained with T2weighted conventional spin-echo (CSE) and fast spin-echo (FSE) sequences in 34 consecutive female patients who underwent clinical pelvic MR examination at 1.5 T. The MR images from each patient were compared side by side, blindly and independently, by two radiologists experienced in MR imaging who used a standardized score sheet for anatomic and pathologic findings. The FSE sequences were rated superior significantly more often than the CSE sequences in most categories of findings (P < .05), including overall image quality and reduction of motion artifact. The examination time for the FSE sequences was 1 minute 46 seconds versus an examination time of 9 minutes 14 seconds for the CSE sequences. (Both CSE and FSE Sequences provided 18 sections.) It is concluded that the FSE sequence provides T2-weighted anatomic and pathologic information superior to that provided by the CSE sequence and requires substantially less imaging time. sagittal
MD MD
Imag333 CeNovem7, 1992; February
sequence
induces
phase
changes
necessary for positional information along the phase axis. On the other hand, gradient-induced phase changes cause signal loss. The amount of signal loss is proportional to the strength of the gradient. Therefore, echoes acquired with bow-magnitude phase gradients provide most
of the
signal
images
and
contrast
of CSE
MR
(1-3).
To increase imaging speed, a different phase-encoding gradient can be used for each echo of a multiecho train after a 90#{176} radio-frequency pulse. This increases the amount of data acquired per section per TR interval. Such a sequence was initially described by Hennig et ab (4,5) and was
referred
to as rapid
acquisition
with relaxation enhancement Recently, a variant of this has
been
described
and
(RARE). sequence used
for
clini-
cab MR imaging by Melki et al (6). Their variant of the RARE sequence is referred to as fast-acquisition interleaved SE. More recently, this has been shortened to fast SE (FSE). The FSE sequence acquires a train of up to 16 SEs after each section-selective 90#{176} pulse. This is achieved by applying multiple 180#{176} pulses in rapid succession to generate multiple echoes. A different phase-encoding gradient is used for each echo. (The phase-encoding gradients are applied after each 180#{176} pulse and then rewound prior to the next 180#{176} pulse.) The number of 180#{176} pulses applied is referred to as the echo train length (ETL),
and
the
oes is referred (E-space).
spacing
between
to as the
When
the
echo
phase
ech-
spacing gradients
are ordered so that, for each section, the low-magnitude gradients are used for the nth echo (n 16) of each echo train, an MR image with an effective TE (or pseudo-TE) of n times E-space is produced
(6).
TR, number
of phase
(NPE),
and
number
Therefore,
for
equal
encodings of signals
aver-
aged (NSA), the FSE sequence can reduce imaging time by a factor equal to ETL. The imaging time for the FSE
Abbreviations: CSE = conventional spin echo, E-space = echo spacing, ETL = echo train length, FOV = field of view, FSE = fast spin echo, RARE = rapid acquisition with relaxation enhancement, SE = spin echo, TE = echo time, TR = repetition time.
665
Comparison
of Image
Scorer
Scores
Based
FSE
CSE
I 2
27 29
5 3
1 2
23 26
5 0
on FSE
CSE Sequences
and
E
Overall
P Value
FSE
CSE
NA NA
.001
.001
27 29
3 1
Cervical
.001 .001
18 10
6 5
Margins 2 7 Anatomy
Reduction
4 4
NA NA
Zonal
Anatomy
.001 .001
(CE)
8 17
.01 NS
2 2
CervicalZonal
(IZ)
P Value
N
Artifact
Cervical 4 1
Zonal
E
Motion
Quality 2 2
Cervical
N
Anatomy
(ST)
1 2
2
0
0
32
NS
19
8
6
1
.03
6
0
4
24
.02
12
3
17
2
.02
1 2
16
4
5
9
17
2
10
