Musculoskeletal Imaging • Original Research

Tokuda et al.
MRI of Soft-Tissue Tumors

Musculoskeletal Imaging
Original Research

MRI of Soft-Tissue Tumors:

Fast STIR Sequence as
Downloaded from www.ajronline.org by 36.80.244.245 on 06/14/18 from IP address 36.80.244.245. Copyright ARRS. For personal use only; all rights reserved

Substitute for T1-Weighted

Fat-Suppressed Contrast-Enhanced
Spin-Echo Sequence
Osamu Tokuda1 OBJECTIVE. The purpose of this study was to assess the value of the fast STIR sequence
Yuko Harada in comparison with the T1-weighted fat-suppressed contrast-enhanced sequence in the evalu-
Naofumi Matsunaga ation of soft-tissue tumors.
MATERIALS AND METHODS. Sixty-seven soft-tissue tumors imaged with both
Tokuda O, Harada Y, Matsunaga N STIR and T1-weighted fat-suppressed contrast-enhanced sequences were evaluated. The sig-
nal-to-noise and contrast-to-noise ratios of the tumors in comparison with normal muscle,
bone marrow, and fat were measured. Subjective image contrast between soft-tissue tumors
and the nearest normal tissue was evaluated by two observers. The observers classified the
soft-tissue tumors as benign or malignant using a 5-point scale, and sensitivity, specificity,
and accuracy were calculated. The results of the two readings were assessed with receiver op-
erating characteristic analysis.
RESULTS. The contrast-to-noise ratios of all tumors in comparison with muscle (p < 0.01),
bone marrow (p < 0.05), and fat (p < 0.05) were significantly higher on the fast STIR imag-
es than on the T1-weighted fat-suppressed contrast-enhanced images. Both observers’ mean
ratings of benign, malignant, and all tumors in comparison with muscle on fast STIR images
were significantly higher than those on T1-weighted fat-suppressed contrast-enhanced images.
For both observers, the mean sensitivity, specificity, accuracy, and area under the receiver op-
erating characteristic curve in evaluation of the fast STIR images did not differ significantly
from those in evaluation of the T1-weighted fat-suppressed contrast-enhanced images.
CONCLUSION. The fast STIR sequence is excellent for evaluation of soft-tissue tu-
mors, and contrast-enhancement is not always needed.

Keywords: contrast-enhanced, contrast-to-noise ratio,
MRI, signal-to-noise ratio, soft-tissue tumor, STIR
DOI:10.2214/AJR.09.2675
Received March 2, 2009; accepted after revision
May 13, 2009.
1
All authors: Department of Radiology, Yamaguchi
University Graduate School of Medicine, 1-1-1
Minamikogushi, Ube, Yamaguchi 755-8505, Japan.
Address correspondence to O. Tokuda
(chinupapa@goo.jp).
AJR 2009; 193:1607–1614
0361–803X/09/1936–1607
© American Roentgen Ray Society
T
he most promising and impor-
tant sequences for evaluation of
soft-tissue tumors are those that
entail fat suppression. Suppres-
sion of the relatively high signal intensity of
fat leads to more efficient use of the dynamic
range for display of tissue contrast on MR
images. Fat suppression also may be helpful
for reducing the severity of artifacts. The
conspicuousness of phase-encoding errors
caused by motion is proportional to the sig-
nal intensity of the moving structure. Be-
cause fat is a major contributor to many mo-
tion artifacts, fat suppression can be effective
in reducing their prominence [1, 2].
The conspicuousness of abnormal en-
hancement after injection of paramag-
netic contrast material can increase on
T1-weighted images with the use of fat sup-
pression [3–6]. T1-weighted fat-suppressed
contrast-enhanced imaging, which is the
combination of gadopentetate dimeglumine
contrast enhancement and fat-suppression
technique, is a superior method for evalu-
ating adrenal masses [7] and neoplastic and
inflammatory diseases of the spine [5, 8],
kidney [9], and head and neck [10–12]. T1-
weighted fat-suppressed contrast-enhanced
imaging also is useful for evaluation of soft-
tissue tumors because of the greater conspi-
cuity of these lesions after enhancement.
Use of T1-weighted fat-suppressed contrast-
enhanced imaging can improve lesion de-
tection, tissue characterization, and deter-
mination of tumor extent.
STIR technique entails an alternative MRI
sequence that suppresses the signal intensity
of fat and the additive effects of T1 and T2
mechanisms on tissue signal intensity [4, 13–
19]. STIR imaging is commonly used to de-
tect bone marrow lesions because it is sen-
sitive in the detection of tumor, edema, and
infection in bone marrow [3, 4, 17, 18]. Fast
STIR imaging is superior to T1-weighted fat-

AJR:193, December 2009 1607

 Tokuda et al.

suppressed contrast-enhanced imaging in the TABLE 1:  Histologic Types and Location of Tumors (n = 67)
evaluation of all bone marrow components, No. of
and the two techniques have comparable re- Histologic Type Location Lesions
sults in the evaluation of surrounding soft- Benign (n = 34)
tissue components [17]. To our knowledge,
Schwannoma Shoulder, forearm, thigh, lower thigh 9
however, the value of fast STIR imaging in
Neurofibroma Forearm, thigh 2
Downloaded from www.ajronline.org by 36.80.244.245 on 06/14/18 from IP address 36.80.244.245. Copyright ARRS. For personal use only; all rights reserved

comparison with T1-weighted fat-suppressed

contrast-enhanced imaging in the evaluation Glomus tumor Finger, toe 5
of soft-tissue tumors has not been reported. Desmoid Shoulder 1
The purpose of this retrospective study was
Hemangioma Thigh, lower thigh, palm, finger, upper arm 6
to assess the value of the fast STIR sequence
in comparison with the T1-weighted fat-sup- Solitary fibrous tumor Groin, thigh 3
pressed contrast-enhanced sequence in MRI Fibromatosis Palm 1
evaluation of soft-tissue tumors. Morton’s neuroma Third toe 1

