1. Introduction
Magnetic resonance imaging (MRI) of the breast
was first performed in the late 1980s. At first, differentiation between benign
and malignant breast lesions was primarily based on their differences in T1 and
T2 relaxations times (Rausch et al., 2006). Due to the large overlap in T1 and
T2 relaxation times in benign and malignant breast lesions, it became apparent
that contrast administration was mandatory for reliable breast MRI. Heywang et
al. demonstrated that breast carcinomas showed significant enhancement within 5
minutes after contrast administration (Heywang et al., 1989).
Since then, increasing field strengths,
dedicated breast coil designs, and
improvements in sequence protocols have led to a large improvement in
diagnostic accuracy of breast MRI. Currently, the sensitivity of
contrast-enhanced MRI for detecting breast cancer reaches 88%, with a
specificity of 68%. The positive predictive value is reported to be 72%, with a
negative predictive value of 85% (Bluemke et al., 2004). The reported sensitivity
and specificity may vary in different publications due to differences in study
populations, and technical and diagnostic criteria used. Reported sensitivities
therefore vary from 83-100%, with reported specificities varying from 29-100%
(Rausch et al., 2006).
These numbers are superior to mammography and
ultrasound, and are independent of factors such as tumor histology, breast
density, and hormonal therapy use. They also show that breast MRI is highly
accurate for detecting breast cancer. However, due to the rather limited
specificity, false-positive results are frequently observed, requiring
additional imaging or (MR guided) biopsy, in turn causing patient anxiety and
discomfort.
In this chapter, the technical aspects and
proper indications of breast MRI are discussed. In addition, a systematic
approach to the image interpretation of breast MRI is proposed.
2.
Performing
magnetic resonance imaging of the breast
2.1 Patient
handling
Before
performing breast MRI, it is important to instruct the patient thoroughly. It
is important to inform the patient that lying comfortly and motionless is
important for succesfull imaging of the breast. They should be instructed that
administration of the contrast agent can result in various physical sensations,
which may cause patient anxiety (and motion) when not properly instructed.
A dedicated breast coil should be used for
breast MRI. These coils usually consist of a multichannel coil (nowadays up to
32-channel) with two loops in which the breasts are placed while the patient is
lying in prone position. The breasts should be placed as deep as possible in
the coil loops, with the nipples pointing downward if possible. To further
reduce motion artefacts, the breasts can be gently fixated using cushions.
Excessive compression should be avoided, as this might influence breast
perfusion, and thus contrast enhancement pharmacokinetics.
In premenopausal women, the enhancement of the
fibroglandular tissue after contrast administration is dependent of the
menstrual cycle. MR imaging of the breast in the wrong phase of the menstrual
cycle can result in strong glandular enhancement, complicating the
interpretation of the images. Elective breast MRI is ideally performed in the
first phase of the menstrual cycle, i.e. days 3-14, with day 1 being the first
day of menstruation (Delille et al., 2005). In patients with proven breast
cancer who undergo breast MRI as part of their preoperative staging, MRI should
be performed at the earliest opportunity. In these cases, rapid presurgical
patient work-up is preferred over optimal MR image quality.
2.2 Technical
aspects
2.2.1 Field
strengths
Increasing field strengths are associated with
increased signal-to-noise (SNR) ratios. In order to acquire sufficient spatial
resolution for accurate assessment of lesion morphology, it is generally
accepted that field strengths of more than 1.5 Tesla are recommended for breast
MRI (Weinstein et al., 2010). Theoretically, a higher field strength (e.g. 3
Tesla) increases the SNR for breast MRI. At a similar temporal resolution, this
increased SNR might be used to increase spatial resolution, and thus improve
lesion morphology evaluation and diagnostic accuracy.
In a proof-of-concept study, Kuhl et al.
compared the accuracy of both 1.5 and 3.0 Tesla breast MRI in the same
patients. Although the study population was small (n=37, total of 53 breast
lesions, both malignant and benign), they demonstrated that the overall image
quality scores for the dynamic contrast-enhanced series were higher
(p<0.01). They also demonstrated that at 3.0 Tesla, the differential
diagnosis of enhancing lesions was possible with a higher diagnostic confidence,
as reflected by a larger area under the ROC-curve (Kuhl et al., 2006).
