Eur Radiol (2017) 27:1867–1876

DOI 10.1007/s00330-016-4583-0

UROGENITAL

Endometrial cancer with cervical stromal invasion: diagnostic

accuracy of diffusion-weighted and dynamic contrast enhanced
MR imaging at 3T
Gigin Lin 1,5 & Yu-Ting Huang 1 & Angel Chao 2 & Yu-Chun Lin 1 & Lan-Yan Yang 4 &
Ren-Chin Wu 3 & Hsin-Ying Lu 1,5 & Shu-Hang Ng 1 & Koon-Kwan Ng 1 & Chyong-Huey Lai 2

Received: 30 May 2016 / Revised: 22 August 2016 / Accepted: 26 August 2016 / Published online: 9 September 2016
# European Society of Radiology 2016

Abstract MR imaging— 91.6 %/88 %, 58.3 %/50 %, and

Objectives To compare the diagnostic accuracy of diffusion- 97.2 %/94.4 %. The diagnostic performance of DW
weighted (DW) and dynamic contrast-enhanced (DCE) mag- MR imaging (Reader 1: areas under the receiver operat-
netic resonance (MR) imaging for detecting cervical stromal ing characteristic curve (AUC) = 0.98; Reader 2: AUC =
invasion in endometrial cancer. 0.97) was significantly higher than that of DCE MR
Methods Eighty-three consecutive women with endometrial imaging (p = 0.009 for Reader 2) or T2-weighted MR
cancer underwent preoperative evaluation in a 3-T unit, in- imaging (Reader 1: p = 0.006; Reader 2: p = 0.013).
cluding T2-weighted, DW (b = 0 and 1000 s/mm2), and Patients with cervical stromal invasion showed a signif-
DCE MR imaging. Two radiologists independently assessed icantly greater canal width (p < 0.0001) and myometrial
presence of cervical stromal invasion, with histopathological invasion extent (p = 0.006).
reference as gold standard. Conclusions DW MR imaging has superior diagnostic perfor-
Results For assessing cervical stromal invasion, the diagnos- mance compared with DCE MR imaging in the detection of
tic accuracy, sensitivity, and specificity, respectively for cervical stromal invasion.
Reader 1/Reader 2, were as follows: DW MR imaging— Key Points
95.2 %/91.6 %, 91.7 %/100 %, and 95.8 %/90.1 %; DCE • DWI demonstrates a higher accuracy than DCE in detecting
cervical stromal invasion.
• Tumour ADC values are similar between patients without or
* Gigin Lin with cervical invasion.
[email protected] • Canal widening causes false-negativity on DCE and T2W
but not on DWI.
1
Department of Medical Imaging and Intervention, Institute for
Radiological Research, Chang Gung Memorial Hospital at Linkou Keywords Cervical stromal invasion . Diffusion weighted
and Chang Gung University, 5 Fuhsing St., Guishan, Taoyuan, imaging . Endometrial cancer . Magnetic resonance imaging .
Taiwan 33382 Myometrial invasion
2
Department of Obstetrics and Gynecology and Gynecologic Cancer
Research Center, Chang Gung Memorial Hospital at Linkou and
Chang Gung University, 5 Fuhsing St., Guishan, Taoyuan, Taiwan
33382
Abbreviations
3
ADC Apparent diffusion coefficient
Department of Pathology, Chang Gung Memorial Hospital at Linkou
and Chang Gung University, 5 Fuhsing St., Guishan, Taoyuan,
AUC Areas under the receiver operating
Taiwan 33382 characteristics curve
4
Clinical Trial Center, Chang Gung Memorial Hospital at Linkou and
CA-125 Cancer antigen 125
Chang Gung University, 5 Fuhsing St., Guishan, Taoyuan, Taiwan DCE Dynamic contrast-enhanced
33382 DW Diffusion-weighted
5
Clinical Phenome Center, Chang Gung Memorial Hospital at FIGO International Federation of Gynecology
Linkou, 5 Fuhsing St., Guishan, Taoyuan, Taiwan 33382 and Obstetrics
1868 Eur Radiol (2017) 27:1867–1876

