Download View/Download PDF

ORIGINAL ARTICLE
Improved Diagnostic Accuracy With Multiparametric Magnetic
Resonance Imaging of the Breast Using Dynamic
Contrast-Enhanced Magnetic Resonance Imaging,
Diffusion-Weighted Imaging, and 3-Dimensional Proton
Magnetic Resonance Spectroscopic Imaging
Katja Pinker, MD,* Wolfgang Bogner, PhD,Þ Pascal Baltzer, MD,* Stephan Gruber, PhD,Þ
Hubert Bickel, MD,* Benedikt Brueck, MD,* Siegfried Trattnig, MD,Þ Michael Weber, PhD,*
Peter Dubsky, MD,þ Zsuzsanna Bago-Horvath, MD,§ Rupert Bartsch, MD,|| and Thomas H. Helbich, MD*
Introduction: The purpose of this study was to compare the diagnostic accuracy
of dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) as a
single parameter to multiparametric (MP) MRI with 2 (DCE MRI and diffusionweighted imaging [DWI]) and 3 (DCE MRI, DWI, and 3-dimensional proton
magnetic resonance spectroscopic imaging [3D 1H-MRSI]) parameters in breast
cancer diagnosis.
Materials and Methods: This prospective study was approved by the institutional review board. Written informed consent was obtained in all patients.
One hundred thirteen female patients (mean age, 52 years; range, 22Y86 years)
with an imaging abnormality (Breast Imaging Reporting and Data System 0,
4Y5) were included in this study. Multiparametric MRI of the breast at 3 T with
DCE MRI, DWI, and 3D 1H-MRSI was performed. The likelihood of malignancy was assessed for DCE MRI and MP MRI with 2 (DCE MRI and DWI)
and 3 (DCE MRI, DWI, and 3D 1H-MRSI) parameters separately. Histopathology was used as the standard of reference. Appropriate statistical tests
were used to assess sensitivity, specificity, and diagnostic accuracy for each
assessment combination.
Results: There were 74 malignant and 39 benign breast lesions. Multiparametric MRI with 3 MRI parameters yielded significantly higher areas under the
curve (0.936) in comparison with DCE MRI alone (0.814) (P G 0.001).
Multiparametric MRI with just 2 parameters at 3 T did not yield higher areas
under the curve (0.808) than did DCE MRI alone (0.814). Multiparametric MRI
with 3 parameters resulted in elimination of false-negative lesions and significantly reduced the false-positives ones (P = 0.002).
Conclusions: Multiparametric MRI with 3 parameters increases the diagnostic accuracy of breast cancer in comparison with DCE-MRI alone and MP
MRI with 2 parameters.
Key Words: breast cancer diagnosis, multiparametric MRI, dynamic
contrast-enhanced magnetic resonance imaging, diffusion-weighted imaging,
3D proton MR spectroscopic imaging
(Invest Radiol 2014;49: 421Y430)
D
ynamic contrast-enhanced (DCE) magnetic resonance imaging
(MRI) of the breast has been the mainstay of breast MRI with an
Received for publication October 9, 2013; and accepted for publication, after revision,
November 25, 2013.
From the Departments of *Biomedical Imaging and Image-guided Therapy, Division of
Molecular and Gender Imaging, †Biomedical Imaging and Image-guided Therapy,
MR Centre of Excellence, ‡Surgery, §Pathology, and ||Internal Medicine, Division
of Oncology, Medical University of Vienna, Vienna, Austria.
Conflicts of interest and sources of funding: Supported by the Austrian National
Bank Jubiläumsfond Project nos. 13652 and 15082.
The authors report no conflicts of interest.
Reprints: Thomas H. Helbich, MD, Department of Biomedical Imaging and Imageguided Therapy, Division of Molecular and Gender Imaging, Medical University
of Vienna, Waehringer Guertel 18-20, 1090 Vienna, Austria. E-mail: thomas.helbich@
meduniwien.ac.at.
Copyright * 2014 by Lippincott Williams & Wilkins
ISSN: 0020-9996/14/4906Y0421
Investigative Radiology
& Volume 49, Number 6, June 2014
excellent sensitivity.1Y5 However, it is limited by a low specificity.6,7
Dynamic contrast-enhanced MRI of the breast provides morphologic
information and limited functional information regarding angiogenesis and tumor vascularity. Available data suggest that the use of
other functional MRI parameters such as diffusion-weighted imaging
(DWI) with quantitative apparent diffusion coefficient (ADC) mapping and proton magnetic resonance (MR) spectroscopic imaging
(1H-MRSI) may provide additional specificity.8Y15 Diffusion-weighted
imaging reflects the diffusivity of water molecules in tissue. Typically,
the movement of water is restricted in malignant cells.16,17 Proton MR
spectroscopic imaging reflects tumor metabolism. Choline (Cho), a
marker of cell turnover, is increased in breast cancer.12,18Y20
To date, research groups have investigated multiparametric
(MP) MRI of the breast using combinations of DCE-MRI with one
additional functional MRI parameter, for example, DWI or 1H-MRSI.
The results obtained with MP MRI of the breast using 2 parameters
were promising.9Y13,21,22 However, to our knowledge, the diagnostic
accuracy of a combination of 3 parameters (DCE MRI, DWI, and 3dimensional [3D] 1H-MRSI) has not been performed.
The number of parameters necessary to significantly improve
the accuracy of breast cancer diagnosis is unknown. In this study, we
aimed to assess the diagnostic accuracy of 3 parameters (DCE MRI,
DWI, and 3D 1H-MRSI) in comparison with dual-parameter (DCE
MRI and DWI) and single-parameter DCE MRI.