5
.006 .001
24 27
2 2
1 1
.001 .001
1
19
5
5
5
2
23
2
6
3
.004 .001
18 21
6 4
Cervical
Zonal
Uterine
Zonal
Anatomy
Left Ovary 1
16 11
2
Anatomy
2 4
13 12
1
2 1
6 10
4 1
2
1 2
0 2
1 0
.025 .002
5 9
1 0
NS NS
Ovarian
Right NS NS
1
4
7
1 0
1 2
6
1
5
1
I 2
3
29
NS
2
29
NS
Ovarian
Mass
31
NS
2
31
NS
1
26
.06
4
24
NS
3 0
0 0
Venous
Vaginal
Cysts 0
34
2
32
7 11
NS NS
22 22
.06 NS
22 32
NS NS
Margins 9 2
Ovarian
0
Plexus
5 10
Cysts
0
Lesions
2
6 1
0
.07 NS
0
1 1
0
15 16
0
NS .01
Bartholin
.03 NS
1
28 26
Nabothian
10 9
Implants
Endometrial
Perivaginal
1 1
.01 .001
1
Adenomyosis
2
4 2
Stroma
2 1
1 3
Mass Lesions 32 32
Ovarian
5 6 Right
29 31
OlE)
Identification
5 7
12 11
Implants
1 0
Ovary
Right 10 12
2 1
Left Ovarian
14 9
Anatomy
6 7
Stroma
Endometrial
1 1
Zonal
Right
.001 .06
Margins 7 4
Uterine
(M/J)
10 9
7 9
Left Ovarian
Uterine
Identification
Left Ovarian 1 2
(OZ)
Leiomyomas
3
5
19
NS
1
7
15
.003
Note-Numbers in the FSE and CSE columns indicate the number of cases in which each of the sequences was rated superior. cE = inner bright zone, E = number of cases for which the FSE and CSE sequences were rated equal, IZ = adjacent zone of intermediate signal intensity, J/E = junctional zone/endometrium, M/J = myometrium/junctional zone, N = number of cases in which a finding was not identified with either sequence, NA = not applicable, NS = not significant, OZ = outer zone continuous with the myometrium, ST = dark zone of cervical stroma.
sequence
is given
by the following (TR x NSA X NPE)/
equation: Tscan ETL. Depending and the E-space, tions
may
on the choice of ETL the number of sec-
be reduced
compared
with
a CSE sequence of equal TR. When an FSE sequence is used, weighted images can be acquired less ing
than times
motion
artifact.
report
of the
sibibity
666
These reduce
The
is to provide
tion male
1 minute. should
clinical
of the
pelvis. Radiology
#{149}
FSE
T2in
short imagphysiologic
purpose
of this
an
evalua-
initial
usefulness sequence
MR images
and in the
obtained
CSE
sequences
were
directly
com-
Committee
group
cab pelvic MR examinations. A direct comparison of image quality and detection of normal anatomic structures was made. It was not the purpose of this study to compare lesion detection
ages
ranged
age,
35 years).
tamed
from
in specific
diseases.
consisted
patients
the
fe-
with
MATERIALS Approval from
the
for this University
to us
during from
interview
study Human
study
for
MR
all before
during
examination
Their
(mean
consent
patients
of
period.
17 to 70 years
Informed
female imaging
was our
obroutine
with
MR
imaging. All MR
imaging
was
(Signa;
body GE
Medical
performed
coil and
with
a 1.5-T
Systems,
Mibwau-
series of coronal was obtained,
obtained
sagittal
T2-weighted
Investigation
quences
and were
axial
CSE
performed
with
a
sys-
kee). After a bocalizer weighted MR images
METHODS was
Our
a 10-week
whole-volume
AND
institution.
of 34 consecutive
referred
pelvis
tern
fea-
at our
pared with MR images obtained with FSE sequences at identical locations in a series of patients undergoing clini-
Tise-
the
September
follow-
1992
ian cysts, anatomy,
uterus margins, uterine leiomyoma of the uterus,
identification,
ovarian
stroma,
zonal ovary
individual
ovarian cysts, ovarian mass lesions, ovarian endometrial implants, adenomyosis, vaginal margins, perivaginal venous plexus, and Bartholin cysts. Each finding was
scored
for
superior,
conspicuity,
B superior,
fled. The
assessment
as follows:
equal,
or
of overall
not
A
identi-
image
quality
included quality of visualization of any abnormalities present. This enabled diagnosis whenever abnormalities were seen
on MR images quence
a.
b.
