European Journal of Radiology 107 (2018) 111–118

Contents lists available at ScienceDirect

European Journal of Radiology

journal homepage: www.elsevier.com/locate/ejrad

Research article

Breast MRI for prediction of lymphovascular invasion in breast cancer T

patients with clinically negative axillary lymph nodes
⁎
Takao Igarashi , Hisayo Furube, Hirokazu Ashida, Hiroya Ojiri
Department of Radiology, The Jikei University School of Medicine, 3-25-8, Nishi-Shimbashi, Minato-ku, Tokyo 105-8461, Japan

A R T I C LE I N FO A B S T R A C T

Keywords: Objective: To retrospectively assess magnetic resonance imaging (MRI) findings that can predict lymphovascular
Breast invasion (LVI) in invasive breast cancer patients who were diagnosed with clinically negative axillary lymph
MRI nodes (LNs) preoperatively.
Lymphovascular invasion Methods: This study included 140 lesions of 140 patients who underwent preoperative breast MRI and breast
surgery, with omission of axillary LN dissection. Clinical characteristics and MRI findings were evaluated. The T2
signal intensity (SI) ratio (mean T2 SI of the tumor/mean T2 SI of the muscle), tumor apparent diffusion
coefficient (ADC) value, peritumoral ADC value, peritumor-tumor ADC ratio (peritumoral maximum ADC value/
tumor mean ADC value), and ADC value of the contralateral breast parenchyma were retrospectively assessed.
Statistical analyses were performed to identify significant factors for predicting LVI. Inter-observer variability
was calculated.
Results: The tumor ADC value (all ages: p = 0.005; age ≤ 55: p < 0.001), peritumoral ADC value (age ≤ 55: p
= 0.04), and peritumor-tumor ADC ratio (all ages: p < 0.001; age ≤ 55: p < 0.001) were significantly asso-
ciated with LVI on univariate analysis. Multivariate logistic regression analysis revealed significant differences in
the pathological size of the invasive component and the tumor ADC value for predicting LVI (odds ratio [OR]:
3.43; 95% confidence interval [CI]: 1.41–8.32; p = 0.007; OR: 16.0; 95% CI: 1.89–136; p = 0.01, respectively).
Inter-observer agreement was substantial for the tumor ADC value (intraclass correlation coefficient
[ICC] = 0.77; 95% CI: 0.70–0.83) and the ADC value of the contralateral breast parenchyma (ICC = 0.68; 95%
CI: 0.59–0.76). There was moderate agreement for the peritumoral ADC value (ICC = 0.53; 95% CI: 0.40–0.64)
and the peritumor-tumor ADC ratio (ICC = 0.49; 95% CI: 0.35–0.61) and fair agreement for the T2 SI ratio
(ICC = 0.30; 95% CI: 0.15–0.45).
Conclusion: We found that the tumor ADC value, peritumoral ADC value, and peritumor-tumor ADC ratio were
predictive MRI findings for LVI in patients aged ≤55. The tumor ADC value was the most significant predictor
for LVI; moreover, inter-observer agreement for the tumor ADC value was substantial between two blinded
observers with differences in interpretation experience.

1. Introduction essential for breast cancer patients to undergo preoperative imaging of

Lymphovascular invasion (LVI) has been pathologically defined as
the presence of tumor cells within both lymphatic and vascular space in
the area surrounding an invasive carcinoma [1]. LVI has been accepted
as sufficiently reliable to define risk of relapse and survival rate in
lymph node (LN)-negative breast cancer patients [2–5]. A strong asso-
ciation between axillary nodal metastasis and LVI was found in a large
study, in which positive LNs were detected in 51% of LVI-positive pa-
tients, compared with 19% of LVI-negative patients [6]. It is, therefore,
LVI and axillary LN in terms of prognostic prediction [3,7]. Since LVI is
confirmed by histopathological examination, based on an excised spe-
cimen that contains primary lesion and perilesional breast tissue, it is
difficult to confirm LVI by preoperative biopsy, which contains only
primary lesion. Therefore, preoperative magnetic resonance imaging
(MRI) is expected to provide an imaging biomarker that can predict LVI.
Some recent studies have shown that apparent diffusion coefficient
(ADC) value is a prognostic factor related to tumor aggressiveness in
patients with breast cancer [8–10]. Recent retrospective cohort studies

Abbreviations: MRI, magnetic resonance imaging; LVI, lymphovascular invasion; LN, lymph node; SI, signal intensity; ADC, apparent diffusion coefficient; OR, odds
ratio; CI, confidence interval
⁎
Corresponding author.
E-mail addresses: igarashi-t@jikei.ac.jp (T. Igarashi), maedahisayo@jikei.ac.jp (H. Furube), ashiyan@jikei.ac.jp (H. Ashida), ojiri@jikei.ac.jp (H. Ojiri).

https://doi.org/10.1016/j.ejrad.2018.08.024
Received 2 May 2018; Received in revised form 1 August 2018; Accepted 26 August 2018
0720-048X/ © 2018 Elsevier B.V. All rights reserved.