MATERIALS AND METHODS
Patients
This prospective; single-institution study was approved by the
institutional review board of the Medical University of Vienna. Written
informed consent was obtained from all patients before their MRI examination. From September 2008 to December 2012, a total of 133
consecutive female patients were enrolled in this study. They underwent
MP MRI of the breast with DCE-MRI, DWI, and 3D 1H-MRSI.
The inclusion criteria were as follows: age of 18 years or older
and an abnormality at mammography or breast ultrasound (asymmetric
density, architectural distortion, suspicious microcalcifications, or
breast mass; Breast Imaging Reporting and Data System (BI-RADS)
category 0 or 4Y5). The exclusion criteria were pregnancy, lactation,
prior treatment (eg, breast biopsy before MRI, neoadjuvant chemotherapy), and contraindications to MRI or contrast agents.1
In addition to the previously mentioned criteria, 12 patients
were excluded because of the technical failure of 3D 1H-MRSI (frequency shift 91 ppm resulting in insufficient water and/or fat suppression) and 8 patients were excluded because of insufficient data
quality because of motion artifacts, resulting in a total of 113 (mean
age, 52 years; range, 22Y86 years) patients to be included in the study.
Regardless of the results of MP MRI, histopathological verification of
the lesion in question was performed. The initial BI-RADS category
www.investigativeradiology.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
421
Investigative Radiology
Pinker et al
distributions of the lesions before MRI were the following: BI-RADS
0 for 43, BI-RADS 4 for 38, and BI-RADS 5 for 32 lesions.
MR Imaging
All patients underwent MP MRI of the breast. All MR examinations were performed in the prone position using a 3-T MR scanner
(Tim Trio; Siemens, Erlangen, Germany) and a dedicated 4-channel
breast coil (InVivo, Orlando, FL). In premenopausal women, MRI
was performed in the second week of the menstrual cycle.1 The MRI
sequence protocol was acquired in the following order:
a) T2-weighted turbo spin echo sequence with fat suppression (repetition
time [TR]/echo time [TE]/inversion time [TI], 4800/61/230
milliseconds; field of view (FOV), 340 mm; 34 slices at 4 mm;
matrix, 314 320; 1 average; time of acquisition (TA) 2
minutes 26 seconds);
b) Diffusion-weighted, double-refocused, single-shot echo-planar imaging with inversion recovery fat suppression (TR/TE/inversion
time, 13700/83/220 milliseconds; FOV, 340 117 mm; 40 slices
at 3.5 mm; matrix 192 64 [50% oversampling]; 2 averages;
b values, 50 and 850 s/mm2; TA, 3 minutes 19 seconds)8;
c) Point-resolved spectroscopic sequence with spectral water and fat
suppression and spatial outer volume suppression (TR/TE, 750/
145 milliseconds; FOV, 12 12 12 cm3; matrix, 12 12 12; interpolated voxel size, 7.5 7.5 7.5 mm3; 5 averages;
TA, 11 minutes 17 seconds) before the application of contrast
agent application to avoid an influence of the contrast agent on
the Cho signal12;
d) A split dynamics protocol combining high-spatial and high-temporal
resolution was used. Parameters applied were as follows: T1weighted volume interpolated breathhold examination sequences
(TR/TE, 3.61/1.4 milliseconds; FOV, 320 mm; 72 slices; 1.7-mm
isotropic; matrix, 192 192; 1 average; 13.2 seconds per volume)
with a total TA of 15 minutes 20 seconds and T1-weighted turbo
fast low-angle shot 3D sequences with selective water excitation
(TR/TE, 877/3.82 milliseconds; FOV, 320 mm; 96 slices; 1-mm
isotropic; matrix, 320 134; 1 average; 2 minutes).23
All patients were injected with a single dose of the contrast
agent. Gadoterate meglumine (Dotarem; Guerbet, Roissy, France) was
injected intravenously as a bolus (0.1 mmol/kg of body weight) at
4 mL/s, followed by a 20-mL saline flush using a power injector (Spectris
Solaris EP; Medrad, Pittsburgh, PA). Application of the contrast agent
was started at 75 seconds after starting the first coronal T1-weighted
volume interpolated breathhold examination. The total examination
time for the MP MRI protocol was approximately 34 minutes.
Data Analysis
Multiparametric MRI data were prospectively evaluated by 2
experienced breast radiologists (K.P., 7 years of experience in breast
MRI; T.H., 16 years of experience in breast MRI) in consensus as
previous studies have shown very good intraobserver and interobserver
agreement by using the following evaluation of each MRI parameter.8,14 The readers were aware that the patients had a breast lesion, but
they were not provided with the previous mammographic and sonographic imaging data or histopathological results.
MRI Parameters
DCE-MRI of the Breast
To distinguish between benign and malignant contrastenhancing lesions descriptors according to the American College of
Radiology, BI-RADS lexicon was used.24,25 Lesions were classified as
masses or nonYmass like enhancing lesions (NMLEs). According to the
American College of Radiology MRI BI-RADS lexicon, the following
descriptors were assessed for masses: shape (round, oval, lobulated,
422
www.investigativeradiology.com
& Volume 49, Number 6, June 2014
irregular), margin (smooth, irregular, spiculated), enhancement pattern
(homogenous, heterogeneous), and enhancement kinetics (persistent
enhancement [type I], initial strong enhancement and plateau phase
[type II], initial strong enhancement and wash-out [type III]).26
For NMLE, the distribution (focal, regional, mutiple regions;
segmental, ductal, linear, and diffuse), the pattern of enhancement
(homogenous, heterogeneous, clumped, and stippled), and the symmetry were assessed.
Lesion enhancement kinetics were used to assess functional
information. Regions of interest (ROIs) were manually drawn in the
most enhancing part of a lesion, and intensity courses were plotted
against time. For NMLE, the enhancements kinetics were not taken
into account. The probability of malignancy was determined by
assigning a final BI-RADS category.6,7,24 In addition, the lesion size
measuring the largest diameter in 1 plane was assessed.