In another proof-of-concept study by Pinker et
al., contrast-enhanced breast MRI was performed on a 3 Tesla MRI scanner in 34
patients (having 55 breast lesions). Their imaging protocol enabled accurate
detection and assessment of breast lesions, with a sensitivity of 100% (95%
confidence interval 90.6-100.0%. The specificity was 72.2%, with a 95%
confidence interval of 49.1-87.5% (Pinker et al., 2009). Although these preliminary
results are promising, there is no
strong evidence to date of the superiority of 3.0 over 1.5 Tesla breast MR
imaging.
2.2.2 Imaging
planes
In the past, breast MR imaging was usually
performed in a sagittal plane. The advantage of this imaging plane was that a
relatively small field-of-view could be selected to cover the breast, resulting
in an improved spatial resolution. However, simultaneous contralateral breast
cancer can be detected in 3% of the cases (Lehman et al., 2007), indicating
that bilateral breast imaging is strongly recommended. Bilateral sagittal
imaging of the breast can lead to decrease of SNR and spatial resolution (Kuhl,
2007). Therefore, current bilateral imaging protocols use the transverse or
coronal plane. Coronal imaging of the breast tends to give more respiratory
motion artifacts. Also, nipple and chest wall involvement is more difficult to
detect on coronal images. Therefore, the transverse imaging plane is preferred
when bilateral breast imaging is performed (Kuhl, 2007).
2.2.3 Spatial
and temporal resolution
Breast MRI needs to be performed with adequate
spatial resolution in order to assess lesion morphology accurately. It is
widely adopted that an optimal breast MRI should have a minimum size threshold
for detection of lesions of 5 mm. Therefore, a voxel size of at least 2.5 mm in any direction should be used (Mann
et al., 2008). However, higher in-plane spatial resolution results in more
accurate lesion morphology assessment. Therefore, the minimal in-plane spatial
resolution as recommended by the American College of Radiology is < 1 mm (Weinstein et al., 2010).
2.2.4
Temporal
resolution and contrast-enhanced dynamic T1 weighted imaging sequences
Gadolinium (Gd, atom number 64) is a chemical
that belongs to the element category of the lanthanides. Due to it’s paramagnetic properties, it is often used as
an intravenous contrast agent in MRI. However, free Gd-atoms are highly toxic
and as a result, gadolinium-based contrast agents consist of a chelated
Gd-complex to render it non-toxic. Gd-based contrast agents lower T1, T2, and
T2* relaxation times. Since the decrease is highest for T1 relaxation times,
contrast-enhanced MR imaging sequences are mostly T1-weighted.
The contrast agent is administered
intravenously with an automated injector to ensure a continuous inflow of
contrast. Although the optimal dose is unknown, a dose of 0.1-0.2 mmol per
kilogram of body weight and a flow rate of 3 mL/second is generally accepted
(Kuhl, 2007, Rausch et al., 2006). The administration is followed by a saline
flush to ensure complete administration of the dose.
After
intravenous administration, the contrast agent leaks through immature (‘leaky’)
microvessels that were formed by tumor angiogenesis (Carmeliet et al., 2000,
Hashizume et al., 2000, Jansen et al., 2009). As a result, breast lesions tend
to demonstrate a peak enhancement between 90-120 seconds. In order to assess
the pharmacokinetic enhancement curves (see paragraph 4 on ‘Image
interpretation’), a minimum of three different time points should be included:
first, a non-enhanced scan; second, a scan which captures the peak enhancement
of the lesion, and third, a scan with shows the delayed enhancement
characteristics of the lesion. In order to capture the peak enhancement of the
lesion, temporal resolution of the acquisitions performed should be in the
order of 60-120 seconds, but they should not compromise the in-plane spatial
resolution (which must be used for lesion morphology). In order to acquire a
reliable measurement of the delayed enhancement characteristics, it is
recommended to continue imaging until approximately 8 minutes after contrast
administration (Weinstein et al., 2010).
2.2.5 T2-weighted imaging sequences
This sequence is often used as ‘problem solver’
sequence, since it provides additional relevant information on different breast
lesions, narrowing down the differential diagnostic considerations.