Downloaded for A. Muh. Yasser Mukti (andimuh_yassermukti@dr.com) at University of Muslim Indonesia from ClinicalKey.com by Elsevier on July 22, 2019.
For personal use only. No other uses without permission. Copyright ©2019. Elsevier Inc. All rights reserved.
T. Igarashi et al.
found that ADC values were significantly lower in the LVI-positive
group than the -negative group [11,12]. A lower ADC value is asso-
ciated with reduction in the speed of osmosis in tumor tissue and with
greater tumor cell proliferation. LVI was associated with a high tumor
cell proliferation level or high Ki-67 labeling index [13]. Peritumor-
tumor ADC ratio has been reported to be a feasible method of MRI, and
is device-independent. This method could objectively assess lymphe-
dema caused by LVI [12].
Little has been reported regarding the combination of preoperative
findings, including MRI, associated with prediction of LVI in breast
cancer patients who were preoperatively diagnosed with negative ax-
illary LNs (hereafter defined as “clinically negative nodes”).
Identification of predictive findings that are associated with LVI has
enabled us to determine the need for additional therapy. This study was
undertaken to characterize preoperative MRI findings that can predict
LVI in invasive breast cancer patients with clinically negative nodes.
2. Material and methods
2.1. Patients
This retrospective study was approved by the Committee on Clinical
Studies at our institution (approval number, 29-268). The requirement
for informed consent was waived. In total, 921 consecutive patients
underwent breast MRI between July 17, 2015 and June 30, 2017 at our
institution. The phase of the menstrual cycle of patients aged ≤55 was
not considered because this was a retrospective study; therefore, it was
not possible to analyze the phase of menstrual cycle during the MRI
examination. The following inclusion criteria were used: (a) patients
who had no suspicious findings, suggestive of metastatic LNs, on both
preoperative breast ultrasonography and MRI; (b) patients who un-
derwent both sentinel lymph node (SLN) biopsy and breast surgery,
with omission of axillary LN dissection. Preoperative imaging findings
suggestive of negative axillary LNs were evaluated based on a previous
report [13], as follows: negative for round-shaped LNs with a com-
pletely or partially effaced fatty hilum; negative for LNs with irregular
margins and/or cortical thickening (≥5 mm; focal or diffuse); negative
for extra-hilar blood vessel flow on color doppler analysis.
The following exclusion criteria were used: (a) patients who did not
undergo US and/or MRI within the six-month period immediately
preceding surgery (n = 47); (b) patients who had a past medical history
of neoadjuvant chemotherapy and/or radiotherapy and/or breast sur-
gery (n = 47); (c) patients who were diagnosed with ductal carcinoma
in situ (n = 39), (d) intracystic papillary carcinoma (n = 2), (d) in-
traductal papilloma (n = 1), (e) Paget’s disease (n = 4), (f) invasive
cancer of < 1 mm (n = 6) on histopathological examination; and (g)
patients with small lesions in which an region of interest (ROI) could
not be placed using a picture archiving and communication systems
(PACS) monitor (n = 2). A total of 140 patients met the inclusion and
exclusion criteria (Fig. 1).
2.2. Image acquisition
MRI was performed by using a 1.5-T system (Symphony, Siemens
Medical Solutions, Erlangen, Germany) with a maximum gradient field
strength of 30 m T/m and a 4-ch CP breast array coil. All patients were
examined in the prone position. First, coronal T2-weighted spectral
attenuated inversion recovery images of both breasts were obtained
with the following parameters: repetition time (TR)/echo time (TE),
4000 ms/73 ms; flip angle, 150°; field of view, 35 × 35 cm; matrix size,
806 × 896; slice thickness, 4 mm; gap, 0; number of excitations (NEX),
1. Subsequently, coronal diffusion-weighted images of both breasts
were obtained at b values of 0, 1000, and 1500 s/mm2 with the fol-
lowing parameters: TR/TE, 5600 ms/87 ms; field of view, 35 × 35 cm;
matrix size, 234 × 320; slice thickness, 3.5 mm; slice gap, 0; NEX, 5.
ADC maps were automatically generated on the operating console using
112
Downloaded for A. Muh. Yasser Mukti (andimuh_yassermukti@dr.com) at University of Muslim Indonesia from ClinicalKey.com by Elsevier on July 22, 2019.
For personal use only. No other uses without permission. Copyright ©2019. Elsevier Inc. All rights reserved.
European Journal of Radiology 107 (2018) 111–118
the least squares method with all three images and b values of 0, 1000,
and 1500 s/mm2. Dynamic contrast-enhanced MR (CEM) images were
acquired with a three-dimensional fat-suppressed volumetric inter-
polated breath-hold examination sequence by using the following
parameters: TR/TE, 5 ms/2.41 ms; flip angle, 15°; field of view,
340 mm; matrix, 768 × 768; receiver bandwidth, 340 kHz/pixel; mean
partition thickness, 0.9 mm; time of acquisition, 60 s; NEX, 1. The
section thickness varied depending on the size of the breast. Sections
were acquired without a gap. Both breasts were included in the images.
Three contrast-enhanced acquisitions were acquired at 60, 120, and
240 s after the start of the IV administration of 0.1 mmol/kg gado-
pentetate dimeglumine or gadobutrol (Magnevist or Gadovist; Bayer
HealthCare, Berlin, Germany); administration was at a rate of 1 mL/s,
followed by a 15 mL saline flush, and was performed with an automatic
injector.
2.3. Image analysis
One radiologist with 16 years of experience retrospectively eval-
uated the clinical characteristics for which clinical and pathological
information were available. Then, another radiologist with 16 years of
experience in breast MR imaging (first blinded observer) re-evaluated
the continuous variables of MRI findings on a PACS monitor while
blinded to the clinical and pathological information. All cases were
anonymized for re-evaluation. Another radiologist with 4 years of ex-
perience in breast MRI (second blinded observer) evaluated the con-
tinuous variables of the MRI findings to assess the reproducibility of the
first observer’s evaluation. The second observer was blinded to the re-
sults of the first observer and to the clinical and histopathological in-
formation.
Continuous variables of MRI findings included the T2 signal in-
tensity (SI) ratio, tumor ADC value, peritumoral ADC value, peritumor-
tumor ADC ratio, and ADC value of the contralateral breast par-
enchyma. ROIs were placed on targeted regions by referring to the
findings of the dynamic CEM images. In a case of a multifocal or
multicentric lesion, the largest tumor was accepted for analysis. For
measurements of T2 SI of the tumor, the largest tumor cross section was
selected, and an oval or round ROI, as large as possible, was placed
inside the tumor. Another ROI was also placed on the pectoralis major
muscle to measure the T2 SI of the muscle and calculate the T2 SI ratio
between the lesion and the muscle. The T2 SI ratio was calculated ac-
cording to the following formula:
T2 SI ratio = mean T2 SI of the tumor/mean T2 SI of the muscle. (1)
The tumor ADC value, peritumoral ADC value, and ADC value of the
contralateral breast parenchyma measurements were calculated using
ADC maps. To measure the tumor ADC value, ROIs were manually
placed within the solid component of the tumor. To measure the peri-
tumoral ADC value and the ADC value of the contralateral breast par-
enchyma, ROIs were manually placed in the breast parenchyma; fatty
tissue was avoided to the extent possible. ROIs of the peritumoral ADC
value were placed within 2 cm from the border between the tumor and
the normal breast parenchyma. At least three measurements were taken
for all continuous variables on MR images. The mean value among the
three T2 SI ratios, the minimum value among the three mean tumor
ADC values, the maximum value among the three maximum peritu-
moral ADC values, and the minimum value among the three mean ADC
values of the contralateral breast parenchyma were accepted as the T2
SI ratio, tumor ADC value, peritumoral ADC value, and ADC value of
the contralateral breast parenchyma, respectively. We calculated the
peritumor-tumor ADC ratio, which has been reported to be a feasible
method, according to the following formula [12]:
Peritumor-tumor ADC ratio = peritumoral maximum ADC value/tumor
mean ADC value. (2)
T. Igarashi et al. European Journal of Radiology 107 (2018) 111–118