Materials and Methods Lipoma Thigh, neck 4

A retrospective review was conducted of 85
MRI examinations performed from 2001 to 2007
to identify imaging findings of soft-tissue tumors
with histopathologic confirmation in a biopsy or
surgical specimen. The inclusion criterion was
availability of both fast STIR and T1-weighted fat-
suppressed contrast-enhanced images. Six patients
were excluded because of severe motion artifact,
and 12 patients were excluded because of inho-
mogeneous fat suppression on T1-weighted fat-
suppressed contrast-enhanced images. The study
group included 67 patients (39 men and boys, 28
women and girls; mean age, 55.1 years; range,
1–91 years). The histologic types and locations of
the soft-tissue tumors are shown in Table 1.
MR images were obtained with a 1.5-T su-
perconducting MRI system (Signa Horizon, GE
Healthcare). T1-weighted spin-echo (TR, 400–
600 milliseconds; effective TE, 10–12 millisec-
onds) and T2-weighted fast spin-echo (TR/TE,
4,000–4,500/96) imaging was performed in two
planes. A fast spin-echo STIR sequence (4,000–
6,000/60–100; inversion time, 150 milliseconds;
echo-train length, 10; field of view, 140–400
mm2; matrix size, 256–384 × 224–256; section
thickness, 3–5 mm; and intersection gap, 0.3–1.0
mm) also was performed in two planes (transax-
ial and coronal or sagittal). The number of sig-
nals acquired was four, and the acquisition time
for the fast STIR sequence varied between 5 min-
utes 12 seconds and 5 minutes 42 seconds. All pa-
tients agreed to receive an additional IV injection
of contrast material. A T1-weighted spin-echo se-
quence (400–600/10–12) with fat suppression
was performed in two planes (transaxial and cor-
onal or sagittal) after administration of gadopen-
tetate dimeglumine (0.1 mmol/kg body weight IV,
Magnevist, Bayer Schering Pharma). The section
thickness was 4–5 mm; intersection gap, 0.4–1.0
mm; field of view, 140–400 mm; image matrix
size, 256–384 × 224–256, identical to the param-
eters used for the fast STIR sequence. The number
Myxoma Thigh 2
Malignant (n = 33)
Myxoid liposarcoma Thigh 1
Well-differentiated liposarcoma Thigh 2
Pleomorphic liposarcoma Buttocks, forearm 3
Epithelioid sarcoma Axilla 1
Malignant fibrous histiocytoma Groin, thigh, lower thigh, forearm 12
Malignant peripheral nerve sheath tumor Pelvic wall, buttocks, forearm 2
Malignant lymphoma Groin 1
Extraskeletal osteosarcoma Knee 1
Malignant rhabdoid tumor Thigh 1
Malignant melanoma Upper arm, foot 2
Multiple myeloma Sternoclavicular joint 1
Myxoid chondrosarcoma Axilla, groin, shoulder, lower thigh 4
Rhabdomyosarcoma Lower thigh 1
Synovial sarcoma Thigh 1
of signals acquired was two, and the acquisition lesion was 95.3 cm 2 (range, 32.0–303.5 cm 2). The
time for the fat-suppressed T1-weighted sequence average signal intensity of the region was mea-
varied between 3 minutes 30 seconds and 5 min- sured on both fast STIR and T1-weighted fat-sup-
utes 12 seconds. pressed contrast-enhanced images. The signal in-
Semiquantitative imaging analysis was per- tensity of normal muscle, bone marrow, fat, and
formed. The contrast-to-noise ratios (CNRs) of air near the tumors was measured. All measure-
the soft-tissue tumors were measured. One slice ments were performed twice, and the mean values
position showing a prominent tumorous compo- were adopted.
nent was chosen, and a region of abnormal signals All measurements were performed manually by
in the tumor was identified for analysis. A region one musculoskeletal radiologist (18 years of experi-
of interest was placed within the zones of abnor- ence) using a standard MRI console (Image VINS
mal signal intensity of soft-tissue tumors, the near- Pro, Yokogawa Electric Corporation). The CNR of
est normal tissue (muscle, bone marrow, and fat), a tumor was calculated as the mean signal intensity
and air near the lesion. The mean size of each re- of the lesion minus the mean signal intensity of the
gion of interest within the zones of abnormal sig- normal muscle, bone marrow, and subcutaneous fat
nal intensity of soft-tissue tumors was 304.1 cm 2 near the lesion divided by the SD of air near the le-
(range, 51.2–1,343.2 cm 2), within normal mus- sion. The signal-to-noise (SNR) of the tumors was
cle was 143.0 cm 2 (range, 28.1–493.2 cm 2), with- calculated as the mean signal intensity of the lesion
in normal bone marrow was 113.7 cm2 (range, divided by the SD of air near the lesion.
17.9–379.6 cm 2), within normal fat was 76.9 cm2 An additional qualitative analysis was performed
(range, 28.1–493.2 cm 2), and within air near the by two musculoskeletal radiologists (18 and 7