For example, breast cysts (when inflammed) can
show rim enhancement after administration of contrast agent. In these cases,
signal intensity of the cyst is often slightly increased on the non-enhanced
T1-weighted image due to the proteinacious content of the cyst. Due to the high
water content and, consequently, the longer T2 relaxation times, cysts show a
very high signal intensity on T2-weighted images, and can thus be distinguished
(in combination with their sharp margins) from malignant breast lesions (Figure
1).
In 1999, Kuhl et al. demonstrated the
additional value of T2-weighted imaging in breast MRI by examining 205 benign
and malignant tumors. By means of visual assessment of the lesion appearance on
T2-weighted fast spin echo images, they were able to distinguish between
fibroadenomas and breast cancers, with a respective (age-dependent)
sensitivity, specificity, positive predictive value, and negative predictive
value for patients over 50 years of age of 89%, 62%, 85%, and 68% (Kuhl et al.,
1999a).
In another recent study, Baltzer et al.
evaluated 316 patients, of which 65 showed nonmass like enhancement on breast
MRI. BI-RADS predictors could not discriminate between benign and malignant
lesions with respect to nonmass like enhancement. However, the signal intensity
of T2-weighted images and the presence of cysts improved the diagnostic
accuracy, with a sensitivity of 91% and a specificity of 65% (Baltzer et al.,
2011).
Fig. 1.
Example of the added value of T2-weighted breast imaging. (A) shows the primary
metaplastic tumor in the right breast. At MRI, a suspicious lesion was observed
in the contralateral breast (B), with a corresponding high signal intensity on
T2-weighted imaging (C). Second look ultrasound demonstrated a small simple
cyst at this site, which was subsequently aspirated (D).
However, both benign and malignant breast
lesions may show increased signal intensity on T2-weighted images. In a review
of the histopathologic findings in such a group of lesions, Santamaria et al.
stated that MR signal hyperintensity is most likely to be associated with the
following conditions: extensive necrosis, (micro)cysts, fatty or sebaceous
components, mucinous stroma, loose myxoid stroma, edema or hemorrhage
(Santamaria et al., 2010). But also other benign entities, such as myxoid
fibroadenomas, oil cysts, and intramammary lymph nodes are known to show an
increased signal intensity on these sequences (Kuhl, 2007). In addition, some
malignant lesions might also demonstrate an increased signal intensity on
T2-weighted images, especially mucinous carcinomas due to their mucinous
content (Santamaria et al., 2010).
3. Indications
for breast MRI
Breast MRI can be used for a variety of
diagnostic problems. Proper indications for performing breast MRI (as supported
by the European Society of Breast Cancer Specialists and the European Society
of Breast Imaging) are: inconclusive findings in conventional imaging,
preoperative staging, unknown primary cancer, evaluation of therapy response in
neoadjuvant chemotherapy, imaging of the breast after conservative therapy,
screening of the high risk patient, breast implant imaging, and MR-guided
interventions, such as biopsy and lesion localization (Mann et al., 2008,
Sardanelli et al., 2010, Yeh, 2010).
3.1 Inconclusive
findings in conventional imaging
In a study by Berg et al., 177 malignant
lesions in 121 breast were evaluated with mammography, ultrasound, and MRI.
They showed that the sensitivity for detecting tumors decreased from 100% in
fatty breasts, to only 45% in extremely dense breasts. The sensitivity of
mammography was highest for invasive ductal carcinoma (89%), versus 55% for
ductal carcinoma in situ, and only 34% for invasive lobular carcinoma.
Ultrasound demonstrated a higher sensitivity for both invasive ductal (94%) and
invasive lobular carcinoma (86%). Sensitivity for detecting ductal carcinoma in
situ was worse for ultrasound (47%), presumably owing to the fine
microcalcifications associated with ductal carcinoma in situ, which are much
better visualized on mammography. However, MRI was superior to all other
modalities and for all tumor types: it detected 95% of the cases of invasive
ductal carcinoma, 96% of the cases of invasive lobular carcinoma, and 89% of
the cases of ductal carcinoma in situ (Berg et al., 2004). Due to this superior
ability to detect breast cancer, MRI can be used as a problem-solving modality,
when inconclusive findings in conventional imaging are encountered. For
example, patients can be reffered from the mammography screening programm with
abnormalities owing to a presumable superposition of fibroglandular tissue.
These patients can undergo a single breast MRI to exclude possible underlying
malignancies. Also, if there are discrepancies between clinical examination,
mammography, and/or ultrasound, MRI can serve as a powerful problem-solving
entity.