Fig. 1. Flow diagram for the patient selection process.

Inter-observer variability was assessed for each continuous variable.
An age-dependent evaluation was conducted for the continuous vari-
ables with a cutoff at age 55, in consideration of the fact that natural
menopause occurs at up to 55 years old for women worldwide.
2.4. Assessment of sentinel node
SLN biopsy was performed with a radiocolloid and a dye. Mapping
agents were injected into the subdermal plexus. 99mTC-labeled phytate
colloid was injected on the day of surgery (0.25 mL, 15 MBq) or the day
before surgery (0.5 mL, 30 MBq), and lymphoscintigraphy was per-
formed. During the surgery, 5 mL of isosulfan blue dye (Lymphazurin;
Covidien, Mansfield, MA, USA) or 3 mL of indocyanine green dye
(Diagnogreen; Daiichi Sankyo, Tokyo, Japan) was injected. SLNs were
identified as those with dye uptake, radiotracer uptake, or both.
2.5. Histopathological assessment
All SLNs were sectioned along their short axis at 2-mm intervals.
The nodal tissue was quickly frozen in liquid nitrogen, and a single 5-
μm-thick section, stained with hematoxylin and eosin (H & E), was
examined intraoperatively (frozen-section analysis). In our institution,
breast surgery with omission of axillary LN dissection was performed
for the following cases: (a) SLNs had no metastatic cells or only isolated
tumor cells; (b) SLNs had only 1-2 lymph nodes with metastatic cells,
for patients who had undergone breast-conserving surgery. In this
study, a positive SLN was defined as a LN with isolated tumor cells,
micrometastases (diameter of metastatic deposit, 0.2–2 mm), or mac-
rometastases (metastatic tumor with a maximum diameter > 2 mm).
LVIs were classified into four grades: ly/v 0 (no LVI), ly/v 1 (minimal
LVI), ly/v 2 (moderate LVI), and ly/v 3 (marked LVI). We divided LVIs
into LVI-positive (ly/v 1-3) and LVI-negative groups (ly/v 0).
The following variables were evaluated in all primary tumors: his-
topathological type; immunohistochemical staining for estrogen re-
ceptors (ER), progesterone receptors (PgR), and human epidermal
growth factor receptor 2 (HER2); Ki-67 proliferation index, nuclear
grade, size of the invasive component, and the presence of LVI. The size
of the invasive component was defined as the maximum diameter of the
focal invasion, excluding the intra-ductal component, on histopatho-
logical examination. Hormonal status was considered positive if > 1%
of tumor cells were positive for ER or PgR [14]. HER2 positivity was
defined as a score of 3+ on immunohistochemical staining, or as a
score of 2+ with confirmation of HER2 gene amplification by fluor-
escence in situ hybridization. Ki-67 proliferation index was determined
based on immunohistochemical analysis; an index of < 14% was con-
sidered low, and an index of ≥14% was considered high [15].