Figure
1.
the same structures
MR images
obtained with (a) CSE and patient. Note the improved visualization in b. A nabothian cyst is present within
(b) FSE sequences of the uterine the cervix.
at identical margins and
locations in uterine zone
but
with tected from
with
one se-
not on MR images
obtained
obtained
the other sequence. with both sequences this study. Evaluation
findings
was
the
and left anatomy
right zonab
me
performed
uations of the metrium/junctional
individually
for
sides. Evaluation included separate
interface
between and
zone
zone/endometnum. structures
Lesions not dewere excluded of all ovarian
the myojunctional
Evaluation
included
separate
of utereval-
of cervical evaluations
of four distinct zones: an inner bright zone, an adjacent zone of intermediate signal
intensity,
stroma,
a dark
and
an outer
zone
with the myometnum. ovarian endometriosis
accordance
with
Structures section
not locations
of cervical
zone
continuous
Adenomyosis were identified
previous present were
reports on the scored
and in
(7-12).
evaluated as not iden-
tified. Interobserver gory of finding
weighted each rately. as A
b.
Figure 2. (a) MR image obtained with CSE sequence shows poor definition margins because of motion artifact. (b) MR image obtained with FSE sequence cation fundal
in the same leiomyoma
patient (arrow).
as in a shows
clear
delineation
of the
uterine
and
of the uterine
margins
at identical as well
loas a
variability was analyzed
kappa
statistic
scorer were then In each category,
for
catea
(13). The
data
for
analyzed sepathe cases scored
B or not identified
=
the remaining
each with
were
cases
discarded,
were
statistical significance by tailed binomial distribution. were considered significant
tested
for
means of a oneThe results at P < .05.
RESULTS ing parameters: 2,000/20, 80;
section 2.5 128
thickness,
mm;
two
5 mm;
signals
intersection
averaged;
and
gap, a 256
matrix.
These parameters provided sections in 9 minutes 18 seconds. Respiratory compensation, no phase wrap, and superior and inferior saturation were used. Glucagon was not tered before MR imaging. After performed,
the
In each case, ing parameters
TR msec/TE msec = field of view (FOV), 28 cm;
CSE imaging sagittal and
x 18
were se-
quences were performed at the nine central section locations in each plane, which were selected on the basis of the CSE sequences. The nine central sections were chosen to include the uterus and ovaries. The FSE MR images were obtained with the following parameters: 2,900/126; ETL, 16; E-space, 18 msec; FOV, 28 cm; section thickness, 5 mm; intersection gap, 2.5 mm; two signals averaged; a 256 x 128 matrix; and no phase wrap. Superior and inferior saturation parameters tions
pulses were also used. provided nine section
184
Number
#{149}
locations identical
3
and
the
nine
for the CSE nine section
sequence rate films.
were
central
sequence locations
photographed
All photography
was
These boca-
images but were chosen trast and make the images
imagthe
onto sepaperformed did
not differThe standardCSE MR
to optimize as similar
conin
appearance as possible. Each case then consisted of two axial and two sagittab films of nine images per film. The FSE CSE or B case two ized
and
films were then randomly labeled A for each case. The four films for each were then compared side by side by radiologists experienced in MR imag-
ing (T.R.M., score
SM.), sheet
anatomic
and
quality,
motion cervical
who that
pathologic artifact zonal
used
listed
a standardthe
findings: reduction, anatomy,
All of the as
section and the for the FSE
radiologist (R.C.S., who interpret the images) to minimize ences in photographic technique. window and level settings were ized separately for the FSE and
margins,
in 53 seconds.
Volume
images,
and from
by one
pulses adminis-
sequences axial FSE
MR
all patient data were removed
following overall cervical naboth-
either
categories
were
rated
substantial agreement .80) or almost perfect (.81 K 1) except
(.61 K agreement for cervical zonal structures, ovarian stroma, and right ovary identification. The results of the comparisons of the MR images obtained with FSE sequences and those obtained with
CSE
sequences
are shown
separately
for each ble. This
of the two scorers in the table shows the number
cases
in
rior,
CSE
which was
CSE
were
was
not
Levels
rated visible
FSE was rated rated superior, equal, with
of significance
each category. The FSE sequences
supeFSE and
or the either
are
were
Taof
finding
sequence.
shown
for
superior
to the CSE sequences and enabled shorter examination times. The FSE sequences were rated superior significantly more often by both scorers for
Radiology
#{149} 667
the following categories of findings: overall image quality, reduction of motion artifact, and depiction of left ovarian stroma, uterine margins, uterme zonab anatomy, and cervical margins.