FOV Field of view membranes [25]. However, DW MR imaging has not been
MR Magnetic resonance included in the standard staging guideline for endometrial
NCCN National Comprehensive Cancer Network cancer [26], indicating a need to assess the next frontier in
ROI Region of interest the evaluation of cervical stromal invasion. Preliminary re-
TR Repetition time ports have attempted to compare DW with contrast-
TE Echo time enhanced MR imaging, but direct comparison with DCE
MR imaging is sparse [27–29]. Although the added value of
DWI in the endometrial cancer staging has been extensively
Introduction evaluated, whether DW MR imaging is superior to DCE MR
imaging in the diagnosis of cervical stromal invasion remains
Endometrial cancer is the most common malignancy of the unknown.
female reproductive system in the United States [1]. Stromal In this study we aim to evaluate the diagnostic accuracy of
invasion of the cervix is associated with a higher risk of lymph DW MR imaging with reference to the ADC map in compar-
nodal spread [2, 3] and is a poor prognostic factor for both ison with DCE MR imaging for detecting cervical stromal
progression-free survival and overall survival [4–7]. invasion in endometrial cancer patients.
Therefore, the updated 2009 International Federation of
Gynaecology and Obstetrics (FIGO) staging system specified
cervical stromal invasion as stage II disease [8]. According to Materials and methods
the National Comprehensive Cancer Network guidelines [9],
radical hysterectomy along with bilateral salpingo-oophorec- Patients
tomy, peritoneal lavage, and lymph node dissection if indicat-
ed, improves local control and survival compared with total Our institutional review board approved the protocol of this
hysterectomy alone, for endometrial cancer patients with cer- diagnostic accuracy study and informed consent was obtained
vical stromal invasion [10]. The presence of cervical stromal from all patients. The study was conducted in a tertiary referral
invasion is an indication for oncologists to administer adju- centre with a dedicated gynaecologic oncology interdisciplin-
vant brachytherapy [2]. In addition, the clinical exclusion of ary team. Inclusion criteria included (1) histologically proven
cervical stromal invasion is critical for nulliparous patients and untreated endometrial carcinoma for which operations
younger than 40 years with pathologically proven type 1 can- were scheduled and (2) an age of 18 years or older.
cer and normal serum levels of cancer antigen 125 (CA-125), Exclusion criteria included (1) MR contraindications (cardiac
because fertility-preserving treatment might be an option [11]. pacemaker, insulin pump, cochlear implant, and metal shrap-
Magnetic resonance (MR) imaging is the imaging modality nel); (2) the presence of pelvic or hip metal prostheses; (3)
of choice in evaluating the uterine cervix and is considered impaired renal function with estimated glomerular filtration
superior to computed tomography [12, 13] and transvaginal rates less than 60 mL/min/1.73 m2; and (4) an inability to
ultrasound [14–16]. MR imaging is also more accurate than provide informed consent. We recorded relevant clinical in-
endocervical curettage [17] and hysteroscopy [15] in the pre- formation, including age, menopausal status, tumour stage,
diction of stromal invasion. According to a recent meta-anal- and types of treatment. Data collection was planned before
ysis, the pooled specificity of MR imaging in the prediction of the MR imaging acquisition and a standard pathological ex-
cervical stromal invasion has reached 95 %, but the sensitivity amination was performed. The study cohort was previously
is only 57 % [18]. Diagnostic accuracy also depends on the reported in an investigation of the influence of menopausal
imaging technique and quality [19]. Stromal invasion of the status on the diagnostic accuracy of MR imaging for the eval-
cervix is typically the most clearly depicted on sagittal and uation of myometrial invasion [30]. In the current study, we
axial oblique T2-weighted images as an area of hypointense examined 83 patients who had undergone hysterectomy and
cervical stroma disrupted by an intermediately intense or hy- compared the accuracy of DW and DCE MR imaging in the
perintense tumour [20, 21]. Dynamic contrast-enhanced diagnosis of cervical stromal invasion in endometrial cancer.
(DCE) MR imaging is accepted as the state-of-the-art standard
for tumour delineation [18, 22]. However, diagnostic difficul- MR methods
ty is commonly encountered when endometrial tumours enter
the endocervical canal and widen the internal os, obliterating MR studies were conducted using a 3-T MR imaging unit
the interface between tumour and the cervical stroma [23]. (Tim Trio, Siemens, Erlangen, Germany) before the patients
Diffusion-weighted (DW) MR imaging has emerged as a use- were scheduled for operations. For all examinations, the pa-
ful tool in evaluating the myometrial invasion depth of endo- tients were placed in the supine position with the bladder
metrial cancer [24, 25] on the basis of the contrast of water emptied before examination. The lower nine elements of the
mobility, tissue cellularity, and the integrity of the cellular integrated spine coil and the lower six elements of the body-
Eur Radiol (2017) 27:1867–1876 1869

phased array coil were used to cover the entire pelvis. In each Imaging analysis
examination, T2-weighted, DW, and DCE MR images of the
pelvis were acquired with sequences obtained with identical Two radiologists with 11 (G.L.) and 8 (Y.T.H.) years’ experi-
slice thicknesses and gaps in the sagittal and axial oblique ence in gynaecological radiology and blinded to clinical and
planes perpendicular to the cervical canal. High-resolution histological information interpreted MR images independent-
T2-weighted imaging was performed with fast spin-echo se- ly on offline Mac OS workstations (MacBook Pro, Apple,
quences (repetition time [TR] ms/echo time [TE] ms, 5630/87; Cupertino, CA, USA). T2-weighted and high-b-value DW
average, 3; matrix, 256 × 320; field of view (FOV), 20 cm) MR images (b = 1000 s/mm2) were fused using dedicated im-
[24]. DW MR imaging was performed using a single-shot age processing software (OsiriX v5.6, Los Angeles, CA,
echo-planar technique with fat suppression (TR/TE, USA) [24]. Each reader interpreted the presence of cervical
3300/79; average, 4; section thickness, 4 mm; gap, 1 mm; stromal invasion on T2-weighted images, DW MR images,
matrix, 128 × 128; FOV, 20 cm). Apparent diffusion coeffi- and DCE MR images independently while blinded to other
cient (ADC) maps were generated from isotropic DW images images. On T2-weighted images, cervical stromal invasion
with b values of 0 and 1000 s/mm2 by calculating the slope of was defined as hypointense cervical stroma disrupted by an
the logarithmic decay curve for signal intensity against the b intermediately intense or hyperintense tumour [20].
value (Syngo, Siemens, Erlangen, Germany). DCE MR imag- Preservation of the normal hypointense cervical stroma inter-
ing was performed at 0, 30, 60, 90, and 120 seconds in the face, regardless of tumour extension into the endocervical
sagittal plane (80/3; average, 1; matrix, 256 × 320; FOV, canal or widening of the internal os, was regarded as no stro-
20 cm) and 180 seconds in the sagittal and axial oblique mal invasion. On the DCE MR images, cervical stromal inva-
planes (567/10; flip angle, 150°; average, 2; matrix, sion was defined as interruption of the enhancement of the
256 × 320; FOV, 20 cm) with an intravenous bolus injection normal cervical epithelium [22, 31]. We planned not to eval-
of 0.1 mmol/kg body weight of the contrast medium uate DCE images in association with T2-weighted images, in
(gadopentetate dimeglumine, Magnevist, Schering, Berlin, order to determine whether intravenous contrast medium ad-
Germany) followed by a 20-mL saline flush with an injection ministration is necessary in a MR exam. On DW MR images,
rate of 3 mL/s [12]. This study was performed during minimal cervical stromal invasion was defined as the presence of
breathing. No premedication was administered. higher signal intensity on high-b-value DW MR images and