1608 AJR:193, December 2009


TABLE 2: Contrast-to-Noise Ratio Comparison of Fast STIR and T1-Weighted Fat-Suppressed Contrast-Enhanced
Images of Soft-Tissue Tumors
Soft-Tissue Tumor Fast STIR
Muscle
Benign tumors 64.50 ± 56.90
Downloaded from www.ajronline.org by 36.80.244.245 on 06/14/18 from IP address 36.80.244.245. Copyright ARRS. For personal use only; all rights reserved
Malignant tumors 72.43 ± 48.47
All tumors 68.51 ± 51.99
Bone marrow
Benign tumors 68.10 ± 53.30
Malignant tumors 68.13 ± 47.14
All tumors 68.11 ± 50.13
Fat
Benign tumors 69.93 ± 46.74
Malignant tumors 68.13 ± 47.14
All tumors 63.22 ± 47.89
Note—Values are mean ± SD.
years of experience) blinded to the clinical data.
They evaluated 59 of the soft-tissue tumors on
coronal images, five tumors on axial images, and
three tumors on sagittal images. Both observers
separately reviewed the fast STIR images and 2
weeks later separately reviewed the T1-weight-
ed fat-suppressed contrast-enhanced images. The
images were presented randomly to each of the
readers at each session. Subjective image contrast,
which differentiates soft-tissue tumors and normal
tissue (muscle, bone marrow, and fat) near the tu-
mors, was graded (1, poor; 2, acceptable; 3, good;
4, excellent). The observers then classified the
soft-tissue tumors as benign or malignant using a
5-point scale to assign a confidence level (1, defi-
nitely benign; 2, probably benign; 3, equivocal; 4,
probably malignant; 5, definitely malignant).
The confidence level ratings of the images were
used to calculate sensitivity, specificity, and accu-
racy for each observer in the diagnosis of malig-
nancy using MR images. Ratings of 1 and 2 indi-
cated a benign tumor; 4 or 5, a malignant tumor;
and 3, an incorrect reading. In addition, MRI of
seven tumors (two myxomas, one myxoid lipo­
sarcoma, four myxoid chondrosarcomas) showed
a cystic mass. Sensitivity, specificity, and accu-
racy for each observer in differentiating benign
from malignant tumors were calculated for these
seven cystic masses.
The paired Student’s t test was used to compare
the CNR and SNR of soft-tissue tumors on fast
STIR images with those on T1-weighted fat-sup-
pressed contrast-enhanced images; the observer
ratings of the subjective image contrast of the tu-
mors on the 4-point scale on both types of images;
the observer ratings of confidence level regarding
benign or malignant; and the sensitivity, specific-
AJR:193, December 2009 1609
MRI of Soft-Tissue Tumors
T1-Weighted Fat-Suppressed
Contrast-Enhanced
33.38 ± 33.02
38.21 ± 29.30
35.82 ± 30.73
51.90 ± 39.96
47.52 ± 32.45
49.92 ± 36.47
45.13 ± 36.17 –11.94 to 35.60
49.01 ± 33.50
47.00 ± 34.04
ity, and accuracy of fast STIR imaging with those
of T1-weighted fat-suppressed contrast-enhanced
imaging. Software (Statcel version 7.0, OMS) was
used to calculate p; p < 0.05 was used to indicate a
statistically significant difference.
To assess interobserver and intraobserver vari-
ability in assignment of a confidence level to le-
sion status, the weighted kappa value [20] was
calculated. The level of agreement was defined
as follows: a kappa value less than 0 indicated
no agreement; 0.00–0.40, poor agreement; 0.41–
0.75, good agreement; 0.76–1.00, excellent agree-
ment. Kappa values were calculated with statisti-
cal software (Excel Statistics 2008 for Mirosoft
Windows, SSRI).
Receiver operating characteristic curves were
calculated to compare the fast STIR reading
with the T1-weighted fat-suppressed contrast-en-
hanced readings. True-positive cases were defined
as malignant tumors correctly designated as such.
False-positive cases were defined as benign tu-
mors incorrectly designated as such. Diagnostic
capability was determined by calculation of the
area under each reader-specific receiver operating
characteristics curve (A z). The results were ex-
pressed as mean ± SD. Calculation of A z was per-
formed with software (Rockit beta version 0.9.1,
C. E. Metz and B. A. Herman, University of Chi-
cago). The A z values for fast STIR readings were
compared with those for T1-weighted fat-sup-
pressed contrast-enhanced readings by use of one-
factor analysis of variance for both observers.
Results
CNR comparisons for the fast STIR ver-
sus T1-weighted fat-suppressed contrast-en-
hanced images of soft-tissue tumors (benign,
95% CI p
8.49 to 53.75 < 0.01
14.81 to 53.62 < 0.01
18.53 to 46.84 < 0.0001
–6.53 to 38.91 0.16
3.33 to 37.90 < 0.05
3.96 to 32.43 < 0.05
0.32
3.77 to 38.07 < 0.05
2.32 to 30.13 < 0.05
malignant, and all tumors) are shown in Ta-
ble 2. In comparison with the CNR of muscle,
the mean CNRs of benign, malignant, and
all tumors were significantly higher on fast
STIR images than on T1-weighted fat-sup-
pressed contrast-enhanced images (p < 0.01)
(Figs. 1 and 2). In comparison with bone
marrow, the mean CNRs of malignant and
all tumors were significantly higher on fast
STIR than on T1-weighted fat-suppressed
contrast-enhanced images (p < 0.05). There
was no significant difference, however, be-
tween the mean CNR of benign tumors on
the fast STIR images and that on T1-weight-
ed fat-suppressed contrast-enhanced images
(p = 0.16) (Fig. 3). In comparison with the
CNR of fat, the mean CNRs of malignant
and all tumors were significantly higher on
fast STIR images than on T1-weighted fat-
suppressed contrast-enhanced images (p <
0.05). However, there was no significant dif-
ference between the mean CNR of benign tu-
mors on fast STIR images and that on T1-
weighted fat-suppressed contrast-enhanced
images (p = 0.32).
There were no significant differences be-
tween the mean SNRs of benign, malignant,
and all tumors in the muscle, bone marrow,
and fat on fast STIR images and those on
T1-weighted fat-suppressed contrast-en-
hanced images.
The observer ratings of subjective image
contrast of the tumors in comparison with
normal muscle, bone marrow, and fat on fast
STIR and T1-weighted fat-suppressed con-
trast-enhanced images are shown in Table 3.
In the comparison between tumor and muscle,
 Tokuda et al.