Diffusion-Weighted Imaging
When assessing water diffusivity, highYb value (ie, 850 s/mm2)
images were visually evaluated for hyperintense regions that corresponded to enhancing lesions on DCE MRI. Three-dimensional ROIs
were drawn manually on ADC maps in all lesions. Partial volume effects due to normal parenchyma at the border of the lesion and areas of
necrotic tissue identified in the morphological and contrast-enhanced
images were avoided. The ROIs were defined as slightly smaller than
the actual lesions to avoid partial-volume effects.
Bogner et al8 evaluated the diagnostic quality of DWI regarding
apparent ADC accuracy. On the basis of receiver operating characteristic (ROC) curves, the optimal ADC threshold of 1.25 10j3 mm2/s
for the differentiation between benign and malignant breast lesions was
determined. This ADC threshold was applied in the current study. Lesions were classified as benign if ADC values were equal to or greater
than 1.25 10j3 mm2/s and malignant if less than 1.25 10j3 mm2/s.
3D 1H-MRSI
An experienced spectroscopist (S.G., more than 10 years of
experience) evaluated the 3D 1H-MRSI data. For the assessment of
metabolic information, all 3D 1H-MRSI voxels in the lesion volume
as detected with DCE MRI were evaluated for elevated levels of Cho.
The voxel with the maximum Cho signal-to-noise ratio (SNR) was
determined. Gruber et al12 evaluated the diagnostic accuracy of 3D
1
H-MRSI for the differentiation of benign and malignant breast lesions, on the basis of Cho SNR threshold levels, defining a threshold
of 2.6. Choline SNR measurements are a commonly accepted approach in MR spectroscopy of the breast.20 This Cho SNR threshold
level was applied in the current study. A lesion was classified as
malignant if SNR was equal to or greater than 2.6 and benign if SNR
was lower than 2.6.
MP MRI With 2 Parameters
Dynamic Contrast-Enhanced MRI and DWI
Multiparametric MRI with DCE MRI and DWI was considered
positive if one or more MRI parameter was indicative of malignancy.
Dynamic Contrast-Enhanced MRI and 3D 1H-MRSI
Multiparametric MRI with DCE MRI and 3D 1H-MRSI was
considered positive if one or more MR imaging parameter was indicative of malignancy.
DWI and 3D 1H-MRSI
In the current study, interpretation of DWI and 3D 1H-MRSI
data was always performed after identification of lesions on DCE MR
images. Therefore, these 2 parameters were not assessed individually
or in combination without the information provided by DCE MRI.
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Investigative Radiology
& Volume 49, Number 6, June 2014
MP MRI With 3 Parameters
For the assessment of MP MRI with all 3 parameters (DCE MRI,
DWI, and 3D 1H-MRSI), the following reading scheme was used:
If the results of DCE MRI, DWI, and 1H-MRSI of the breast
were positive, MP MRI was considered positive for malignancy
(Fig. 1).
If the results of DCE MRI, DWI, and 1H-MRSI of the breast
were negative, MP MRI was considered negative for malignancy.
Diagnostic Accuracy of Multiparametric Breast MRI
If 2 of 3 parameters were positive, MP MRI was considered
positive for malignancy (Figs. 2, 3). If 2 of the 3 parameters were
negative, MP MRI was considered negative for malignancy.
Histopathology
The final diagnosis was established through histopathology.
All cases were read by 1 pathologist (Z.B-H, 6 years of experience in
breast pathology).
FIGURE 1. Invasive ductal carcinoma G3 in a 67-year-old woman retroareolar in the left breast. A to D, The indistinct
irregular mass lesion (large arrow) demonstrates heterogeneous initial strong enhancement followed by a wash-out and is classified
as BI-RADS 5 (suspicious finding) through DCE MRI of the breast. Note the clumped NMLE in the anterior aspect of the lesions
representing an extensive intraductal component (small arrow) (E). The mass demonstrates low ADC values (0.988 10j3 mm2/s).
F, In 3D 1H-MRSI, the lesion presents with a Cho peak at 3.23 ppm (SNR, 7.45). Multiparametric MRI with 3 parameters
accurately classifies the tumor as malignant.
* 2014 Lippincott Williams & Wilkins
www.investigativeradiology.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
423
Investigative Radiology
Pinker et al
& Volume 49, Number 6, June 2014
FIGURE 2. Melanoma metastasis in a 39-year-old woman in the left breast retroareolar. A to D, The oval smooth mass (arrow) shows
a persistent homogenous enhancement and is classified through DCE MRI of the breast as BI-RADS 2 (benign). E, The lesion
demonstrates decreased ADC values (0.646 10j3 mm2/s). F, In 3D 1H-MRSI, elevated Cho is observed at 3.23 ppm (SNR, 10.2).
Multiparametric MRI with 3 parameters accurately classifies the tumor as malignant.
All patients underwent image-guided needle biopsy,27 surgical
biopsy, mastectomy, or lumpectomy. In case of a benign histopathological diagnosis at image-guided needle biopsy, the final diagnosis
was benign (n = 31). In case of discordant findings between histopathological diagnosis and imaging findings, the final diagnosis was
established with open surgery (n = 4). In case of a high-risk lesion,
which has an uncertain potential for malignancy, the final diagnosis
was established with open surgery (n = 4).27,28
Statistical Methods
Statistical analysis was performed by a statistician (M.W.),
using Statistical Package for the Social Sciences 19.0 and CIA 2.2.0.
424
www.investigativeradiology.com
All calculations were performed on a per-lesion basis. To calculate
the sensitivity and specificity of DCE MRI of the breast, the assigned
BI-RADS classifications for MRI data were dichotomized. Magnetic
resonance imaging BI-RADS 1, 2, and 3 were considered as benign.