Fig. 1 Flow diagram of the study cohort

1870 Eur Radiol (2017) 27:1867–1876

lower signal intensity on ADC maps, compared with the ad- invasion ratio, tumour size, and mean ADC value of endome-
jacent normal cervical parenchyma [32]. The deepest point to trial tumours between groups with or without cervical stromal
which a tumour extended into the cervical stroma was mea- invasion.
sured in either the axial oblique or sagittal plane by using T2-
weighted images providing anatomical reference details.
Regions of interest (ROI) were drawn on a hyperintense tu- Results
mour area on high-b-value DW MR imaging, on the slice with
maximal tumour area, and applied on ADC maps to obtain Patient characteristics
mean ADC values of tumour. ADC values were averaged
between the two readers at the final analysis stage. Cervical From July 2010 to April 2012, a consecutive cohort of
stroma findings were interpreted visually by using a 5-point 107 patients was enrolled, and a total of 83 patients were
grading system (scores 0-4): score 0, no lesion; score 1, defi-
nite benign lesion; score 2, possibly benign lesion; score 3, Table 1 Patient demographics and histopathology data
superficial stromal invasion; score 4, deep stromal invasion. A
score less than or equal to 2 was considered negative for cer- Cervical stromal invasion Present Absent
vical stromal invasion, whereas a score of 3 or higher was n, 12 71
considered positive. Each reader recorded potential confound- Age (y), median (range) 56 (36-74) 56 (30-87)
ing factors pertinent to accurately determining the depth of Menopausal status
cervical stromal invasion, including tumours located at the Premenopausal 6 34
lower half of the uterine segment and cervical canal widening
Postmenopausal 6 37
at the internal os, as indicated by the diameter being more than
Surgery
8 mm at its widest point [33].
Total abdominal hysterectomy 12 68
Laparoscopic hysterectomy 0 3
Histopathology
Overall FIGO
IA 0 49
All of the enrolled patients underwent standard operations
IB 0 9
consistent with current FIGO surgicopathological staging.
II 6 0
Cervical stromal invasion was diagnosed through a consensus
IIIA 0 2
of two pathologists specialising in gynaecological oncology
IIIC1 2 5
without reference to the MR images. When tumours were
IIIC2 3 6
found in both the cervix and endometrium, additional immu-
IV 1 0
nohistochemistry studies were performed to confirm the diag-
Myometrial invasion
nosis of endometrial cancer rather than cervical cancer [34].
The depth of myometrial invasion, lymphovascular invasion, None 2 35
and histopathological types were recorded. Superficial 5 21
Deep 5 15
Statistical analysis Cervical stromal invasion 12 0
Lymphovascular invasion 7 12
Data were analysed using MedCalc for Windows, Version Histopathology
9.2.0.0 (MedCalc Software, Mariakerke, Belgium). Reader Endometrioid 9 66
agreement regarding the invasion depth was analysed using Well differentiated 0 2
weighted kappa statistics (0.00 ≤ k < 0.40 indicated poor Moderately differentiated 2 34
agreement; 0.40 ≤ k ≤ 0.70 indicated fair agreement; k > 0.70 Poorly differentiated 7 30
indicated excellent agreement) [35]. The overall sensitivity, Clear cell 0 2
specificity, and diagnostic accuracy of DW and DCE MR Serous papillary 3 1
imaging were determined on the basis of the histopathological Undifferentiated 0 2
presence of cervical stromal invasion and were represented Tumour type*
with 95 % confidence intervals (CIs) [36]. Areas under the Type 1 cancer 2 36
receiver operating characteristic curve (AUCs) were calculat- Type 2 cancer 10 35
ed to compare diagnostic performance in each group. The
Note—FIGO, International Federation of Gynecology and Obstetrics. *
McNemar test was used to compare the accuracy, sensitivity,
Type 1 cancer: well or moderately differentiated endometrioid adenocar-
and specificity between groups. The Mann-Whitney U test cinoma; type 2 cancer: poorly differentiated endometrioid, clear cell,
was used to compare the age, canal width, myometrial serous papillary or undifferentiated carcinoma
Eur Radiol (2017) 27:1867–1876 1871