Typical
Figures findings differences
examples
1-4. no
anatomic the
in
of no
findings
CSE MR significantly
images more
FSE MR images
which
shown
categories
statistically significant existed. There were
or pathologic
which superior
the
are
In the other
nor
a pathologic
for
were often
any
rated than
cases
abnormality
in was
detected with one sequence but not with the other. Evaluation of vaginal margins and perivaginab venous plexus was limited by the fact that only the nine centrab section locations were compared. Thus, the axial MR images often did not cover this region of interest because
the
sections
dude
the
uterus
were
and
chosen
a. Figure Note
cystic
b. 3.
CSE (a) and
the
improved
right
FSE (b) MR images
obtained
visualization
of the
mass
to be a paratubal
adnexal
proved
ovarian
at identical
stroma
and
locations
uterine
in the same
margins.
patient.
At surgery,
the
cyst.
to in-
ovaries.
DISCUSSION This study provides an initial evabuation of the capacity of the FSE sequence to generate clinically useful T2-weighted
pelvis.
MR
Our
sequence T2-weighted
results provides MR
images
indicate
of the
that
female
this
superior quality images with signifi-
cantly shorter examination times than CSE sequences. This study did not formally compare the two techniques in detection or depiction of specific disease entities. The FSE sequence used in this study provides nine sections in 53 seconds with a TR of 2,900 msec. To provide more coverage of the pelvis and thereby include both the true and false pelvis, a concatenated acquisition 106 seconds long (providing 18 section locations) would be required. Alternatively,
nonconcatenated
acqui-
sitions with a TR of 5,800 msec and an ETL of 16 or with a TR of 2,900 msec and an ETL of 8 could be used. Both of these would provide 18 sections in 106 seconds. However, the time of data acquisition for each section bocation is doubled with the nonconcatenated sequences because the concatenated sequence acquires all of the data for half of the sections during the first 53 seconds and all of the data for the other half during the second 53 seconds. Thus, the nonconcatenated acquisitions would have more motion artifact. A longer TR (2,900 vs 2,000 msec) was used for the FSE sequence to provide adequate imaging coverage while a short imaging time was maintamed. Depending on the choice of 668
Radiology
#{149}
a. Figure cervical
b. 4.
MR images cancer.
The
(a) CSE MR image
obtained
in pregnant
examination
was
is severely
degraded
woman
performed
because
aged
for staging
23 years before
of fetal motion
with
a clinical
termination
artifact.
Fetal
diagnosis
of
of pregnancy.
and
placental
detail is absent. An area of increased signal intensity within the cervix (arrow) likely represents the tumor. (b) On the FSE MR image, obtained in 53 seconds, the discrete margins of the tumor and the surrounding fibrous stroma (curved arrow) can be confidently identified. Fetal and placental structures are delineated; the heart (small straight white arrow), liver (large straight white arrow), spine, cerebral ventricles, and umbilical cord (black arrows) are now discernible.
the
ETL,
the
number
can be obtained may be less than
with the
of sections an FSE number
with a CSE sequence with TR. We chose the maximum ETL to minimize longer TR only signal-to-noise
that sequence obtained
an identical possible
imaging time. The minimally increases ratio. Furthermore, in
the FSE sequence, we used an effective TE that exceeded the TE used in the
CSE
sequence.
The
effective
TE
was chosen to make the contrast characteristics of the gynecobogic structures as similar as possible between the two sequences. Identical values of these parameters do not yield images of identical contrast.
With
few
exceptions,
the
contrast
characteristics of the CSE and FSE MR images in this study are similar. The contrast characteristics of structures
affected by specific diseases on FSE MR images were not studied. The rebative signal from subcutaneous fat is greater on FSE MR images than on CSE MR images, a finding previously described not been knowledge,
(1,3,6). One finding previously described, and that was not
explored
in our
study
is that
that has to our formally
skeletal
muscle, leiomyomas, and intervertebral disks are more hypointense on the FSE than on the CSE MR images. This may be associated with the par-
September
1992
ticular choice of parameters used in our study. It has been noted that FSE sequences can cause image blurring when a bong EU and short effective TE are used (3,6). Even though a long ETL (16) was used, we did not observe any significant image blurring in comparison with the CSE MR images. This is consistent with the long effective TE chosen (126 msec). Although proton density-weighted MR images without blurring can be obtamed with the FSE sequence and a shorter ETL, we believe that proton density-weighted MR images rarely add any additional information. To ascertain whether a lesion contains lipid or blood, Ti-weighted MR images can be obtained with fat or water suppression (14). While proton den-
FSE sequences and have had no problems with power deposition. However, it is currently recommended that the FSE sequence not be used in patients over 250 lb (113.5 kg) or in patients whose body configuration is such that the area being imaged would touch the sides of the bore of the magnet. This precaution is prelimmary and may be changed in the future. Since completion of our study, we have noted improved quality of FSE images obtained with the body coil and a 256 x 192 matrix. This matrix increases imaging time by 50% compared with a 256 x 128 matrix. However, this amounts only to an increase from approximately 2 minutes per imaging
plane
to 3 minutes
per
184
#{149} Number
3
1.