both observers rated the fast STIR images sig-

nificantly higher than the T1-weighted fat-sup-
pressed contrast-enhanced images of benign
tumors (observer 1, p < 0.0001; observer 2, p <
0.05), malignant tumors (observer 1, p < 0.001;
observer 2, p < 0.05), and all tumors (observer
Downloaded from www.ajronline.org by 36.80.244.245 on 06/14/18 from IP address 36.80.244.245. Copyright ARRS. For personal use only; all rights reserved

1, p < 0.0001; observer 2, p < 0.01).

In the comparison between tumor and bone
marrow, observer 1 rated the fast STIR images
significantly higher than the T1-weighted fat-
suppressed contrast-enhanced images of be-
nign tumors (p < 0.001) and all tumors (p <
0.01). There were no significant differences,
however, between the mean ratings of fast
STIR images and those of T1-weighted-fat-
suppressed contrast-enhanced images of ma-
lignant tumors. For observer 2, there were no
significant differences between the mean rat- A B
ings of the fast STIR images and T1-weight- Fig. 1—55-year-old woman with malignant peripheral nerve sheath tumor of right buttock.
A, Coronal fast STIR MR image (TR/TE, 4,000/120; inversion time, 150 milliseconds; echo-train length, 10)
ed fat-suppressed contrast-enhanced images shows well-defined mass of heterogeneously high signal intensity. Markedly hyperintense areas correspond
of benign, malignant, and all tumors. to cystic necrosis (short arrow). Tumor invasion of surrounding area of high signal intensity (long arrow) was
In the comparison of fat with benign and pathologically proven.
B, Coronal T1-weighted fat-suppressed contrast-enhanced MR image (400/10) shows strongly heterogeneous
all tumors, observer 1 gave the fast STIR im-
enhancement. Multiple unenhanced compartments correspond to cystic necrosis (arrows).
ages significantly higher mean ratings than
the T1-weighted fat-suppressed contrast-en-
hanced images (p < 0.05). In the compari-
son of fat with malignant tumors by observ-
er 1, no significant difference (p = 0.54) was
found between the mean rating of the fast
STIR images and that of the T1-weighted
fat-suppressed contrast-enhanced images.
For observer 2, there were no significant dif-
ferences between the mean ratings of the fast
STIR images of benign, malignant, and all
tumors and those of the T1-weighted fat-sup-
pressed contrast-enhanced images.
The sensitivity, specificity, and accura-
cy for differentiation of benign from malig-
nant tumors on fast STIR and T1-weighted
fat-suppressed contrast-enhanced images by
both observers are shown in Table 4. The sen-
sitivity, specificity, and accuracy of observer
1 reading the fast STIR images were 93.9%,
A B
73.5%, and 80.6% and reading the T1-weight-
ed fat-suppressed contrast-enhanced images Fig. 2—24-year-old man with schwannoma of left thigh.
A, Coronal fast STIR MR image (TR/TE, 4,000/120; inversion time, 150 milliseconds; echo-train length, 10)
were 97.0%, 58.8%, and 77.6%. The sensi- shows well-defined mass of marked heterogeneously high signal intensity with intermediate hyperintense
tivity, specificity, and accuracy of observer 2 area. Marked hyperintense area (long arrow) corresponds to Antoni type B pattern. Partial intermediate
reading the STIR images were 81.8%, 52.9%, hyperintense area (short arrow) corresponds to Antoni type A pattern.
B, Coronal T1-weighted fat-suppressed contrast-enhanced MR image (400/10) shows weakly heterogeneous
and 67.2%, and reading the T1-weighted fat-
enhancement with ill-defined margin. Arrow indicates strongly enhanced area corresponding to Antoni type A
suppressed contrast-enhanced images were pattern.
84.8%, 47.1%, and 65.7%. For both observ-
ers, there were no significant differences be- of observer 1 reading fast STIR images were and 57.1% (4/7) and reading T1-weighted fat-
tween the sensitivity, specificity, and accura- 60.0% (3/5), 0.0% (0/2), and 42.9% (3/7) and suppressed contrast-enhanced images were
cy of reading fast STIR images and those of for reading T1-weighted fat-suppressed con- 100% (5/5), 0% (0/2), and 71.4% (5/7).
reading T1-weighted fat-suppressed contrast- trast-enhanced images were 100% (5/5), 0% The weighted kappa values for the rat-
enhanced images. (0/2), and 71.4% (5/7). The sensitivity, speci- ings of the subjective image contrast in eval-
In the analysis of the seven cystic mass- ficity, and accuracy of observer 2 reading fast uation of soft-tissue tumors in muscle were
es, the sensitivity, specificity, and accuracy STIR images were 80.0% (4/5), 0.0% (0/2), 0.749 (95% CI: 0.599–0.899) for fast STIR