113

Downloaded for A. Muh. Yasser Mukti (andimuh_yassermukti@dr.com) at University of Muslim Indonesia from ClinicalKey.com by Elsevier on July 22, 2019.
For personal use only. No other uses without permission. Copyright ©2019. Elsevier Inc. All rights reserved.
T. Igarashi et al. European Journal of Radiology 107 (2018) 111–118

2.6. Statistical analysis Table 1

Clinical characteristics.
All statistical analyses were performed with Ekuseru-Toukei 2015 Variable Patients Patients without p
(SSRI, Tokyo, Japan) and R (The R Project for Statistical Computing, with LVI LVI
www.r-project.org, version 3.3.0) software. Qualitative variables were (n = 41) (n = 99)
analyzed with Fisher’s exact test. Fisher’s exact test with the Benjamini-
Patient age (years) 0.20
Hochberg method was used for multiple pairwise comparisons. ≤55 23 43
Continuous variables were analyzed with the Mann-Whitney U test. > 55 18 56
Diagnostic performance was estimated by using receiver operating Histopathological type
characteristic (ROC) analysis by calculating the area under the ROC Scirrhous carcinoma 28 48
curve (AUC) for each continuous variable that showed a significant Papillotubular carcinoma 8 22
difference on univariate analysis. A cutoff value differentiating the LVI- Solid-tubular carcinoma 3 13
Mucinous carcinoma (pure 3
positive and -negative groups was chosen to maximize the Youden
type)
index. Based on the cutoff values, continuous variables were fit to the Mucinous carcinoma 1
multivariate logistic regression analysis. Multivariate logistic regression (mixed type)
analysis was performed with variables related to MRI findings that Apocrine carcinoma 1
showed significant differences in univariate analyses. p < 0.05 was Metaplastic carcinoma 1
Invasive micropapillary 1
considered statistically significant. Inter-observer variability of the carcinoma
continuous variables was assessed by using intraclass correlation coef- Invasive lobular carcinoma 1 10
ficients (ICCs). An ICC of 0.81–1.0 was considered to indicate almost Intrinsic subtype 0.03
perfect agreement; 0.61–0.80, substantial agreement; 0.41–0.60, mod- Luminal A 6 36
Luminal B 34 52
erate agreement; 0.21–0.40, fair agreement; and ≤0.20, slight agree-
HER2 1 5
ment. Triple-negative 0 6

Ki-67 index
All ages 25 (16–40) 19.3 (10.2–30.6) 0.03
3. Results Age, ≤ 55 25.7 (18–42.55) 17.1 (9.7–32) 0.02
Age, > 55 24.4 20 (11.5–30) 0.37
Histopathological examination revealed patients with (n = 41) and (15.08–30.75)
Nuclear grade 0.21
without LVI (n = 99) among the 140 lesions. Univariate analyses of the
Low or intermediate 32 86
qualitative and continuous variables are summarized in Tables 1 and 2. High 9 13
Results reported regarding continuous variables on MRI are based on Clinical stage 0.002
assessments by the first blinded observer. The median (interquartile T1 17 73
range, 25–75 percentile) of all ROIs was 20 mm2 (15–27) for the first T2 22 24
T3 2 2
observer and 28 mm2 (19–38) for the second observer. In the univariate
analyses of clinical characteristics, the pathological size of the invasive Pathological size of invasive component (mm)
All ages 24 (16–32) 15 (10–22) < 0.001
component (all ages: p < 0.001; age ≤ 55: p < 0.001), Ki-67 index (all
Age, ≤ 55 25 (16.5–36) 12 (9.5–17) < 0.001
ages: p = 0.03; age ≤ 55: p = 0.02), and SLN positivity (p < 0.001) Age, > 55 23 (16.25–25) 16.5 0.06
were significantly associated with LVI. For the intrinsic subtype and (10.75–23.25)
clinical stage, Luminal B tumor and clinical stage T2 are significantly Lesion type 0.11
Mass 40 87
more likely than luminal A and clinical stage T1 to exhibit LVI (adjusted
Non-mass 1 12
p = 0.03 and adjusted p = 0.002, respectively). For continuous vari- SLN < 0.001
ables on MR images, the tumor ADC value (all ages: p = 0.005; age ≤ Positive 21 19
55: p < 0.001), peritumoral ADC value (age ≤ 55: p = 0.04), and Negative 20 80
peritumor-tumor ADC ratio (all ages: p < 0.001; age ≤ 55: p < 0.001)
Abbreviations: LVI: lymphovascular invasion; SLN: sentinel lymph node.
were significantly associated with LVI (Figs. 2 and 3). The only sig-
The size of the invasive component and the Ki-67 index are expressed as the
nificant variable identified in patients aged > 55 years was the ADC
age-based median and interquartile range (25–75 percentile).
value of the contralateral breast parenchyma (p = 0.02). The AUC for
In a multiple comparison of the intrinsic subtype and clinical stage, after per-
the pathological size of the invasive component was 0.72, for the tumor forming the Fisher's exact test with the variables above, the results were cor-
ADC value it was 0.64, and for the peritumor-tumor ADC ratio it was rected using the Benjamini-Hochberg method. Only the lowest adjusted p-va-
0.70. Multivariate logistic regression analysis revealed significant dif- lues are shown. There were significant differences between Luminal A and
ferences in the pathological size of the invasive component, SLN posi- Luminal B, and between clinical stage T1 and T2.
tivity and the tumor ADC value for the prediction of LVI (odds ratio
[OR]: 3.43; 95% confidence interval [CI]: 1.41–8.32; p = 0.007; OR: 4. Discussion
2.96; 95% CI: 1.15–7.62; p = 0.02; OR: 16.0; 95% CI: 1.89–136; p =
0.01, respectively; Table 3). The diagnostic performance, based on the Our results are consistent with a recent retrospective cohort study
data when applying the cutoff values of 2 cm for the pathological size of that reported that breast cancers with LVI had significantly lower ADC
the invasive component and 0.862 × 10−3 mm2/s for the tumor ADC values than breast cancers without LVI, independent of the histological
value, is summarized in Table 4. subtype, grades of the ductal histological type, and mass lesions [11].
Inter-observer agreement was substantial for the tumor ADC value Our results also agree with a previous study that reported that the ADC
(ICC = 0.77; 95% CI: 0.70–0.83) and the ADC value of the contralateral value, peritumor-tumor ADC ratio, tumor diameter, Ki-67 index, and
breast parenchyma (ICC = 0.68; 95% CI: 0.59–0.76). There was mod- axillary LN metastasis were significantly different between patients
erate agreement for the peritumoral ADC value (ICC = 0.53; 95% CI: with and without LVI [12]. The minimum-ADC value has been sug-
0.40–0.64) and the peritumor-tumor ADC ratio (ICC = 0.49; 95% CI: gested as an effective parameter for the prediction of LVI status, and it
0.35–0.61) and fair agreement for the T2 SI ratio (ICC = 0.30; 95% CI: was reported that LVI status had a strong positive correlation with LN
0.15–0.45). status [16]. Thus, our results showed that the minimum mean tumor