Mulkern
RV, Melki
PS, Jakab
N, Higuchi
0,
Jolesz FA. Phase-encode order and its effect on contrast and artifact in single-shot rare sequences. Med Phys 1991; 18:10322.
3.
4.
5.
6.
1037. Twieg
DB.
lation
of the NMR imaging
The
k-space
trajectory
formu-
process
with
applications in analysis and synthesis of imaging methods. Med Phys 1983; 10:610621. Mulkern RV, Wong STS, Winalski C, Jolesz FA. Contrast manipulation and artifact assessment of 2D and 3D RARE sequences. Magn Reson Imaging 1990; 8:557-566. HennigJ, Nauerth A, Friedburg H. RARE imaging: a fast imaging method for clinical MR. Magn Reson Med 1986; 3:823-833. Hennigj, Friedburg H. Clinical applications and methodological developments of the RARE technique. Magn Reson Imaging 1988; 6:391-395. Melki PS, Mulkem RV, Panych LP, Jolesz
FA. Comparing the FALSE method with conventional dual-echo sequences. JMRI
imag-
ing plane. In addition, the minimum sity-weighted images might be helpecho spacing of the FSE sequence has ful in these circumstances, they been reduced to 13 msec. This prowould not be diagnostic. In patients vides additional sections per TR interwho undergo evaluation for bladder val. The FSE sequence can be used to carcinoma (a group of patients not included in this study), proton denacquire axial, sagittab, or coronal T2sity-weighted MR images can be very weighted MR images of the female helpful. pelvis with an examination time of 1 Because of the large number of 180#{176} minute 46 seconds, providing 18 section locations in each plane. A coronal pulses applied in rapid succession T2-weighted FSE sequence can be whenever the FSE sequence is used, power deposition becomes a concern. used as a localizing sequence. Sagittal In a previous study by Melki et ab (6), and axial T2-weighted FSE sequences a saline phantom was used to evalucan then be followed with a convenate power deposition associated with tional axial Ti-weighted sequence. FSE sequences. Temperature changes The total examination time for all four sequences is less than 10 minutes. If induced in the phantom were well one allows time for setup before exbelow the guidelines of the Food and Drug Administration, which recently amination and between sequences, an approved the FSE sequence used in entire pelvic MR study can be compbeted in less than 20 minutes. Therethis study for clinical imaging. Since completion of this study, we have exfore, FSE sequences enable marked amined more than 350 patients with reduction of imaging time compared with CSE sequences, while improving overall image quality. U
Volume
References
1991;
1:319-326.
7.
Mark AS, Hricak
8.
Adenomyosis and leiomyoma: differential diagnosis with MR imaging. Radiology 1987; 163:527-529. Arriv#{233}L, Hricak H, Martin MC. Pelvic endometriosis: MR imaging. Radiology 1989; 171:687-692. Togashi K, Ozasa H, Konishi I, et al. En-
9.
larged
uterus:
nomyosis and ing. Radiology
H, Heinrichs
LW, et al.
differentiation
between
leiomyoma with 1989; 171:531-534.
10.
Zawin
11.
Endometriosis: appearance and detection at MR imaging. Radiology 1989; 171:693696. Nishimura K, Togashi K, Itoh K, et al. Endometrial cysts of the ovary: MR imaging. Radiology 1987; 162:315-318. McCarthy S. MR imaging of the uterus.
12.
M, McCarthy
Radiology 13.
Kramer tistics.
dance. 14.
5, Scoutt
ade-
MR imag-
L, Comite
F.
1989; 171:321-322. MS. Feinstein
AR.
Clinical
biosta-
LIV. The biostatistics of concorClin Pharmacol Ther 1981; 29:111-
123. Kier R, Smith RC, McCarthy S. Value of lipid- and water-suppression MR images in distinguishing between blood and lipid within ovarian masses. AJR 1992; 158:321325.
Radiology
#{149} 669