Fig. 2 Endometrial cancer in a 43-year-old woman (endometrioid type; intravenous gadolinium contrast administration at 30, 60, 90, and 120 s.
Grade 3). Sagittal (a) T2-weighted MR image (5630/87), (b) DW MR This lesion invaded posterior cervical stroma (arrow) and was correctly
image (3300/79, b = 1000 s/mm2), (c) fused T2-weighted and DW MR diagnosed on all the images
images, (d) ADC map, and (e-h) dynamic T1-weighted image (80/3) after

found to be eligible for final analysis with ages ranging were diagnosed. The 12 patients with cervical stromal
from 30 to 87 years (median, 56 years). A flow diagram invasion showed a significantly greater canal width and
of the study cohort is depicted in Fig. 1. The interval myometrial invasion extent than those of patients without
between the MR examination and surgery ranged from 0 cervical stromal invasion (mean ± standard deviation,
to 30 days (median, 11 days). Table 1 lists the clinical and 13.08 ± 9.45 mm vs 3.07 ± 5.73 mm; p < 0.0001, 50.67
demographic characteristics of the study population. No ± 41.03 % vs 21.61 ± 31.51 %; p = 0.006, respectively).
cases of minimal deviated adenocarcinoma or cervicitis No significant differences were found between the groups

Fig. 3 Endometrial cancer in a 51-year-old woman (endometrioid type; intravenous gadolinium contrast administration at 30, 60, 90, and 120 s.
Grade 3). Sagittal (a) T2-weighted MR image (5630/87), (b) DW MR This lesion invaded anterior cervical stroma (arrow) and was correctly
image (3300/79, b = 1000 s/mm2), (c) fused T2-weighted and DW MR diagnosed on DW MR imaging (b-d) but was not clearly seen on T2-
images, (d) ADC map, and (e-h) dynamic T1-weighted image (80/3) after weighted (a) and DCE MR imaging (e-h)
1872 Eur Radiol (2017) 27:1867–1876

Fig. 4 Endometrial cancer in a 70-year-old woman (endometrioid type; This lesion (arrow) invaded focally to the endocervical canal and
Grade 2). Sagittal (a) T2-weighted MR image (5630/87), (b) DW MR widened the internal os. It was correctly diagnosed on DW MR
image (3300/79, b = 1000 s/mm2), (c) fused T2-weighted and DW MR imaging (b-d) but was not clearly seen on T2-weighted (a) and DCE
images, (d) ADC map, and (e-h) dynamic T1-weighted image (80/3) after MR imaging (e-h). Note the large subserosal myoma in the posterior wall
intravenous gadolinium contrast administration at 30, 60, 90, and 120 s. of the uterus

in age (p = 0.796), tumour size (p = 0.501) and tumour diagnostic accuracy and sensitivity of DW MR imaging were
ADC value (p = 0.213). significantly higher than those of DCE and T2-weighted MR
imaging for both Reader 1 and Reader 2. The overall diagnos-
DW, DCE, and T2-weighted MR imaging in the detection tic performance of DW MR imaging (Reader 1: AUC = 0.98;
of cervical stromal invasion 95 % CI: 0.93, 1.00; Reader 2: AUC = 0.97, 95 % CI: 0.90,
1.00) was significantly higher than that of DCE MR imaging
The reader agreement rate was excellent (k = 0.82) for DCE for Reader 2 (p = 0.009) and T2-weighted MR imaging
MR imaging and good for DW (k = 0.74) and T2-weighted (Reader 1: p = 0.006; Reader 2: p = 0.013) (Fig. 5) Table 3.
MR imaging (k = 0.74). Examples of MR imaging with cervi-
cal stromal invasion are depicted in Figures 2, 3 and 4. For Diagnostic pitfalls
detecting cervical stromal invasion, DW MR imaging
achieved an accuracy, sensitivity, and specificity of 95.2 %, Canal widening was significantly associated with false-
91.7 %, and 95.8 % for Reader 1 and 91.6 %, 100 %, and negative results on DCE and T2-weighted MR imaging but
90.1 % for Reader 2, respectively, as detailed in Table 2. The not on DW MR imaging for both Reader 1 and Reader 2

Table 2 Diagnostic result of detecting cervical stromal invasion in endometrial cancer patients

n TP TN FP FN Accuracy Sensitivity Specificity PPV NPV

Reader 1
DW 83 11 68 3 1 95.2 (88.1-98.7) 91.7 (61.5-99.8) 95.8 (88.1-99.1) 78.6 (49.2-95.3) 98.6 (92.2-100.0)
DCE 83 7 69 2 5 91.6 (83.4-96.5)* 58.3 (27.7-84.8)* 97.2 (90.2-99.7) 77.8 (40.0-97.2) 93.2 (84.9-97.8)
T2-weighted 83 6 61 10 6 80.7 (70.6-88.6)* 50.0 (21.1-78.9)* 85.9 (75.6-93.0) 37.5 (15.2-64.6) 91.0 (81.5-96.6)
Reader 2
DW 83 12 64 7 0 91.6 (83.4-96.5) 100.0 90.1 (80.7-95.9) 63.2 (38.4-83.7) 100.0
DCE 83 6 67 4 6 88.0 (79.0-94.1)* 50.0 (21.1-78.9)* 94.4 (86.2-98.4) 60.0 (26.2-87.8) 91.8 (83.0-96.9)
T2-weighted 83 8 55 16 4 75.9 (65.3-84.6)* 66.7 (34.9-90.1)* 77.5 (66.0-86.5) 33.3 (15.6-55.3) 93.2 (83.5-98.1)