114

Downloaded for A. Muh. Yasser Mukti (andimuh_yassermukti@dr.com) at University of Muslim Indonesia from ClinicalKey.com by Elsevier on July 22, 2019.
For personal use only. No other uses without permission. Copyright ©2019. Elsevier Inc. All rights reserved.
T. Igarashi et al. European Journal of Radiology 107 (2018) 111–118

Table 2
Results of continuous variables on MR images.
Variable All ages Age ≤ 55 Age > 55

Patients with LVI Patients without p Patients with LVI Patients without p Patients with LVI Patients without p
(n = 41) LVI (n = 23) LVI (n = 43) (n = 18) LVI (n = 56)
(n = 99)

T2 SI ratio 3.62 (2.75–4.57) 3.45 (2.73–4.88) 0.65 3.03 (2.78–4.02) 3.33 (2.71–4.73) 0.71 4.19 (2.80–5.16) 3.71 (2.93–5.27) 0.90
Tumor ADC value 0.738 0.785 0.005 0.733 0.809 < 0.001 0.754 0.782 0.20
(×10−3 mm2/s) (0.665–0.815) (0.715–0.884) (0.665–0.790) (0.713–0.892) (0.652–0.822) (0.719–0.843)
Peritumoral ADC 1.849 1.717 0.08 1.985 1.743 0.04 1.799 1.707 0.83
value (1.661–2.080) (1.556–1.967) (1.795–2.151) (1.497–1.939) (1.618–1.936) (1.582–1.976)
(×10−3 mm2/s)
Peritumor-tumor ADC 2.60 (2.25–2.72) 2.16 (1.91–2.50) < 0.001 2.65 (2.40–3.02) 2.10 (1.88–2.49) < 0.001 2.44 (2.12–2.67) 2.26 (1.94–2.50) 0.15
ratio
ADC value of the 1.761 1.779 0.95 1.941 1.774 0.14 1.579 1.794 0.02
contralateral (1.547–1.950) (1.608–1.883) (1.747–2.013) (1.608–1.898) (1.424–1.780) (1.613–1.881)
breast
parenchyma
(×10−3 mm2/s)

Abbreviations: MR: magnetic resonance; LVI: lymphovascular invasion; SI: signal intensity; ADC: apparent diffusion coefficient. Continuous variables are expressed as
the median and interquartile range (25–75 percentile).

ADC value may contribute to the prediction of lymphatic invasion. Our
results also showed that the measurement of the tumor ADC value could
be conducted while ensuring reproducibility between blinded readers
with differences in interpretation experience. Previous reports sug-
gested that the tumor ADC value was correlated with tumor
aggressiveness indicators, such as Ki-67, in ER-positive breast cancers
[17,18]. Jacquemier et al. reported that higher Ki-67 values were as-
sociated with vascular and lymphatic invasion [19]. It has also been
reported that premenopausal breast cancer patients tend to have higher
Ki-67 values, while postmenopausal patients tend to have lower Ki-67

Fig. 2. MR images of a 74-year-old woman with invasive

ductal carcinoma (Luminal A) beneath the left nipple.
Lymphovascular invasion and nodal macrometastasis were
confirmed at the histopathological examination. (A)
Coronal diffusion-weighted image (b = 1500) showing a
margin-dominant high-intensity mass measuring approxi-
mately 20 mm. (B) Coronal apparent diffusion coefficient
(ADC) map showing a mass with low signal suggestive of
restricted diffusion (arrow). It can be confirmed visually
that the region surrounding the mass shows higher signal
intensity than the other region including the right mam-
mary gland in the ADC map (arrowheads). The tumor ADC
value was 0.680 (×10−3 mm2/s), the peritumor-tumor
ADC ratio was 2.825.