Magnetic resonance imaging BI-RADS 4 and 5 were considered as
malignant. Sensitivity, specificity, accuracy, negative predictive value
(NPV), positive predictive value (PPV), and their 95% confidence
intervals for DCE MRI of the breast and MP MRI with 2 and 3 parameters were calculated.
Histopathology was used as the criterion standard. Significant
differences in sensitivity, specificity, NPV, PPV, and diagnostic accuracy were assessed through logistic regression for multiple measures
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Investigative Radiology
& Volume 49, Number 6, June 2014
Diagnostic Accuracy of Multiparametric Breast MRI
FIGURE 3. Fibroadenoma in a 48-year-old woman in the 2-o’clock position of the right breast. A to D, The slightly irregular
partly indistinct mass (arrow) with a heterogeneous initial strong enhancement and plateau curve is classified through DCE MRI of
the breast as BI-RADS 4 (probably malignant). E, The mass lesion shows high ADC values (1.55 10j3 mm2/s). F, There is no
Cho peak depicted at 3.23 ppm in 3D 1H-MRSI. The lesion is false positive in DCE MRI but true negative in DWI and 3D 1H-MRSI.
Therefore, MP MRI with 3 parameters accurately classifies the tumor as benign.
(generalized estimating equations). Receiver operating characteristic
curves were plotted, and the area under the curve (AUC) was determined. Statistical differences between the AUCs were assessed using
the method proposed by DeLong et al.29 A P value of less than or equal
to 0.05 was considered a significant result.
RESULTS
There were 113 lesions with a lesion size ranging from 5 to 98 mm
(mean, 29 mm). Histopathologic diagnosis revealed 74 lesions to be
* 2014 Lippincott Williams & Wilkins
malignant and 39 to be benign (Table 1). There were 98 enhancing
masses (size range, 5Y98 mm; mean, 28.1 mm) and 15 NMLE (size
range, 5Y89 mm; mean, 36.8 mm) at DCE MRI of the breast.
Dynamic contrast-enhanced MRI classified 87 lesions as malignant (MRI BI-RADS 4 and 5) and 26 lesions as benign (MRI BIRADS 2 and 3). The benign mass lesions showed type 1 curve in 15
(46.9%) of 32, type 2 curve in 13 (40.6%) of 32, and type 3 curve in 4
(12.5%) of 32. Curve types in the malignant mass lesions were distributed as follows: type 1 in 14 (21.2%) of 66, type 2 in 22 (33.3%) of
66, and type 3 in 30 (45.5%) of 66.
www.investigativeradiology.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
425
Investigative Radiology
Pinker et al
TABLE 1. Detailed Histopathological Diagnoses of all Malignant
(n = 74) and Benign (n = 39) Breast Lesions
Malignant
n
%
74
65.5
Histopathological subtype
IDC
ILC
Mucinous carcinoma
Papillary carcinoma
Phylloid high grade
Metastasis
DCIS
52
11
2
1
1
1
6
46
9.7
1.8
0.9
0.9
0.9
5.3
Benign
n
%
39
3
34.5
2.7
18
2
9
7
15.9
1.8
7.9
6.2
& Volume 49, Number 6, June 2014
Multiparametric MRI with 2 parameters, that is, DCE MRI
and DWI or 3D 1H-MRSI, did not improve diagnostic accuracy
(AUC, 0.808) compared with DCE MRI alone (P = 0.323).
Diagnostic accuracy of MP MRI with 3 parameters was significantly higher compared with MP MRI with 2 parameters (P G 0.001).
Sensitivity, Specificity, NPV, and PPV
High risk (complex sclerosing lesion, FEA, CCC with
focal atypia)
FA/FAH
Papilloma
DH, CCC, FCC, focal fibrosis, nodular sclerosing adenosis
Miscellaneous (chronic abscess, gynecomastia, fat necrosis,
pseudoangiomatosis)
CCC indicates columnar cell changes; DCIS, ductal carcinoma in situ; DH,
ductal hyperplasia; FA, fibroadenoma; FAH, fibroadenomatous hyperplasia;
FCC, fibrocystic changes; FEA, flat epithelial atypia; IDC, invasive ductal
carcinoma; ILC, invasive lobular carcinoma.
Multiparametric MRI with 3 parameters, that is, DCE MRI,
DWI, and 3D 1H-MRSI, yielded the highest sensitivity (100%), which
was maximized compared with DCE MRI with 98.6%. Multiparametric MRI with 3 parameters significantly increased specificity to
87.2% compared with DCE-MRI with 64.1% (P G 0.001). Multiparametric MRI with 2 parameters also maximized sensitivity to 100%
but showed a decrease in specificity of 61.5%.
Positive predictive values of DCE MRI alone and MP MRI with 2
or 3 parameters for breast cancer diagnosis are summarized in Figure 5.
The PPV was significantly higher with MP MRI with 3 parameters at
93.7% compared with DCE MRI alone at 83.9% (P G 0.001). Multiparametric MRI with 2 parameters demonstrated lower PPVs of 83.1%
compared with DCE MRI alone.
The NPV of MP MRI with either 2 or 3 parameters was increased to 100% compared with DCE MRI at 96.2%.
Multiparametric MRI with all 3 parameters allowed the elimination of the false-negative finding and a significant reduction in
false-positive findings from 14 with DCE MRI and 15 with MP MRI
with 2 parameters to 5 with 3 parameters (P = 0.002). Detailed histopathological results for all false-positive and false-negative lesions
for DCE MRI and MP MRI are listed in Table 3.
DISCUSSION
The ADC values of the benign lesions ranged from 0.98 to 2.54 10j3 mm2/s (mean, 1.61 10j3 mm2/s); those of the malignant lesions,
from 0.4 to 1.62 (mean, 0934 10j3 mm2/s).