This was demonstrated by Moy et al., who
retrospectively reviewed all MRI examinations (n=115) of the breast that were
performed for inconclusive findings at mammography. They found no suspicious
correlate on MRI in 87% of the cases. In the remaining 15 cases (13%), 6
malignancies were found. However, 18 incidental lesions were also observed on
these examinations (Moy et al., 2009). Similar results were observed by Yau et
al., who reviewed 3001 MRI exams and found 204 MRI exams that were performed
for ‘problem solving’. Of these 204 exams, 42 were graded as BI-RADS category 4
or 5 (see also paragraph 4.4). Malignant
lesions were found in 14 cases, whereas benign findings or follow-up imaging
encompassed the remaining 28 cases. 162 exams were graded as BI-RADS category
0, 1, 2, or 3. In this group, biopsy was performed in 28 cases, revealing 1
malignant lesions. In the remaining 134 cases, no biopsy was performed within
the following 12 months (Yau et al., 2011). Both studies concluded that MRI is
a valuable tool for evaluation of inconclusive mammography findings, but
patient selection criteria should be strict because of the high incidence of
incidental lesions seen on MRI.
3.2 Preoperative
staging
The assessment of tumor size and additional
tumor foci is essential for establishing the proper surgical and post-surgical
treatment of each individual patient.
Recently, Uetmatsu et al. compared the ability
to assess breast cancer extension for mammography, ultrasound, breast MRI, and
even multidetector row computed tomography (MDCT). In this study of 210 breast
tumors, they showed that the accuracy for establish the tumor extent (compared
to histopathological results) was highest for breast MRI: 76%. The accuracy of
establishing the tumor extent was lower for the other modalities: MDCT 71%,
ultrasound 56%, and mammography 52%. However, they showed that MRI and
ultrasound had a substantial risk of overestimating the tumor size. With
respect to ductal carcinoma in situ extent, their study showed that the
accuracy of breast MRI was also highest: 89% (followed by MDCT (72%),
ultrasound (61%), and mammography (22%)). They concluded that breast MRI had
the highest accuracy for assessing the true breast cancer extent, but emphasize
that there is a risk of overestimation, which should be considered in
pre-surgical planning (Uematsu et al. 2008). In line with these results, the
superiority of assessing the proper breast tumor extension was also
demonstrated by several other studies (Mann et al., 2008, 2008b).
Also, MRI can be helpful for detecting
additional tumor foci (Figure 2). In a study of 969 patients by Lehman et al.,
simultaneous contralateral breast cancer was detected by breast MRI in 3% of
the cases (Lehman et al., 2007).
Tumor multifocality or multicentricity can also
be accurately assessed by MRI (Figure 3). For instance, this was demonstrated
by Drew et al. in their study of 334 women, with 178 confirmed cancer cases.
With preoperative breast MRI, multifocal or multicentric breast cancers was
suggested in 38% of the cases. In this particular group, histology eventually
demonstrated multifocality or multicentricity in 74% of the cases. Unifocal
breast cancer was found in 22% of the cases, benign breast disease in 4%. Their
observations resulted in a sensitivity of breast MRI for detecting
multifocal/multicentric cancer of 100%, with corresponding specificity,
positive predictive value, and negative predictive value of 86%, 73%, and 100%,
respectively (Drew et al., 1999).
Although
these results seem promising, the effectiveness of performing pre-operative
breast MRI was not evaluated until recently. In 2010, the COMICE trial, by
Turnbull et al., randomly assigned a total of 1623 patients to undergo either
pre-operative breast MRI (n=816) or no breast MRI (n=807). They demonstrated
that next to the conventional triple
Fig. 2. Detection of contralateral
breast cancer by breast MRI. (A) shows the primary index tumor in the right
breast, presenting as an irregular mass with rim enhancement. The tumor shows a
surrounding area of nonmass-like enhancement, with skin enhancement (open
arrow) and pectoral muscle ingrowth (arrow head). (B) shows an additional small
enhancing mass in the left breast (arrow), which corresponded with a small
hypoechoic mass on second look targeted ultrasound (C). Histologic biopsy of
this small mass revealed invasive ductal carcinoma, similar to the primary mass
in the right breast.
assessment performed in breast cancer, addition
of a pre-operative breast MRI did not result in a significantly reduced
re-operation rate (odds ratio 0.96, 95% confidence interval 0.751.24, p=0.77,
Turnbull et al., 2010).