1610 AJR:193, December 2009

 MRI of Soft-Tissue Tumors

malignant tumors by observer 1 was 0.836

(95% CI: 0.744–0.928) and for that by observ-
er 2 was 0.837 (95% CI: 0.744–0.930). These
findings suggested excellent intraobserver
agreement for both observers for differentia-
tion of benign and malignant tumors.
Downloaded from www.ajronline.org by 36.80.244.245 on 06/14/18 from IP address 36.80.244.245. Copyright ARRS. For personal use only; all rights reserved

The mean Az for observer 1 was 0.925 ±

0.036 (95% CI: 0.825–0.974) for the fast
STIR images and 0.906 ± 0.033 (95% CI:
0.843–0.981) for the T1-weighted fat-sup-
pressed contrast-enhanced images (Fig. 4A).
The Az values for observer 2 were 0.919 ±
0.035 (95% CI: 0.828–0.968) for the fast
STIR images and 0.928 ± 0.033 (95% CI:
0.838–0.974) for the T1-weighted fat-sup-
pressed contrast-enhanced images (Fig. 4B).
Neither observer had significant differences
A B between the Az values for fast STIR and those
Fig. 3—53-year-old woman with solitary fibrous tumor of left inguinal region. for T1-weighted fat-suppressed contrast-en-
A, Coronal fast STIR MR image (TR/TE, 4,000/120; inversion time, 150 milliseconds; echo-train length, 10) shows hanced images (observer 1, p = 0.453; ob-
heterogeneous mass of low signal intensity with ill-defined margin (arrow). server 2, p = 0.239).
B, Coronal T1-weighted fat-suppressed contrast-enhanced MR image (400/10) shows strongly heterogeneous
enhancement with well-defined margin (arrow).
Discussion
images and 0.684 (95% CI: 0.546–0.821) for ratings of subjective image contrast. The In the STIR technique, an initial 180° in-
T1-weighted fat-suppressed contrast-enhanced weighted kappa values for differentiation verting radiofrequency pulse is followed by
images, in bone marrow were 0.769 (95% CI: of benign and malignant tumors were 0.766 a standard 90–180° spin-echo sequence. The
0.619–0.919) for fast STIR images and 0.682 (95% CI: 0.671–0.861) for the fast STIR im- time allowed to elapse between the inversion
(95% CI: 0.552–0.812) for T1-weighted fat- ages and 0.795 (95% CI: 0.706–0.884) for pulse and the 90° pulse is chosen to approxi-
suppressed contrast-enhanced images, and the T1-weighted fat-suppressed contrast-en- mate the null point of fat, resulting in sup-
in fat were 0.752 (95% CI: 0.575–0.928) for hanced images. These findings suggested ex- pression of the signal intensity of fat. Most
fast STIR images and 0.479 (95% CI: 0.258– cellent interobserver agreement for the dif- often, the inversion time is determined with
0.700) for T1-weighted fat-suppressed con- ferentiation between benign and malignant field-strength-dependent reference values.
trast-enhanced images. These findings sug- tumors on both types of images. The weighted A further important point is that high-field-
gested good interobserver agreement on the kappa value for differentiation of benign and strength systems are required for use of the

TABLE 3: Observer Ratings of STIR and T1-Weighted Fat-Suppressed Contrast-Enhanced Images

of Soft-Tissue Tumors
Observer 1 Observer 2
T1-Weighted T1-Weighted
Fat-Suppressed Fat-Suppressed
Soft-Tissue Tumor Fast STIR Contrast-Enhanced 95% CI p Fast STIR Contrast-Enhanced 95% CI p
Muscle
Benign tumors 3.26 ± 1.14 2.53 ± 0.99 0.33 to 1.14 < 0.0001 3.47 ± 0.99 2.97 ± 1.14 0.05 to 0.95 < 0.05
Malignant tumors 3.52 ± 0.62 3.06 ± 0.90 0.17 to 0.74 < 0.001 3.82 ± 0.39 3.39 ± 0.79 0.08 to 0.77 < 0.05
All tumors 3.39 ± 0.92 2.79 ± 0.98 0.35 to 0.84 < 0.0001 3.64 ± 0.77 3.18 ± 0.10 0.19 to 074 < 0.01
Bone marrow
Benign tumors 3.32 ± 1.15 2.76 ± 1.13 0.15 to 0.97 < 0.001 3.35 ± 1.04 3.21 ± 1.01 −0.30 to 0.59 0.51
Malignant tumors 3.52 ± 0.62 3.24 ± 0.87 −0.05 to 0.60 0.095 3.73 ± 0.52 3.64 ± 0.65 −0.11 to 0.30 0.37
All tumors 3.42 ± 0.92 3.00 ± 1.03 0.16 to 0.68 < 0.01 3.54 ± 0.84 3.42 ± 0.87 −0.12 to 0.36 0.33
Fat
Benign tumors 3.36 ± 0.98 3.07 ± 1.63 0.04 to 0.90 < 0.05 3.50 ± 0.93 3.26 ± 1.05 −0.15 to 0.63 0.23
Malignant tumors 3.45 ± 0.83 3.36 ± 0.96 −0.21 to 0.39 0.54 3.64 ± 0.74 3.55 ± 0.83 −0.17 to 0.35 0.48
All tumors 3.36 ± 0.98 3.07 ± 1.06 0.02 to 0.54 < 0.05 3.56 ± 0.84 3.40 ± 0.95 −0.07 to 0.40 0.16
Note—Values are mean ± SD.