Note—Data are numbers. In parentheses are 95 % confidence intervals. DCE = dynamic contrast-enhanced T1-weighted imaging, DW = diffusion-
weighted, MRI = magnetic resonance imaging, TP = true positive, TN = true negative, FP = false positive, FN = false negative, PPV = positive predictive
value, NPV = negative predictive value. *, p < 0.05
Eur Radiol (2017) 27:1867–1876 1873

Fig. 5 Receiver operating

characteristic curve for cervical
stromal invasion with DW, DCE,
and T2-weighted MR imaging for
Reader 1 (a) and Reader 2 (b)

(Table 4). Nabothian cysts in the cervix resulted in focal high and Reader 2, respectively, and none of them had cervical
signal intensity on DW MR imaging because of the T2 shine- stromal invasion. Subgroup analysis showed the respective
through effect, which was readily corrected with reference to specificity, NPV and accuracy of DW, DCE and T2-
the ADC map. The occurrence of nabothian cysts and lower weighted MR imaging achieved 100 % for both Reader 1
tumour location caused false diagnostic cases on DCE and T2- and Reader 2, whilst the sensitivity, PPV and AUC could not
weighted MR imaging, but the association was not significant. be calculated.
Review of the only false-negative case in DW MR imaging
revealed microscopic cervical stromal invasion. After exclud-
ing cases with nabothian cyst, lower tumour location or cer- Discussion
vical canal widening, there were 18 and 15 cases for Reader 1
We report herein the usefulness of DW MR imaging with
reference to ADC mapping in detecting cervical stromal
invasion of endometrial cancer. The clinical implications
Table 3 Pairwise comparisons of ROC curves in prediction of cervical of our results may facilitate clinical decision making re-
stromal invasion
garding the choice between radical and simple hysterec-
Variable AUC Value P Value Bonferroni tomy [10], or offering fertility-preserving treatment op-
P Value tions for candidates without cervical stromal invasion
[11]. In addition, preoperative pinpointing of the probable
Reader 1
area of cervical stromal invasion on imaging can guide
DW 0.98
adequate pathological tissue sampling, because cervical
DCE 0.94
stromal invasion may affect only small areas within the
T2-weighted 0.85
circumference of the cervix.
DW vs DCE .055 .328
Significant improvement in diagnostic accuracy and sensi-
DW vs T2-weighted .006* .034†
tivity were found in DW MR imaging compared with DCE
DCE vs T2-weighted .051 .306
and T2-weighted MR imaging in this study. The diagnostic
Reader 2
performance of T2-weighted MR imaging in this study is con-
DW 0.97
sistent with that reported in the literature with an accuracy,
DCE 0.88
sensitivity, and specificity of 87 %, 58 %, and 95 %, respec-
T2-weighted 0.87
tively [37]. Two previous large-scale studies demonstrated the
DW vs DCE .009* .051
accuracy, sensitivity, and specificity of contrast-enhanced MR
DW vs T2-weighted .013* .075 imaging as 82 %–90 %, 33 %–42 %, and 95 %–97 %, respec-
DCE vs T2-weighted .410 2.462 tively [38, 39]. However, DCE MR imaging has been reported
*The differences were not significant according to Bonferroni correction to have variable results in the literature, with accuracy and
for multiple comparisons. sensitivity of 46 %–88 % and 19 %–61 % at 1 T [14, 31,
†The differences were significant according to Bonferroni correction for 40], and 81 %–95 % and 42 %–84 % at 1.5 T [12, 15, 22,
multiple comparisons 27, 37, 41, 42], respectively. The limited accuracy and
1874 Eur Radiol (2017) 27:1867–1876

Table 4 False-positive and false

negative results Reader 1 Reader 2

DW DCE T2-weighted DW DCE T2-weighted

FP FN FP FN FP FN FP FN FP FN FP FN

Nabothian cyst 2 0 1 2 5 2 4 0 2 2 7 2
Location 3 0 2 3 8 4 7 0 4 1 14 0
Canal widening 1 1 2 4* 8 4* 6 0 4 6* 12 4*