115

Downloaded for A. Muh. Yasser Mukti (andimuh_yassermukti@dr.com) at University of Muslim Indonesia from ClinicalKey.com by Elsevier on July 22, 2019.
For personal use only. No other uses without permission. Copyright ©2019. Elsevier Inc. All rights reserved.
T. Igarashi et al. European Journal of Radiology 107 (2018) 111–118

Fig. 3. MR images of a 60-year-old woman with invasive

ductal carcinoma (Luminal B) in the medial upper region
of the left breast. No lymphovascular invasion or nodal
metastasis were confirmed on the histopathological ex-
amination. (A) Coronal diffusion-weighted image (b =
1500) showing a margin-dominant high-intensity mass
measuring approximately 12 mm. (B) Coronal apparent
diffusion coefficient (ADC) map showing a mass with low
signal suggestive of restricted diffusion (arrow). The re-
gion surrounding the mass is iso-intensity compared to the
other region including the right mammary gland in the
ADC map (arrowhead). The tumor ADC value was 0.713
(×10−3 mm2/s), peritumor-tumor ADC ratio was 2.157.

Table 3 values [20]; that is, postmenopausal patients tend to have low pro-
Results of multivariate logistic regression analysis. liferative activity. Both the tumor ADC value and the Ki-67 index of
Variable OR 95% CI p
patients aged ≤55 years showed significant differences regarding the
LVI status, but there were no significant differences in patients
Pathological size of invasive component 3.43 1.41–8.32 0.007 aged > 55 years on univariate analysis. Our study indicated that the
(> 2 cm vs. ≤ 2 cm) lower tumor ADC value and higher Ki-67 index in patients aged ≤55
Ki 67 index (> 14.9 vs. ≤ 14.9) 1.39 0.51–3.85 0.52
SLN (positive vs. negative) 2.96 1.15–7.62 0.02
years were predictors of LVI positivity. Considering previous reports
Tumor ADC value (≤0.862 vs. > 0.862) 16.0 1.89–136 0.01 that showed tumor ADC value did not correlate with LN status [11,9],
(×10−3 mm2/s) our results therefore infer predicting LVI status with measurements of
Peritumor-tumor ADC ratio (> 2.372 vs. ≤ 2.372) 1.48 0.59–3.72 0.41 tumor ADC value can be useful for prognostic prediction. Mori et al.
reported that there was a significant difference between the tumor ADC
Abbreviations: OR: odds ratio; CI: confidence interval; SLN: sentinel lymph
values of LVI-positive and -negative postmenopausal patients [12],
node; ADC: apparent diffusion coefficient.
while in our study, they were not significantly different in patients
aged > 55 years. Considering the correlation of ADC with Ki-67 index
Table 4 [17,18], our findings, which showed that tumor ADC value and Ki-67
The diagnostic performance of the pathological size of the invasive component index in patients aged > 55 years were not significantly different, are
and the tumor ADC value. theoretically consistent.
Sensitivity Specificity PPV NPV Accuracy Peritumoral ADC values of patients aged ≤55 years showed a sig-
nificant difference between patients with and without LVI. However, no
Pathological size of the 61 74 49 82 70 similar results were identified in patients of all ages or in those
invasive component
aged > 55 years. Mori et al. reported that the peritumoral ADC values
(> 2 cm vs. ≤ 2 cm)
Tumor ADC value (≤0.862 98 30 37 97 50 of all patients and of postmenopausal patients showed a significant
vs. > 0.862) difference regarding the presence of LVI [12]. It is conceivable that the
(×10−3 mm2/s) reason our results were different from those of Mori et al. is that we did
not exclude cases with scattered and fatty breast tissue surrounding the
Abbreviations: PPV: positive predictive value; NPV: negative predictive values;
tumor. Although ROIs were placed in the peritumoral breast par-
ADC: apparent diffusion coefficient. Data are expressed as percentages.
enchyma and avoided fat to the extent possible, they might have con-
tained fat components. Mori et al. excluded patients who had no breast