AJR:193, December 2009 1611

 Tokuda et al.

TABLE 4: Differentiation of Benign From Malignant Tumors on Fast STIR and and T1-weighted fat-suppressed contrast-en-
T1-Weighted Fat-Suppressed Contrast-Enhanced MR Images hanced images were observed for either ob-
Sequence Observer 1 Observer 2 Mean server in regard to sensitivity, specificity, and
accuracy for differentiation of benign from
Fast STIR
malignant bone tumors. In our study, the
Sensitivity 93.9 (31/33) 81.8 (27/33) 87.9 sensitivity, specificity, and accuracy for dif-
Downloaded from www.ajronline.org by 36.80.244.245 on 06/14/18 from IP address 36.80.244.245. Copyright ARRS. For personal use only; all rights reserved

Specificity 73.5 (25/34) 52.9 (18/34) 63.2 ferentiation of benign from malignant soft-
Accuracy 80.6 (56/67) 67.2 (45/67) 73.9 tissue tumors on fast STIR and T1-weight-
ed fat-suppressed contrast-enhanced images
T1-weighted fat-suppressed contrast-enhanced
were comparable for the two observers. In
Sensitivity 97.0 (32/33) 84.8 (28/33) 90.9 addition, the two observers had no signifi-
Specificity 58.8 (20/34) 47.1 (16/34) 53.0 cant differences in A z values for fast STIR
Accuracy 77.6 (52/67) 65.7 (44/67) 71.7 and T1-weighted fat-suppressed contrast-en-
hanced images. These findings suggest that
Note—Data are percentages. Numbers in parentheses are raw data.
the ability to differentiate benign from ma-
lignant soft-tissue tumors on fast STIR im-
1.0 1.0 ages may be almost the same as the ability
to differentiate them on T1-weighted fat-sup-
0.8 T1WI-FSCE 0.8 pressed contrast-enhanced images but that
STIR
True-Positive Fraction

True-Positive Fraction

STIR
0.6
0.4
0.2
0 0
0 0.2 0.4 0.6 0.8
False-Positive Fraction
Fig. 4—Receiver operating characteristic (ROC) curves of observer confidence in differentiating benign from
malignant tumors on STIR and T1-weighted fat-suppressed contrast-enhanced (T1WI-FSCE) images.
A, Graph shows ROC curve for observer 1.
B, Graph shows ROC curve for observer 2.
frequency-selective fat-suppressed sequence,
whereas the fast STIR sequence can be per-
formed on low- or high-field-strength MRI
units [21].
Most soft-tissue tumors are visualized as
areas of markedly increased signal intensity
on both fast STIR and T1-weighted fat-sup-
pressed contrast-enhanced images. Conse-
quently, the most important point of compar-
ison between the fast STIR and T1-weighted
fat-suppressed contrast-enhanced sequences
for evaluation of soft-tissue tumors is the uni-
formity of fat suppression. In this study, the
mean CNRs of malignant and all tumors in
comparison with normal muscle, bone mar-
row, and fat were significantly higher on fast
STIR than on T1-weighted fat-suppressed
contrast-enhanced images. There were no
significant differences, however, between the
mean SNRs of benign, malignant, and all tu-
mors on fast STIR images and those on T1-
weighted fat-suppressed contrast-enhanced
images. These findings may have occurred be-
T1WI-FSCE
0.6
0.4
0.2
1.0 0 0.2 0.4 0.6 0.8 1.0
False-Positive Fraction
A B
cause the fast STIR sequence was superior to
the T1-weighted fat-suppressed contrast-en-
hanced sequence in terms of uniformity of fat
suppression. Nakatsu et al. [16] reported that
the fast STIR sequence afforded fat suppres-
sion in the cervical and thoracic regions that
was closer to homogeneous than that obtained
with the fat-suppressed (chemsat) fast spin-
echo sequence. In the cervical and thoracic
regions, magnetic field inhomogeneity due to
the susceptibility effects of complex anatom-
ic relations and air-containing structures often
disturbs uniform fat suppression. In addition,
magnetic field uniformity can be a problem in
MRI of extremities because of the relatively
off-center location of the imaging object and
the proximity of air–tissue boundaries.
The sensitivity, specificity, and accura-
cy for differentiation of benign from malig-
nant bone tumors have been calculated for
fast STIR and T1-weighted fat-suppressed
contrast-enhanced images [17]. In that study,
no significant differences between fast STIR
for depiction of soft-tissue tumors, as op-
posed to differentiating benign from malig-
nant, fast STIR imaging is superior.
In this study, both observers rated the
mean subjective image contrast of soft-tissue
tumors in comparison with muscle on fast
STIR images significantly higher than on T1-
weighted fat-suppressed contrast-enhanced
images. Observer 2 rated the mean subjec-
tive image contrast of soft-tissue tumors in
comparison with bone marrow on fast STIR
images comparable with that on T1-weight-
ed fat-suppressed contrast-enhanced images.
Furthermore, interobserver and intraobserv-
er agreement in ratings of subjective im-
age contrast was good, and this comparable
agreement also suggested that the fast STIR
sequence was equivalent to the T1-weighted
fat-suppressed contrast-enhanced sequence
for evaluation of soft-tissue tumors.
The STIR sequence has several disadvan-
tages. First, the SNR tends to be lower with
the STIR than with a spin-echo sequence
[14]. In our study, however, there were no
significant differences between the mean
SNR of benign, malignant, and all tumors in
the normal muscle, bone marrow, and fat on
the fast STIR images and that on T1-weight-
ed fat-suppressed contrast-enhanced images.
Second, the fast STIR sequence includes rel-
atively long imaging times for acquisition of
a limited number of imaging slices, result-
ing in vulnerability to motion artifacts and
poor SNR [1]. This problem is less severe
in imaging of soft-tissue tumors of the ex-
tremities than imaging of tumors of the chest
and abdomen because the extremities are
less affected by motion-induced noise. Mo-
tion artifact–induced long acquisition times