*The differences were significant according to Chi-square comparison

sensitivity of DCE MR imaging in the present study is con- The role of the ADC value remains controversial in the
sistent with the previously mentioned evidence from the literature. Quartile ADC values were higher [44] but mini-
literature. mum tumour ADC values were lower in tumours with cervical
Initial reports on DW MR imaging in the detection of cer- stromal invasion [45]. However, mean tumour ADC values
vical stromal invasion did not include the element of ADC did not differ between patients with and those without cervical
mapping. Koplay et al. analysed 58 endometrial cancer pa- stromal invasion in the present study. Therefore, value predic-
tients with 11 cases of cervical stromal invasion, demonstrat- tion of cervical stromal invasion by using a b-value cutoff was
ing a higher diagnostic performance of DW MR imaging com- not performed. By using volumetric ADC histogram analysis,
pared with DCE MR imaging (accuracy, sensitivity, and spec- Nougaret et al. demonstrated significantly lower ADC values
ificity of 89.6 %, 82 %, and 91 % vs 86 %, 73 %, and 89 %) at in grade III endometrial cancer and cancers with
1.5 T [27]. However, according to another report of 71 endo- lymphovascular invasion [46]. Whether volumetric ADC his-
metrial cancer patients with six cases of cervical stromal inva- togram analysis predicts cervical stromal invasion warrants
sion, Hori et al. showed virtually identical results in accuracy, further investigation. Nevertheless, the ADC map was useful
sensitivity, and specificity for DW MR imaging and DCE MR for tumour delineation in this study, because visual inspection
imaging at 3 T [28]. More recently, Bonatti et al. reported on a may facilitate differentiating diffusion-restricted lesions from
DW MR imaging study with ADC mapping at 1.5 T, showing those with the T2 shine-through effect caused by benign fluid-
that that DW MR imaging can reliably replace contrast- containing structures, such as inflamed tissue and nabothian
enhanced MR imaging for the local staging of endometrial cysts of the cervix [47].
carcinoma with accuracy, sensitivity, and specificity of Some limitations were encountered in this study. First, only
95 %, 98 %, and 80 %, respectively [29]. Our data further 12 of 83 endometrial carcinoma patients were found to have
support the use of DW MR imaging with ADC mapping at cervical stromal invasion. Such a low incidence may restrict
3 T, which achieves a diagnostic performance comparable to the statistical power of this study. However, to our knowledge,
that of DCE MR imaging. the present study is the largest series aiming for comparing DW
DCE MR imaging facilitates distinguishing stromal inva- and DCE MR imaging in the assessment of cervical stromal
sion from polypoid tumour protrusion from the endometrial invasion in endometrial carcinoma. Second, histopathological
cavity into the endocervix [23]. Normal enhancement of the diagnosis might be difficult in cases of microscopic cervical stro-
cervical mucosa may exclude stromal invasion [22, 23, 43], mal invasion [34]. Therefore, surgical staging for patients with
and thus prevents the false-positive identification that occurs negative findings on MR imaging cannot be completely elimi-
on T2-weighted MR imaging [22]. In this study, however, nated unless the findings are confirmed through hysteroscopic
linear enhancement in the endocervical canal could not ex- examination [18]. We endeavoured to optimise the reference
clude cervical stromal invasion, especially when the tumour standards through additional immunohistochemistry studies in-
dilated the endocervical canal. One might argue that the DCE volving verification by two pathologists independently. Third,
MR imaging criteria were according to the 1988 FIGO classi- tumour volumetry and histogram analysis were not planned.
fication system of the study design [22, 31], which may have Nonetheless, average ADC values on the slice with the maximal
limited the accuracy in DCE images evaluation. Nonetheless, tumour area from two readers were planned to reduce the mea-
we found a significant association between canal widening suring bias [46]. Direct comparison between DW MR imaging
and false-negative results on DCE and T2-weighted MR im- with and without reference to ADC maps was not the intention in
aging. If the clinicians can be informed about the presence of this study. Furthermore, to reduce the adverse effects of medica-
nabothian cyst, lower tumour location and cervical canal wid- tion and patient discomfort, we did not administrate anti-
ening, the pre-treatment diagnostic accuracy may be further peristaltic agents in this prospective study. Nevertheless, no re-
improved by conducting additional examinations. markable bowel motion artefacts affected the quality of DCE or
Eur Radiol (2017) 27:1867–1876 1875