116

Downloaded for A. Muh. Yasser Mukti (andimuh_yassermukti@dr.com) at University of Muslim Indonesia from ClinicalKey.com by Elsevier on July 22, 2019.
For personal use only. No other uses without permission. Copyright ©2019. Elsevier Inc. All rights reserved.
T. Igarashi et al.
parenchyma surrounding the tumor. Accordingly, measurements of the
peritumoral ADC value might have affected the results of our study.
With aging, the breast parenchyma is replaced with fatty tissue [21]. It
has been reported that postmenopausal breast tissue has a significantly
lower ADC than premenopausal tissue [22]. Age-related fatty replace-
ment of breast tissue is associated with decreasing serum estradiol and
progesterone. Hence, ADC values are likely to be reduced due to a re-
duction in the speed of osmosis as breast tissue becomes fatty in post-
menopausal patients. Although the value of the peritumoral-tumoral
ADC ratio for the prediction of LVI and peritumoral edema as a prog-
nostic factor associated with disease recurrence were reported [12,23],
method limitations should be considered when assessing the status of
the peritumoral region in breast cancer patients with scattered or fatty
breast tissue.
Several limitations of our study must be considered. First, this ret-
rospective study included several special histological types. Hence, it is
difficult to compare our results with previous reports that included only
luminal type cancers [17,18]. An additional, prospective study is ne-
cessary to confirm our results. Second, we did not assess correlation of
measurement values for the size of the invasive component between
MRI and histopathological results. Further studies are needed to in-
vestigate whether radiological size measurement could be a predictor of
LVI status, especially in the case of a mass forming lesion, which is
relatively easy to measure. This may be less feasible in the case of a
non-mass lesion, as it is difficult to measure the size of the invasive
component. Third, it is possible that the peritumoral ADC and ADC
values of the contralateral breast parenchyma did not accurately reflect
the true conditions, because the ROIs of some cases may have contained
fat components. Fourth, we included patients with mucinous carci-
noma. Since the ADC value of mucinous carcinoma is different from
that of ordinary invasive ductal carcinoma, it may have been appro-
priate to consider excluding those cases. Fifth, we did not consider fi-
brocystic breast changes. Previous studies have reported that the ADC
value of normal fibroglandular tissues was not significantly affected by
the menstrual cycle [22]. However, the effects of fibrocystic changes on
the ADC values of normal breast tissue have not been clarified. The
pathophysiology of fibrocystic breast change is determined by estrogen
predominance and progesterone deficiency, which results in the hy-
perproliferation of connective tissue, followed by facultative epithelial
proliferation [24]. Izumori et al. reported that the thickness of the
normal breast stroma surrounding lobules and ducts was not always
uniform [25], and we speculate that this non-uniformity could be af-
fected by fibrocystic breast changes. As a result, fibrocystic breast
changes might affect the measurement of ADC values of the breast
parenchyma, especially in patients aged ≤55 years. Sixth, we could not
investigate the history of hormone replacement therapy (HRT) of all
patients at any time in their lives; this should be considered because
normal fibroglandular tissues could be affected by HRT.
5. Conclusions
We found that the tumor ADC value, peritumoral ADC value, and
peritumor-tumor ADC ratio were predictive MRI findings for LVI in
patients aged ≤55. The tumor ADC value showed a significant asso-
ciation with LVI, and the inter-observer agreement for tumor ADC value
was substantial between two blinded observers with differences in in-
terpretation experience. Our study suggests that, even when breast
cancer patients show clinically negative nodes, if positive LVI is sus-
pected on preoperative MRI, a more aggressive management strategy
should be considered; this might include the addition of postoperative
adjuvant chemotherapy and/or radiation therapy.
Conflicts of interest
None.
117
Downloaded for A. Muh. Yasser Mukti (andimuh_yassermukti@dr.com) at University of Muslim Indonesia from ClinicalKey.com by Elsevier on July 22, 2019.
For personal use only. No other uses without permission. Copyright ©2019. Elsevier Inc. All rights reserved.
European Journal of Radiology 107 (2018) 111–118
Funding sources
This research received no specific grant from any funding agency in
the public, commercial, or not-for-profit sectors.
References
[1] J. Rosai, L. Sobin (Eds.), Tumors of the Mammary Gland, Armed Forces Institute of
Pathology, Washington, DC, 1993.
[2] A. Goldhirsch, J.H. Glick, R.D. Gelber, A.S. Coates, B. Thurlimann, H. Senn, Panel
members, Meeting highlights: international expert consensus on the primary
therapy of early breast cancer 2005, Ann. Oncol. 16 (2005) 1569–1583.
[3] I. de Mascarel, F. Bonichon, M. Durand, L. Mauriac, G. MacGrogan, I. Soubeyran,
V. Picot, A. Avril, J.M. Coindre, M. Trojani, Obvious peritumoral emboli: an elusive
prognostic factor reappraised. Multivariate analysis of 1320 node-negative breast
cancers, Eur. J. Cancer 34 (1998) 58–65.
[4] R.A. Mohammed, S.G. Martin, M.S. Gill, A.R. Green, E.C. Paish, I.O. Ellis, Improved
methods of detection of lymphovascular invasion demonstrate that it is the pre-
dominant method of vascular invasion in breast cancer and has important clinical
consequences, Am. J. Surg. Pathol. 31 (2007) 1825–1833.
[5] E.A. Rakha, S. Martin, A.H. Lee, D. Morgan, P.D. Pharoah, Z. Hodi, D. Macmillan,
I.O. Ellis, The prognostic significance of lymphovascular invasion in invasive breast
carcinoma, Cancer 118 (2012) 3670–3680.
[6] C. Gajdos, P.I. Tartter, I.J. Bleiweiss, Lymphatic invasion, tumor size, and age are
independent predictors of axillary lymph node metastases in women with T1 breast
cancers, Ann. Surg. 230 (1999) 692–696.
[7] A.H. Lee, S.E. Pinder, R.D. Macmillan, M. Mitchell, I.O. Ellis, C.W. Elston,
R.W. Blamey, Prognostic value of lymphovascular invasion in women with lymph
node negative invasive breast carcinoma, Eur. J. Cancer 42 (2006) 357–362.
[8] L. Martincich, V. Deantoni, I. Bertotto, S. Redana, F. Kubatzki, I. Sarotto, V. Rossi,
M. Liotti, R. Ponzone, M. Aglietta, D. Regge, F. Montemurro, Correlations between
diffusion-weighted imaging and breast cancer biomarkers, Eur. Radiol. 22 (2012)
1519–1528.
[9] S.H. Park, H.Y. Choi, S.Y. Hahn, Correlations between apparent diffusion coefficient
values of invasive ductal carcinoma and pathologic factors on diffusion-weighted
MRI at 3.0 Tesla, J. Magn. Reson. Imaging 41 (2015) 175–182.
[10] S.B. Thakur, M. Durando, S. Milans, G.Y. Cho, L. Gennaro, E.J. Sutton, D. Giri,
E.A. Morris, Apparent diffusion coefficient in estrogen receptor-positive and lymph
node-negative invasive breast cancers at 3.0T DW-MRI: a potential predictor for an
oncotype Dx test recurrence score, J. Magn. Reson. Imaging 47 (2) (2018) 401–409.
[11] M. Durando, L. Gennaro, G.Y. Cho, D.D. Giri, M.M. Gnanasigamani, S. Patil,
E.J. Sutton, J.O. Deasy, E.A. Morris, S.B. Thakur, Quantitative apparent diffusion
coefficient measurement obtained by 3.0 Tesla MRI as a potential noninvasive
marker of tumor aggressiveness in breast cancer, Eur. J. Radiol. 85 (2016)
1651–1658.
[12] N. Mori, S. Mugikura, C. Takasawa, M. Miyashita, A. Shimauchi, H. Ota, T. Ishida,
A. Kasajima, K. Takase, T. Kodama, S. Takahashi, Peritumoral apparent diffusion
coefficients for prediction of lymphovascular invasion in clinically node-negative
invasive breast cancer, Eur. Radiol. 26 (2016) 331–339.
[13] H. Abe, R.A. Schmidt, K. Kulkarni, C.A. Sennett, J.S. Mueller, G.M. Newstead,
Axillary lymph nodes suspicious for breast cancer metastasis: sampling with US-
guided 14-gauge core-needle biopsy—clinical experience in 100 patients, Radiology
250 (2009) 41–49.
[14] H. Yaziji, C.R. Taylor, N.S. Goldstein, D.J. Dabbs, E.H. Hammond, B. Hewlett,
A.D. Floyd, T.S. Barry, A.W. Martin, S. Badve, F. Baehner, R.W. Cartun, R.N. Eisen,
P.E. Swanson, S.M. Hewitt, M. Vyberg, D.G. Hicks, Members of the Standardization
Ad-Hoc Consensus Committee, Consensus recommendations on estrogen receptor
testing in breast cancer by immunohistochemistry, Appl. Immunohistochem. Mol.
Morphol. 16 (2008) 513–520.
[15] M.C. Cheang, S.K. Chia, D. Voduc, D. Gao, S. Leung, J. Snider, M. Watson, S. Davies,
P.S. Bernard, J.S. Parker, C.M. Perou, M.J. Ellis, T.O. Nielsen, Ki67 index, HER2
status, and prognosis of patients with luminal B breast cancer, J. Natl. Cancer Inst.
101 (2009) 736–750.
[16] W. Yang, J.M. Qiang, H.P. Tian, B. Chen, A.J. Wang, J.G. Zhao, Minimum apparent
diffusion coefficient for predicting lymphovascular invasion in invasive cervical
cancer, J. Magn. Reson. Imaging 45 (2017) 1771–1779.
[17] N. Mori, H. Ota, S. Mugikura, C. Takasawa, T. Ishida, G. Watanabe, H. Tada,
M. Watanabe, K. Takase, S. Takahashi, Luminal-type breast cancer: correlation of
apparent diffusion coefficients with the Ki-67 labeling index, Radiology 274 (2015)
66–73.
[18] H.J. Shin, S.H. Kim, H.J. Lee, G. Gong, S. Baek, E.Y. Chae, W.J. Choi, J.H. Cha,
H.H. Kim, Tumor apparent diffusion coefficient as an imaging biomarker to predict
tumor aggressiveness in patients with estrogen receptor-positive breast cancer,
NMR Biomed. 29 (2016) 1070–1078.
[19] J. Jacquemier, E. Charafe-Jauffret, F. Monville, B. Esterni, J.M. Extra,
G. Houvenaeghel, L. Xerri, F. Bertucci, D. Birnbaum, Association of GATA3, P53,
Ki67 status and vascular peritumoral invasion are strongly prognostic in luminal
breast cancer, Breast Cancer Res. 11 (2009) R23.
[20] E.C. Inwald, M. Klinkhammer-Schalke, F. Hofstädter, F. Zeman, M. Koller,
M. Gerstenhauer, O. Ortmann, Ki-67 is a prognostic parameter in breast cancer
patients: results of a large population-based cohort of a cancer registry, Breast
Cancer Res. Treat. 139 (2013) 539–552.
[21] J.N. Wolfe, Breast parenchymal patterns and their changes with age, Radiology
(1976) 545–552.
T. Igarashi et al. European Journal of Radiology 107 (2018) 111–118