1612 AJR:193, December 2009


are important problems in the evaluation of
soft-tissue tumors of the chest and abdomen.
Finally, nonlipid tissue can be suppressed
if it has a short T1 similar to that of fat be-
cause fat suppression on fast STIR images is
based strictly on relaxation values [1]. Al-
Downloaded from www.ajronline.org by 36.80.244.245 on 06/14/18 from IP address 36.80.244.245. Copyright ARRS. For personal use only; all rights reserved
though it has several disadvantages, the fast
STIR sequence clearly is useful for evalua-
tion of soft-tissue tumors. A further impor-
tant point is that high-field-strength systems
are needed for the frequency-selective fat-
suppressed sequence, whereas the fast STIR
sequence can be performed on low- or high-
field-strength units [21].
According to May et al. [22], the routine
use of gadolinium in the initial MRI evalu-
ation of a possible primary musculoskeletal
neoplasm is not justified. In several cases,
however, contrast enhancement is necessary
for evaluation of soft-tissue tumors. First,
it is difficult to evaluate cystic and myxoid
soft-tissue masses without contrast enhance-
ment. The differential diagnosis of a cystic
mass includes ganglion, infection (abscess or
pyomyositis), myxoma, and myxoid sarco-
ma. Ganglion and abscess have regular rim
enhancement, but myxoid soft-tissue sarco-
ma has thick nodular or irregular peripheral
enhancement with or without central nodu-
lar enhancement. Therefore, the contrast en-
hancement pattern is important in the eval-
uation of cystic masses. In the analysis of
seven cystic masses in our study, the sensi-
tivity and accuracy of both observers reading
fast STIR images were lower than in reading
of T1-weighted fat-suppressed contrast-en-
hanced images. Another reason that contrast
enhancement may be helpful in the evalua-
tion of a soft-tissue mass is to help the ra-
diologist plan the biopsy route, especially in
evaluation of a cystic or necrotic mass. The
area of enhancement representing more vi-
able tumor tissue typically has higher diag-
nostic yield than the unenhanced necrotic ar-
eas of the mass [23].
At some institutions, it is possible for the
radiologist to evaluate unenhanced imag-
es of soft-tissue tumors and decide whether
contrast-enhanced sequences should be per-
formed. Some authors advocate use of dy-
namic contrast enhancement in the initial
evaluation and follow-up of soft-tissue tu-
mors, and some advocate use of MRI spec-
troscopy in the evaluation of these tumors
[24–26]. For practical reasons at busy hos-
pitals and imaging centers, it is not possible
to routinely check images before IV contrast
administration. Therefore, at those institu-
AJR:193, December 2009 1613
MRI of Soft-Tissue Tumors
tions, both STIR and T1-weighted fat-sup-
pressed contrast-enhanced sequences should
be included in standard protocols for evalua-
tion of soft-tissue tumors.
This study had limitations. First, a con-
sequence of the retrospective design was
variability in the MRI parameters. Second,
one musculoskeletal radiologist placed a re-
gion of interest within the soft-tissue tumors,
the nearest normal soft tissue (muscle, bone
marrow, and fat), and air near the lesion. To
avoid sampling error, all measurements were
performed twice, and the mean values were
adopted. Third, the patients were not con-
secutively registered, and the sample size
was modest. All of the patients with soft-tis-
sue tumors who underwent MRI fell into the
following three groups: both fast STIR and
the T1-weighted fat-suppressed contrast-en-
hanced images available; both T2-weight-
ed fat-suppressed and T1-weighted fat-sup-
pressed contrast-enhanced images available;
and only fast STIR images available. Al-
though more than 80% of the patients with
soft-tissue tumors who underwent MRI had
both fast STIR and T1-weighted fat-sup-
pressed contrast-enhanced images, selection
bias is another consideration.
This study showed that fast STIR and T1-
weighted fat-suppressed contrast-enhanced
sequences had comparable results for differ-
entiation of benign from malignant tumors.
The fast STIR technique is an excellent meth-
od for evaluation of soft-tissue tumors, and
contrast enhancement is not always needed.
However, prospective studies with the imag-
ing protocols used in this study are needed to
confirm our conclusion.
References
1. Mirowitz SA. Fast scanning and fat-suppression
MR imaging of musculoskeletal disorders. AJR
1993; 161:1147–1157
2. Mitchell DG, Vintski S, Rifkin MD, Burk DL Jr.
Sampling bandwidth and fat suppression: effects
on long TR/TE MR imaging of the abdomen and
pelvis at 1.5T. AJR 1989; 153:419–425
3. Gronemeyer SA, Kauffman WK, Rocha MS,
Steen RG, Fletcher BD. Fat-saturated contrast-
enhanced T1-weighted MRI in evaluation of os-
teosarcoma and Ewing sarcoma. J Magn Reson
Imaging 1997; 7:585–589
4. Schmid MR, Holder J, Vienne P, Binkert CA, Za-
netti M. Bone marrow abnormalities of foot and
ankle: STIR versus T1-weighted contrast-en-
hanced fat-suppressed spin-echo MR imaging.
Radiology 2002; 224:463–469
5. Georgy BA, Hesselink JR. Evaluation of fat-sup-
pression in contrast-enhanced MR of neoplastic
and inflammatory spine disease. Am J Neurora-
diol 1994; 15:409–417
6. Morrison WB, Schweitzer ME, Bock GW, et al.
Diagnosis of osteomyelitis: utility of fat-sup-
pressed contrast-enhanced MR imaging. Radiol-
ogy 1993; 189:251–257
7. Semelka RC, Shoenut JP, Lawrence PH, et al.
Evaluation of adrenal masses with gadolinium en-
hancement and fat-suppressed MR imaging. J
Magn Reson Imaging 1993; 3:337–343
8. Tien RD, Olson EM, Zee CS. Disease of the lum-
bar spine: findings on fat-suppression MR imag-
ing. AJR 1992; 159:95–99
9. Semelka RC, Hricak H, Stevens SK, Finegold R,
Tomei E, Carroll PR. Combined gadolinium-en-
hanced and fat-saturation MR imaging of renal
masses. Radiology 1991; 178:803–809
10. Barakos JA, Dillon WP, Chew WM. Orbit, skull
base, and pharynx: contrast-enhanced fat-sup-
pression MR imaging. Radiology 1991; 179:191–
198
11. Ross MR, Schomer DF, Chappell P, Enzmann DR.
MR imaging of head and neck tumors: compari-
son of T1-weighted fat-suppressed images with
conventional T2-weighted and fast spin-echo T2-
weighted images. AJR 1994; 163:173–178
12. Hirsch JA, Loevner LA, Yousem DM, et al. Gad-
olinium-enhanced fat-suppressed T1-weighted
imaging of the head and neck: comparison of gra-
dient and conventional SE sequences. J Comput
Assist Tomogr 1998; 22:771–776
13. Tamada T, Nagai K, Iizuka M, et al. Comparison
of whole-body MR imaging and bone scintigra-
phy in the detection of bone metastasis from
breast cancer [in Japanese]. Nippon Igaku Hosha-
sen Gakkai Zasshi 2000; 60:249–254
14. Shuman WP, Patten RM, Baron RL, Liddell RM,
Conrad EU, Richardson ML. Comparison of
STIR and spin-echo MR imaging at 1.5 T in 45
suspected extremity tumors: lesion conspicuity
and extent. Radiology 1991; 179:247–252
15. Golfieri R, Baddeley H, Pringle JS, Souhami R.
The role of the STIR sequence in magnetic reso-
nance imaging examination of bone tumors. Br J
Radiol 1990; 63:251–256
16. Nakatsu M, Hatabu H, Itoh H, et al. Comparison
of short inversion time inversion recovery (STIR)
and fat-saturated (chemsat) techniques for back-
ground fat intensity suppression in cervical and
thoracic MR imaging. J Magn Reson Imaging
2000; 11:56–60
17. Tokuda O, Hayashi N, Matsunaga N. MRI of bone
tumors: fast STIR imaging as a substitute for T1-
weighted contrast-enhanced fat-suppressed spin-
echo imaging. J Magn Reson Imaging 2004; 19:
475–481
18. Mirowitz SA, Apicella P, Reinus WR, Hammerman
 Tokuda et al.