DW MR images in our study cohort. Finally, DCE MR imaging 2. Turan T, Hizli D, Yilmaz SS et al (2012) What is the impact of
cervical invasion on lymph node metastasis in patients with stage
was performed at 0, 30, 60, 90, and 120 seconds only in the
IIIC endometrial cancer? Arch Gynecol Obstet 285:1119–1124
sagittal plane, and more axial oblique acquisition would have 3. Solmaz U, Mat E, Dereli M et al (2015) Lymphovascular space
improved the diagnostic performance. However, the 30-second invasion and cervical stromal invasion are independent risk factors
interval comprised a 17-second breath hold for acquisition and a for nodal metastasis in endometrioid endometrial cancer. Aust N Z J
13-second free breathing for patients’ oxygenation recovery, and Obstet Gynaecol 55:81–86
4. Taskin S, Ortac F, Kahraman K, Goc G, Oztuna D, Gungor M
no additional time for more axial oblique acquisition. Further (2013) Cervical stromal involvement can predict survival in ad-
technical development for faster acquisition or free breathing vanced endometrial carcinoma: a review of 67 patients. Int J Clin
sequences may solve this limitation. Despite these limitations, Oncol 18:105–109
our data clearly demonstrated the impact of DW MR imaging on 5. Kwon JS, Qiu F, Saskin R, Carey MS (2009) Are uterine risk
factors more important than nodal status in predicting survival in
the diagnostic accuracy of cervical stromal invasion in patients endometrial cancer? Obstet Gynecol 114:736–743
with endometrial cancers. 6. Ferriss JS, Brix W, Tambouret R, DeSimone CP, Stoler M, Modesitt
In conclusion, this study demonstrated improved perfor- SC (2010) Cervical stromal invasion predicting survival in endo-
mance of DW MR imaging in the diagnosis of cervical stromal metrial cancer. Obstet Gynecol 116:1035–1041
invasion for endometrial cancer patients, compared with DCE 7. Tewari KS, Filiaci VL, Spirtos NM et al (2012) Association of
number of positive nodes and cervical stroma invasion with out-
MR imaging. Collectively, our evidence suggests that DW MR come of advanced endometrial cancer treated with chemotherapy or
imaging is a legitimate alternative to DCE MR imaging for whole abdominal irradiation: a Gynecologic Oncology Group
assessing cervical stromal invasion. Although MR study is clear- study. Gynecol Oncol 125:87–93
ly not cost-effective as a screening tool, it provides valuable 8. Pecorelli S (2009) Revised FIGO staging for carcinoma of the vul-
va, cervix, and endometrium. Int J Gynaecol Obstet 105:103–104
information especially when ultrasonographic results are inde-
9. NCCN Clinical Practice Guidelines in Oncology: Uterine Cancers.
terminate. The potential savings in both time and cost with the National Comprehensive Cancer Network, USA. Available via
elimination of unnecessary intravenous contrast medium admin- http://www.nccn.org/professionals/physician_gls/pdf/uterine.pdf.
istration during MR examination remain to be assessed through Accessed 23 May, 2016
cost-effectiveness analysis with patients’ clinical circumstances 10. American College of O, Gynecologists (2005) ACOG practice bul-
letin, clinical management guidelines for obstetrician-gynecolo-
taken into account, especially for elderly patients with impaired gists, number 65, August 2005: management of endometrial cancer.
renal function at risk of nephrogenic systemic fibrosis. Obstet Gynecol 106:413–425
11. Eskander RN, Randall LM, Berman ML, Tewari KS, Disaia PJ,
Bristow RE (2011) Fertility preserving options in patients with gy-
Acknowledgments The authors thank all the members of the Cancer necologic malignancies. Am J Obstet Gynecol 205:103–110
Center, Chang Gung Memorial Hospital, for their invaluable help. The 12. Manfredi R, Mirk P, Maresca G et al (2004) Local-regional staging
scientific guarantor of this publication is Chyong-Huey Lai. The authors of endometrial carcinoma: role of MR imaging in surgical planning.
of this manuscript declare no relationships with any companies, whose Radiology 231:372–378
products or services may be related to the subject matter of the article. 13. Kinkel K, Kaji Y, Yu KK et al (1999) Radiologic staging in patients
This study has received funding by Chang Gung Medical Foundation with endometrial cancer: a meta-analysis. Radiology 212:711–718
grant CMRPG-3B1923, CIRPG3E0021, National Science Council 14. Ortashi O, Jain S, Emannuel O, Henry R, Wood A, Evans J (2008)
(Taiwan) 104-2314-B-182A-095-MY3. Dr Lan-Yan Yang kindly provid- Evaluation of the sensitivity, specificity, positive and negative pre-
ed statistical advice for this manuscript. Institutional Review Board ap- dictive values of preoperative magnetic resonance imaging for stag-
proval was obtained. Written informed consent was obtained from all ing endometrial cancer. A prospective study of 100 cases at the
subjects (patients) in this study. Chang Gung IRB No. 101-2187B, 103- Dorset Cancer Centre. Eur J Obstet Gynecol Reprod Biol 137:
7316A3. The authors thank all the members of the Cancer Center, Chang 232–235
Gung Memorial Hospital, for their invaluable help. 15. Cicinelli E, Marinaccio M, Barba B et al (2008) Reliability of di-
Some study subjects or cohorts had been previously reported to inves- agnostic fluid hysteroscopy in the assessment of cervical invasion
tigate the influence of menopausal status on diagnostic accuracy of MR by endometrial carcinoma: a comparative study with transvaginal
imaging for evaluation of myometrial invasion: Lin G, Huang YT, Chao sonography and MRI. Gynecol Oncol 111:55–61
A, et al (2015) Influence of menopausal status on diagnostic accuracy of 16. Mascilini F, Testa AC, Van Holsbeke C, Ameye L, Timmerman D,
myometrial invasion in endometrial cancer: diffusion-weighted and dy- Epstein E (2013) Evaluating myometrial and cervical invasion in
namic contrast-enhanced MRI at 3 T. Clin Radiol 70:1260-1268. In the women with endometrial cancer: comparing subjective assessment
current manuscript we selected the 83 patients who had completed hys- with objective measurement techniques. Ultrasound Obstet
terectomy and compared the accuracy of DW and DCE MR imaging in Gynecol 42:353–358
diagnosis of cervical stromal invasion in endometrial cancer. 17. Haldorsen IS, Berg A, Werner HM et al (2012) Magnetic resonance
Methodology: prospective, diagnostic or prognostic study, performed imaging performs better than endocervical curettage for preopera-
at one institution. tive prediction of cervical stromal invasion in endometrial carcino-
mas. Gynecol Oncol 126:413–418
18. Luomaranta A, Leminen A, Loukovaara M (2015) Magnetic reso-
nance imaging in the assessment of high-risk features of endome-
References trial carcinoma: a meta-analysis. Int J Gynecol Cancer 25:837–842
19. Rauch GM, Kaur H, Choi H et al (2014) Optimization of MR
1. Siegel RL, Miller KD, Jemal A (2016) Cancer statistics, 2016. CA imaging for pretreatment evaluation of patients with endometrial
Cancer J Clin 66:7–30 and cervical cancer. Radiographics 34:1082–1098
1876 Eur Radiol (2017) 27:1867–1876