Choline SNR values of the benign lesions ranged from 0 to 5.6
(mean, 0.89); those of the malignant lesions, from 0 to 56.1(mean, 9.98).
Sensitivities, specificities, PPV, NPV, diagnostic accuracies,
and the AUCs for DCE MRI and MP MRI with 2 or 3 MRI parameters are summarized in Table 2.
Diagnostic Accuracy: ROC Analysis
Diagnostic accuracy was significantly higher with MP MRI
using 3 parameters, that is, DCE MRI, DWI, and 3D 1H-MRSI, with
an AUC of 0.936 compared with DCE MRI alone with an AUC of
0.814 (P G 0.001) (Fig. 4).
The results of our study show that the highest diagnostic accuracy
for breast cancer diagnosis is achieved with MP MRI using 3 parameters,
that is, DCE MRI, DWI, and 3D 1H-MRSI, with an AUC of 0.936
compared with DCE MRI with an AUC of 0.814 (P G 0.001). Multiparametric MRI with 2 parameters does improve diagnostic accuracy in
comparison with DCE-MRI alone.
As demonstrated in our study, each MRI parameter has an incremental value,8,12,23,30 but only MP MRI with 3 parameters allows an
increase in both sensitivity and particularly specificity, resulting in a
significantly improved diagnostic accuracy. With MP MRI using parameters, there were no false negatives and there was a significant reduction in false-positive lesions compared with DCE MRI and MP
MRI with 2 parameters. In addition, MP MRI with 3 parameters has the
potential to eliminate almost two thirds of unnecessary breast biopsies
recommended by DCE MRI alone. The results are in agreement with
TABLE 2. Sensitivities, Specificities, PPV, NPV, Diagnostic Accuracy, AUC, and 95% Confidence Intervals (in Brackets) for DCE-MRI
and MP MRI
Sensitivity
DCE-MRI (1 parameter)
MP MRI with DCE MRI
and DWI (2 parameters)
MP MRI with DCE-MRI and
3D 1H-MRSI (2 parameters)
MP MRI with DCE-MRI, DWI,
and 3D 1H-MRSI
(3 parameters)
98.6%
(92.7%Y99.8%)
100%
(93.6%Y100%)
100%
(93.6%Y100%)
100%
(95.1%Y100%)
Specificity
64.1%
(48.4%Y77.3%)
61.5%
(44.7%Y76.2%)
61.5%
(45.9%Y75.1%)
87.2%*†
(73.3%Y94.4%)
PPV
83.9%
(74.8%Y90.2%)
83.1%
(74%Y89.5%)
83.1%
(74%Y89.5%)
93.7%*†
(86.0%Y97.3%)
NPV
Accuracy
96.2%
(81.1%Y99.3%)
100%
(86.2%Y100%)
100%
(86.2%Y100%)
100%
(89.8%Y100%)
86.7%
(79.2%Y91.8%)
86.7%
(79.2%Y91.8%)
86.7%
(79.2%Y91.8%)
95.6%*†
(90.1%Y98.1%)
AUC
0.814
(0.72Y0.91)
0.808
(0.71Y0.91)
0.808
(0.71Y0.91)
0.936*†
(0.87Y0.99)
*Significantly different from DCE-MRI (P G 0.001).
†Significantly different from MP MRI with 2 MR imaging parameters (P G 0.001).
3D 1H-MRSI indicates 3-dimensional proton spectroscopic imaging; AUC, area under the curve; DCE MRI, dynamic contrast-enhanced magnetic resonance
imaging; DWI, diffusion-weighted imaging; MP, multiparametric; MRI, magnetic resonance imaging; NPV, negative predictive value; PPV, positive predictive value.
426
www.investigativeradiology.com
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Investigative Radiology
& Volume 49, Number 6, June 2014
FIGURE 4. The ROC curves illustrate the higher diagnostic
value (that is, higher sensitivity, specificity, and larger AUC)
that is reached for MP MRI with 3 parameters compared with
DCE MRI of the breast and MP MRI with 2 parameters.
studies that have proven the additional value of MP MRI with more
than 2 parameters in other neoplastic diseases such as brain and prostate
cancer.31Y37 It is evident that MP MRI leads to an increased scan time;
however, the increase in diagnostic accuracy resulting in a reduction of
unnecessary breast biopsies is beneficial for the patient and might even
be cost-effective, although no formal cost-effectiveness analysis was
performed.
Even with the use of MP MRI, there were some false-positive
lesions. These comprised of 3 high-risk lesions at biopsy, which remained lesions with atypia at surgery.38,39 Studies have demonstrated
that high-risk lesions such as atypical ductal hyperplasia are genetically
advanced nonYobligatory precursors of invasive breast cancer.40 The
fact that these lesions remained false positive with MP MRI with all 3
parameters possibly reflects malignant potential. Another false positive
was a very metabolically active juvenile fibroadenoma with a high
cellularity.41,42 The other remaining false-positive lesion was a clinically
Diagnostic Accuracy of Multiparametric Breast MRI
asymptomatic chronic abscess, which demonstrated suspicious morphologic and kinetic features and decreased ADC levels,43,44 but was
negative on 3D 1H-MRSI.
In the past, several studies have evaluated the role of MP MRI
with 2 parameters, that is, DCE MRI and either DWI or 1H-MRSI, in
breast cancer diagnosis.
Kul et al11 investigated the combination of DCE MRI and DWI
in the assessment of breast tumors. The authors reported an increase in
specificity of 86.5% but a decrease in sensitivity of 95.2%. Yabuuchi
et al21,45 and Ei Khouli et al,9 who used a logistic regression model,
reported similar results for MP MRI with DCE MRI and DWI. In all
these studies, the improvement of specificity led to a trade-off in sensitivity. In contrast, the combination approach in this study led to an
increase in sensitivity to 100% but no increase in specificity.