AM. MR imaging of bone marrow lesions: rela-
tive conspicuousness on T1-weighted, fat-sup-
pressed T2-weighted, and STIR images. AJR
1994; 162:215–221
19. Jones KM, Unger EC, Granstrom P, Seeger JF,
Camoby RF, Yoshino M. Bone marrow imaging
Downloaded from www.ajronline.org by 36.80.244.245 on 06/14/18 from IP address 36.80.244.245. Copyright ARRS. For personal use only; all rights reserved
21. Tien RD. Fat-suppression MR imaging in neuro-
radiology: techniques and clinical application.
AJR 1992; 158:369–372
22. May DA, Good RB, Smith DK, Parsons TW.
MR imaging of musculoskeletal tumors and tu-
mor mimickers with intravenous gadolinium:
skeletal neoplasms: static and dynamic Gd-DTPA–
enhanced MR imaging. Radiology 1989; 171:
767–773
25. Vanel D, Shapeero LG, Tardivon A, Western A,
Guinebretiere JM. Dynamic contrast-enhanced
MRI with subtraction of aggressive soft tissue

using STIR at 0.5 and 1.5 T. Magn Reson Imaging
1992; 10:169–176
20. Fleiss JL. The measurement of interrater agree-
ment. In: Fleiss JL. Statistical methods for rates
and proportions, 2nd ed. New York, NY: Wiley,
1981: 212–236
experience with 242 patients. Skeletal Radiol
1997; 26: 2–15
23. Gogna A, Peh WC, Munk PL. Image-guided mus-
culoskeletal biopsy. Radiol Clin North Am 2008;
46:455–473
24. Erlemann R, Reiser MF, Peters PE, et al. Musculo-
tumors after resection. Skeletal Radiol 1998; 27:
505–510
26. Wang CK, Li CW, Hsieh TJ, Chien SH, Liu GC,
Tsai KB. Characterization of bone and soft-tissue
tumors with in vivo 1H MR spectroscopy: initial
results. Radiology 2004; 232:599–605

F O R YO U R I N F O R M AT I O N
The AJR has made getting the articles you really want really easy with an online tool, Really Simple
Syndication, available at www.ajronline.org. It’s simple. Click the yellow RSS button located in the
menu on the left of the page. You’ll be on your way to syndicating your AJR content in no time.

1614 AJR:193, December 2009