[22] J.Y. Kim, H.B. Suh, H.J. Kang, J.K. Shin, K.S. Choo, K.J. Nam, S.W. Lee, Y.L. Jung,
Y.T. Bae, Apparent diffusion coefficient of breast cancer and normal fibroglandular
tissue in diffusion-weighted imaging: the effects of menstrual cycle and menopausal
status, Breast Cancer Res. Treat. 157 (2016) 31–40.
[23] H. Cheon, H.J. Kim, T.H. Kim, H.K. Ryeom, J. Lee, G.C. Kim, J.S. Yuk, W.H. Kim,
Invasive breast cancer: prognostic value of peritumoral edema identified at pre-
operative MR imaging, Radiology 9 (2018) 171157.
[24] H. Vorherr, Fibrocystic breast disease: pathophysiology, pathomorphology, clinical
picture, and management, Am. J. Obstet. Gynecol. 154 (1986) 161–179.
[25] A. Izumori, R. Horii, F. Akiyama, T. Iwase, Proposal of a novel method for observing
the breast by high-resolution ultrasound imaging: understanding the normal breast
structure and its application in an observational method for detecting deviations,
Breast Cancer 20 (2013) 83–91.

118

Downloaded for A. Muh. Yasser Mukti (andimuh_yassermukti@dr.com) at University of Muslim Indonesia from ClinicalKey.com by Elsevier on July 22, 2019.
For personal use only. No other uses without permission. Copyright ©2019. Elsevier Inc. All rights reserved.