Several groups have evaluated the diagnostic accuracy of 1HMRSI with encouraging results. A recent meta-analysis reported
pooled estimates of sensitivity and specificity of 73 and 88% and
demonstrated that 1H-MRSI has a high specificity for diagnosis of
breast cancer.20 Potentially, this specificity may further increase diagnostic accuracy in breast imaging. However, a practical algorithm
for the combination of DCE MRI and 1H-MRSI has not yet been
proposed.
According to Gruber et al,12 3D 1H-MRSI has a higher sensitivity of 97%. We applied the same 3D 1H-MRSI technique, and
our results are concordant with these findings. It has to be kept in
mind that MR spectroscopy even if performed as a 3D technique is
limited to a certain region of one breast. As a consequence, it can
only be considered as an additional tool for the characterization of
lesions in a prior known location. This limits its application for the
characterization of incidental small and nonYmasslike enhancements.
In the current study, interpretation of DWI and 3D 1H-MRSI
data was always performed after identification of lesions on DCE MR
images. There have been approaches described for lesion identification in DWI on high b-value images,22,46 but no data exist for the
application of 3D 1H-MRSI in breast imaging without lesion identification on DCE MR images. Because lesion identification on DCE
MR images is the most commonly used approach, we chose to use
this in the current study; therefore, DWI and 3D 1H-MRSI were not
assessed individually or in combination.
The data suggest that MP MRI with 2 parameters is not as beneficial as MP MRI with 3 parameters with respect to diagnostic accuracy.
However, there are other approaches for the combination of
quantitative multiparametric data than the empirical approach chosen in this study. Methods such as classification trees, logistic regression, or artificial neural networks have been described. These
methods are able to combine raw quantitative data and assign a probability of malignancy. Although desirable, these approaches are limited
FIGURE 5. Positive predictive values of DCE MRI and MP MRI with either 3 or 2 parameters.
* 2014 Lippincott Williams & Wilkins
www.investigativeradiology.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
427
Investigative Radiology
Pinker et al
& Volume 49, Number 6, June 2014
TABLE 3. Detailed Histopathological Results of False-Negative and False-Positive Lesions
1
False Postives
Mass/NMLE
Size (mm)
n
False
Negatives
Mass/NMLE
Size (mm)
n
DCE-MRI (1 parameter)
FA
FA
FA
FAH
DH, CCC
DH, CCC
DH, CCC
Miscellaneous
Miscellaneous
Miscellaneous
High risk
High risk
High risk
High risk
FA
FA
FA
FAH
DH, CCC
DH, CCC
DH, CCC
DH, CCC
Miscellaneous
Miscellaneous
Miscellaneous
High risk
High risk
High risk
High risk
FA
FA
FA
FA
FAH
DH, CCC
DH, CCC
DH, CCC
Miscellaneous
Miscellaneous
Miscellaneous
High risk
High risk
High risk
High risk
FA
Miscellaneous
High risk
High risk
High risk
Mass
Mass
NMLE
Mass
Mass
NMLE
NMLE
Mass
Mass
NMLE
Mass
Mass
Mass
NMLE
Mass
Mass
NMLE
Mass
Mass
Mass
NMLE
NMLE
Mass
Mass
NMLE
Mass
Mass
Mass
NMLE
Mass
Mass
Mass
NMLE
Mass
Mass
NMLE
NMLE
Mass
Mass
NMLE
Mass
Mass
Mass
NMLE
NMLE
Mass
Mass
Mass
NMLE
20
28
39
10
8
10
20
20
90
19
7
8
10
58
20
28
39
10
8
8
10
20
20
90
19
7
8
10
58
20
28
36
39
10
8
10
20
20
90
19
7
8
10
58
20
90
7
8
50
14
metastasis
1
25
1
MP MRI with DCE-MRI and DWI
(2 parameters)
MP MRI with DCE-MRI and 3D 1H-MRSI
(2 parameters)
MP MRI with DCE-MRI, DWI, and 3D 1H-MRSI
(3 parameters)
15
0
15
0
5
0
3D 1H-MRSI indicates 3-dimensional proton MR spectroscopic imaging, CCC, columnar cell changes; DCE-MRI, dynamic contrast-enhanced magnetic
resonance imaging; DH, ductal hyperplasia; DWI, diffusion-weighted imaging; FA, fibroadenoma; FAH, fibroadenomatous hyperplasia; FCC, fibrocystic changes;
IDC, invasive ductal carcinoma; NMLE, nonYmasslike enhancing lesions.
428
www.investigativeradiology.com
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Investigative Radiology
& Volume 49, Number 6, June 2014
because no internationally accepted reference values for quantitative
MP MRI results do exist. Providing such values in a multiparametric
approach would require case numbers beyond the scope of this study.
To date, because of these limitations, only little data on multivariate data
combinations and no data using MP MRI has been published in breast
MRI.47,48 The approach used in the current study was defined before
commencement of the study, was readily applicable in clinical practice,
and achieved high diagnostic accuracies.
There was a relatively small number of pure ductal carcinoma
in situ and invasive lobular cancers as compared with the number of
invasive ductal cancers. Consequently, there was also a relatively
small number of NMLEs, which often represent ductal carcinoma in
situ and invasive lobular carcinoma. This limits the specific insights
into the performance of MP MRI in these subgroups. However, in a
prospective study with consecutive patients, the collective distribution of malignant lesion subtypes is as expected and underlines the
representativeness of our patient sample. In addition, the readers in
this study were experienced breast radiologists with extensive training in the interpretation of breast MR images; therefore, the results
might not be applicable to every radiologist. However, it can be
expected that dedicated training and the addition of computer-aided
detection systems that implement all MR imaging parameters will
be helpful for radiologists in everyday practice. All MR examinations
were performed at 3 T instead of 1.5 T, which is the most commonly
used field strength in breast MRI. This should not be seen as a limitation because 3 T offers advantages such as an improved temporal
and spatial resolution.6,8,12,14,23
CONCLUSIONS
In conclusion, MP MRI with 3 parameters (DCE-MRI, DWI,
and 3D-1H-MRSI) has a greater diagnostic accuracy for breast cancer
diagnosis than DCE-MRI alone or MP MRI with only 2 MRI
parameters.