20. Larson DM, Connor GP, Broste SK, Krawisz BR, Johnson KK 34. McCluggage WG, Hirschowitz L, Wilson GE, Oliva E, Soslow
(1996) Prognostic significance of gross myometrial invasion with RA, Zaino RJ (2011) Significant variation in the assessment of
endometrial cancer. Obstet Gynecol 88:394–398 cervical involvement in endometrial carcinoma: an interobserver
21. Nagar H, Dobbs S, McClelland HR, Price J, McCluggage WG, variation study. Am J Surg Pathol 35:289–294
Grey A (2006) The diagnostic accuracy of magnetic resonance 35. Fleiss JL (1981) Statistical methods for rates and proportions.
imaging in detecting cervical involvement in endometrial cancer. Wiley, New York
Gynecol Oncol 103:431–434 36. Medina LS, Zurakowski D (2003) Measurement variability and
22. Seki H, Takano T, Sakai K (2000) Value of dynamic MR imaging in confidence intervals in medicine: why should radiologists care?
assessing endometrial carcinoma involvement of the cervix. AJR Radiology 226:297–301
Am J Roentgenol 175:171–176 37. Emlik D, Kiresi D, Ozdemir S, Celik C, Karakose S (2010)
23. Freeman SJ, Aly AM, Kataoka MY, Addley HC, Reinhold C, Sala Preoperative assessment of myometrial and cervical invasion in
E (2012) The revised FIGO staging system for uterine malignan- endometrial carcinoma: comparison of multi-section dynamic MR
cies: implications for MR imaging. Radiographics 32:1805–1827 imaging using a three dimensional FLASH technique and T2-
24. Lin G, Ng KK, Chang CJ et al (2009) Myometrial invasion in weighted MR imaging. J Med Imaging Radiat Oncol 54:202–210
endometrial cancer: diagnostic accuracy of diffusion-weighted 38. Antonsen SL, Jensen LN, Loft A et al (2013) MRI, PET/CT and
3.0-T MR imaging–initial experience. Radiology 250:784–792 ultrasound in the preoperative staging of endometrial cancer - a
25. Beddy P, Moyle P, Kataoka M et al (2012) Evaluation of depth of multicenter prospective comparative study. Gynecol Oncol 128:
myometrial invasion and overall staging in endometrial cancer: 300–308
comparison of diffusion-weighted and dynamic contrast-enhanced 39. Duncan KA, Drinkwater KJ, Frost C, Remedios D, Barter S (2012)
MR imaging. Radiology 262:530–537 Staging cancer of the uterus: a national audit of MRI accuracy. Clin
26. Kinkel K, Forstner R, Danza FM et al (2009) Staging of endome- Radiol 67:523–530
trial cancer with MRI: guidelines of the European Society of 40. Vasconcelos C, Felix A, Cunha TM (2007) Preoperative assessment
Urogenital Imaging. Eur Radiol 19:1565–1574 of deep myometrial and cervical invasion in endometrial carcino-
27. Koplay M, Dogan NU, Erdogan H et al (2014) Diagnostic efficacy ma: comparison of magnetic resonance imaging and histopatholog-
of diffusion-weighted MRI for pre-operative assessment of ic evaluation. J Obstet Gynaecol 27:65–70
myometrial and cervical invasion and pelvic lymph node metastasis 41. Cabrita S, Rodrigues H, Abreu R et al (2008) Magnetic resonance
in endometrial carcinoma. J Med Imaging Radiat Oncol 58:538– imaging in the preoperative staging of endometrial carcinoma. Eur J
546 Gynaecol Oncol 29:135–137
28. Hori M, Kim T, Onishi H et al (2013) Endometrial cancer: preop- 42. Teng F, Zhang YF, Wang YM et al (2015) Contrast-enhanced MRI
erative staging using three-dimensional T2-weighted turbo spin- in preoperative assessment of myometrial and cervical invasion,
echo and diffusion-weighted MR imaging at 3.0 T: a prospective and lymph node metastasis: diagnostic value and error analysis in
comparative study. Eur Radiol 23:2296–2305 endometrial carcinoma. Acta Obstet Gynecol Scand 94:266–273
29. Bonatti M, Stuefer J, Oberhofer N et al (2015) MRI for local staging 43. Ascher SM, Reinhold C (2002) Imaging of cancer of the endome-
of endometrial carcinoma: Is endovenous contrast medium admin- trium. Radiol Clin North Am 40:563–576
istration still needed? Eur J Radiol 84:208–214 44. Cao K, Gao M, Sun YS et al (2012) Apparent diffusion coefficient
30. Lin G, Huang YT, Chao A et al (2015) Influence of menopausal of diffusion weighted MRI in endometrial carcinoma-relationship
status on diagnostic accuracy of myometrial invasion in endometri- with local invasiveness. Eur J Radiol 81:1926–1930
al cancer: diffusion-weighted and dynamic contrast-enhanced MRI 45. Shih IL, Yen RF, Chen CA et al (2015) Standardized uptake value
at 3 T. Clin Radiol 70:1260–1268 and apparent diffusion coefficient of endometrial cancer evaluated
31. Sanjuan A, Escaramis G, Ayuso JR et al (2008) Role of magnetic with integrated whole-body PET/MR: correlation with pathological
resonance imaging and cause of pitfalls in detecting myometrial prognostic factors. J Magn Reson Imaging 42:1723–1732
invasion and cervical involvement in endometrial cancer. Arch 46. Nougaret S, Reinhold C, Alsharif SS et al (2015) Endometrial can-
Gynecol Obstet 278:535–539 cer: combined MR volumetry and diffusion-weighted imaging for
32. Tamai K, Koyama T, Saga T et al (2007) Diffusion-weighted MR assessment of myometrial and lymphovascular invasion and tumor
imaging of uterine endometrial cancer. J Magn Reson Imaging 26: grade. Radiology 276:797–808
682–687 47. Padhani AR, Koh DM, Collins DJ (2011) Whole-body diffusion-
33. Kurman RJ (2002) Blaustein’s pathology of the female genital tract, weighted MR imaging in cancer: current status and research direc-
5 edn. Springer, Berlin Heidelberg New York tions. Radiology 261:700–771