ACKNOWLEDGMENTS
The authors thank Dr Oshaani Abeyakoon for helping with the
editing of the manuscript of this article.
REFERENCES
1. Sardanelli F, Boetes C, Borisch B, et al. Magnetic resonance imaging of the
breast: recommendations from the EUSOMA working group. Eur J Cancer.
2010;46:1296Y1316.
2. Kuhl CK. Breast MR imaging at 3T. Magn Reson Imaging Clin N Am. 2007;15:
315Y320.
3. Morris EA. Diagnostic breast MR imaging: current status and future directions.
Magn Reson Imaging Clin N Am. 2007;45:863Y880.
4. Mercado CL, Moy L. Breast MRI. Preface. Magn Reson Imaging Clin N Am.
2010;18(2):xiii. doi: 10.1016/j.mric.2010.02.014.
5. Pediconi F, Miglio E, Telesca M, et al. Effect of preoperative breast magnetic
resonance imaging on surgical decision making and cancer recurrence rates.
Invest Radiol. 2012;47:128Y135.
6. Pinker-Domenig K, Bogner W, Gruber S, et al. High resolution MRI of the
breast at 3 T: which BI-RADS(R) descriptors are most strongly associated with
the diagnosis of breast cancer? Eur Radiol. 2012;22:322Y330.
7. Kuhl CK, Schild HH, Morakkabati N. Dynamic bilateral contrast-enhanced
MR imaging of the breast: trade-off between spatial and temporal resolution.
Radiology. 2005;236:789Y800.
8. Bogner W, Gruber S, Pinker K, et al. Diffusion-weighted MR for differentiation of breast lesions at 3.0 T: how does selection of diffusion protocols
affect diagnosis? Radiology. 2009;253:341Y351.
9. Ei Khouli RH, Jacobs MA, Mezban SD, et al. Diffusion-weighted imaging
improves the diagnostic accuracy of conventional 3.0-T breast MR imaging.
Radiology. 2010;256:64Y73.
10. Partridge SC, DeMartini WB, Kurland BF, et al. Quantitative diffusionweighted imaging as an adjunct to conventional breast MRI for improved
positive predictive value. AJR Am J Roentgenol. 2009;193:1716Y1722.
11. Kul S, Cansu A, Alhan E, et al. Contribution of diffusion-weighted imaging to
dynamic contrast-enhanced MRI in the characterization of breast tumors. AJR
Am J Roentgenol. 2011;196:210Y217.
* 2014 Lippincott Williams & Wilkins
Diagnostic Accuracy of Multiparametric Breast MRI
12. Gruber S, Debski BK, Pinker K, et al. Three-dimensional proton MR spectroscopic imaging at 3 T for the differentiation of benign and malignant breast
lesions. Radiology. 2011;261:752Y761.
13. Bartella L, Morris EA, Dershaw DD, et al. Proton MR spectroscopy with
choline peak as malignancy marker improves positive predictive value for
breast cancer diagnosis: preliminary study. Radiology. 2006;239:686Y692.
14. Bogner W, Pinker-Domenig K, Bickel H, et al. Readout-segmented echo-planar
imaging improves the diagnostic performance of diffusion-weighted MR breast
examinations at 3.0 T. Radiology. 2012;263:64Y76.
15. Eyal E, Shapiro-Feinberg M, Furman-Haran E, et al. Parametric diffusion
tensor imaging of the breast. Invest Radiol. 2012;47:284Y291.
16. Woodhams R, Matsunaga K, Kan S, et al. ADC mapping of benign and malignant breast tumors. Magn Reson Med Sci. 2005;4:35Y42.
17. Chen X, Li WL, Zhang YL, et al. Meta-analysis of quantitative diffusionweighted MR imaging in the differential diagnosis of breast lesions. BMC
Cancer. 2010;10:693.
18. Meisamy S, Bolan PJ, Baker EH, et al. Adding in vivo quantitative 1H MR
spectroscopy to improve diagnostic accuracy of breast MR imaging: preliminary results of observer performance study at 4.0 T. Radiology. 2005;236:
465Y475.
19. Sardanelli F, Fausto A, Di Leo G, et al. In vivo proton MR spectroscopy of the
breast using the total choline peak integral as a marker of malignancy. AJR Am
J Roentgenol. 2009;192:1608Y1617.
20. Baltzer PA, Dietzel M. Breast lesions: diagnosis by using proton mr spectroscopy at 1.5 and 3.0 TVsystematic review and meta-analysis. Radiology.
2013;267:735Y746.
21. Yabuuchi H, Matsuo Y, Okafuji T, et al. Enhanced mass on contrast-enhanced
breast MR imaging: lesion characterization using combination of dynamic
contrast-enhanced and diffusion-weighted MR images. J Magn Reson Imaging. 2008;28:1157Y1165.
22. Yabuuchi H, Matsuo Y, Sunami S, et al. Detection of non-palpable breast
cancer in asymptomatic women by using unenhanced diffusion-weighted and
T2-weighted MR imaging: comparison with mammography and dynamic
contrast-enhanced MR imaging. Eur Radiol. 2011;21:11Y17.
23. Pinker K, Grabner G, Bogner W, et al. A combined high temporal and high
spatial resolution 3 Tesla MR imaging protocol for the assessment of breast
lesions: initial results. Invest Radiol. 2009;44:553Y558.
24. American College of Radiology (ACR). Breast Imaging and Reporting Data
System (BI-RADS). Reston, VA: American College of Radiology; 2003.
25. Tardivon AA, Athanasiou A, Thibault F, et al. Breast imaging and reporting
data system (BIRADS): magnetic resonance imaging. Eur J Radiol. 2007;61:
212Y215.
26. Kuhl CK, Mielcareck P, Klaschik S, et al. Dynamic breast MR imaging: are
signal intensity time course data useful for differential diagnosis of enhancing
lesions? Radiology. 1999;211:101Y110.
27. Wallis M, Tardivon A, Helbich T, et al. Guidelines from the European Society
of Breast Imaging for diagnostic interventional breast procedures. Eur Radiol.
2007;17:581Y588.
28. Krishnamurthy S, Bevers T, Kuerer H, et al. Multidisciplinary considerations in
the management of high-risk breast lesions. AJR Am J Roentgenol. 2012;198:
W132YW140.
29. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two
or more correlated receiver operating characteristic curves: a nonparametric
approach. Biometrics. 1988;44:837Y845.
30. Renz DM, Diekmann F, Schmitzberger FF, et al. Pharmacokinetic approach for
dynamic breast MRI to indicate signal intensity time curves of benign and
malignant lesions by using the tumor flow residence time. Invest Radiol.
2013;48:69Y78.
31. Oliveira Neto JA, Parente DB. Multiparametric magnetic resonance imaging
of the prostate. Magn Reson Imaging Clin N Am. 2013;21:409Y426.
32. Hedgire SS, Oei TN, McDermott S, et al. Multiparametric magnetic resonance
imaging of prostate cancer. Indian J Radiol Imaging. 2012;22:160Y169.
33. Turkbey B, Mani H, Shah V, et al. Multiparametric 3T prostate magnetic resonance imaging to detect cancer: histopathological correlation using prostatectomy specimens processed in customized magnetic resonance imaging
based molds. J Urol. 2011;186:1818Y1824.
34. Franiel T, Stephan C, Erbersdobler A, et al. Areas suspicious for prostate
cancer: MR-guided biopsy in patients with at least one transrectal US-guided
biopsy with a negative findingVmultiparametric MR imaging for detection
and biopsy planning. Radiology. 2011;259:162Y172.
35. Bian W, Khayal IS, Lupo JM, et al. Multiparametric characterization of grade
2 glioma subtypes using magnetic resonance spectroscopic, perfusion, and
diffusion imaging. Transl Oncol. 2009;2:271Y280.
36. Roy B, Gupta RK, Maudsley AA, et al. Utility of multiparametric 3-T MRI for
glioma characterization. Neuroradiology. 2013;55:603Y613.
37. Pinker K, Stadlbauer A, Bogner W, et al. Molecular imaging of cancer: MR
spectroscopy and beyond. Eur J Radiol. 2012;81:566Y577.
38. Kluttig A, Trocchi P, Heinig A, et al. Reliability and validity of needle biopsy
evaluation of breast-abnormalities using the B-categorizationVdesign and
objectives of the Diagnosis Optimisation Study (DIOS). BMC Cancer.
2007;7:100.
www.investigativeradiology.com
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
429
Investigative Radiology
Pinker et al
39. Bianchi S, Caini S, Renne G, et al. Positive predictive value for malignancy on
surgical excision of breast lesions of uncertain malignant potential (B3) diagnosed by stereotactic vacuum-assisted needle core biopsy (VANCB): a large
multi-institutional study in Italy. Breast. 2011;20:264Y270.
40. Riedl CC, Ponhold L, Flory D, et al. Magnetic resonance imaging of the breast
improves detection of invasive cancer, preinvasive cancer, and premalignant
lesions during surveillance of women at high risk for breast cancer. Clin
Cancer Res. 2007;13:6144Y6152.
41. Hochman MG, Orel SG, Powell CM, et al. Fibroadenomas: MR imaging appearances with radiologic-histopathologic correlation. Radiology. 1997;204:
123Y129.
42. Kuijper A, Mommers EC, van der Wall E, et al. Histopathology of fibroadenoma of the breast. Am J Clin Pathol. 2001;115:736Y742.
43. Unal O, Koparan HI, Avcu S, et al. The diagnostic value of diffusion-weighted
magnetic resonance imaging in soft tissue abscesses. Eur J Radiol. 2011;77:
490Y494.
430
www.investigativeradiology.com
& Volume 49, Number 6, June 2014
44. Oto A, Schmid-Tannwald C, Agrawal G, et al. Diffusion-weighted MR imaging
of abdominopelvic abscesses. Emerg Radiol. 2011;18:515Y524.
45. Yabuuchi H, Matsuo Y, Kamitani T, et al. Non-mass-like enhancement on
contrast-enhanced breast MR imaging: lesion characterization using combination of dynamic contrast-enhanced and diffusion-weighted MR images. Eur
J Radiol. 2010;75:e126Ye132.
46. Baltzer PA, Benndorf M, Dietzel M, et al. Sensitivity and specificity of
unenhanced MR mammography (DWI combined with T2-weighted TSE imaging, ueMRM) for the differentiation of mass lesions. Eur Radiol. 2010;20:
1101Y1110.
47. Dietzel M, Baltzer PA, Dietzel A, et al. Artificial neural networks for differential diagnosis of breast lesions in MR-mammography: a systematic approach
addressing the influence of network architecture on diagnostic performance
using a large clinical database. Eur J Radiol. 2012;81:1508Y1513.
48. Baltzer PA, Dietzel M, Kaiser WA. A simple and robust classification tree for
differentiation between benign and malignant lesions in MR-mammography.
Eur Radiol. 2013;23:2051Y2060.
* 2014 Lippincott Williams & Wilkins
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.