In another (randomized controlled) trial of 418
patients (the MONET trial), Peters et al. allocated 207 patients to
preoperative stageing with MRI, and 21 patients to the control group (no
preoperative MRI). They found that the number of re-excisions performed because
of positive resection margins after primary breast conserving therapy was
increased in the MRI group: 34% in the MRI group versus 12% in the control
group (p=0.008). The number of conversions to mastectomy were similar (Peters
et al., 2011).
Fig. 3.
Detection of tumor multifocality and/or multicentricity by breast MRI. (A)
shows the index tumor in the lateral side of the left breast (*), with
additional tumor deposits in the medial part of the breast (arrows), resulting
in a multifocal, multicentric malignancy. (B) shows the index tumor in the
lateral side of the left breast (*), with an additional tumor deposit in the
same quadrant (arrow), resulting in a multifocal malignancy. Both cancers
proved to be invasive ductal carcinomas at biopsy.
However, both studies have some limitations.
For example, the COMICE trial recruited patients from 45 centres, resulting in
a large variation of radiologic experience when evaluating the breast MRI exams.
The MONET trial only evaluated non-palpable breast tumors and a subanalysis of
their results showed that the volume of the lumpectomy specimen was
significantly larger in the control group than in the group which was assigned
to preoperative breast MRI.
3.3 Unknown
primary cancer
This indication refers to the group of patients
who are diagnosed with metastases, but in who a primary tumor cannot be
identified. Schorn et al. demonstrated that MRI was helpful in patients with an
unknown primary cancer and a negative mammography and ultrasound of the
breasts. Breast cancer was detected by MRI in almost 50% of the cases. However,
it should be mentioned that this study only consisted of 14 patients (Schorn et
al. 1999). When looking only at axillary lymph node metastasis, Orel et al.
demonstrated in a study of 38 patients that breast MRI could detect the
previously unknow breast cancer in even 86% of the cases (Orel et al. 1999).
Therefore, in patients diagnosed with metastasis and negative mammography and ultrasound,
breast MRI should be strongly considered.
3.4 Evaluation of therapy respons in
neoadjuvant chemotherapy
In a study by Yeh et al., 31 women who
underwent neoadjuvant therapy for palpable breast cancer were included.
Agreements with the therapy respons rate as measured by clinical examination,
mammography, ultrasound, and breast MRI (as compared with pathology results)
were 19%, 26%, 35%, and 71%, respectively. Of these four modalities, MRI agreed
with the pathology results significantly more often: p<0.002 for all three
comparisons with MRI (Yeh et al., 2005).
Fig. 4.
Evaluation of tumor respons after neoadjuvant chemotherapy. (A) shows the
initial (large) tumor (invasive lobular carcinoma at biopsy) in the right
breast, presenting as a large area of regional nonmass like enhancement. (B)
shows significant reduction in tumor size and enhancing volume after three
gifts of chemotherapy. Thus, adequate chemotherapy respons was proven and
continued in this patient.
In another study, Shin et al. prospectively
included 43 patients with locally advanced or inflammatory breast cancer who
underwent neoadjuvant therapy. The assessment of therapy respons was evaluated
for clinical examination, mammography, ultrasound, and breast MRI. The
intraclass correlation coefficients between predicted tumor size (as assessed
by the different modalities) and the pathologically determined tumor size were
calculated. The values were highest for breast MRI (0.97), followed by
ultrasound (0.78), mammography (0.69), and clinical examination (0.65).
Agreement between the prediction of final therapy respons and the respons
assessed by pathology were expressed as the Kappavalue and were highest for MRI
(0.82), followed by ultrasound (0.50), mammography (0.44), and clinical examination
(0.43, Shin et al., 2010).
These results show that breast MRI is the most
suitable imaging modality to assess chemotherapy respons (Figure 4). In
addition, it is significantly more accurate in assessing the respons than
non-imaging techniques, such as clinical examination.
3.5 Imaging of the breast after
conservative therapy
There are three important reasons to perform
breast MRI after breast conserving therapy: 1) an evaluation tool for detecting
residual disease after positive tumor margins, 2) evaluation when recurrence is
suspected, and 3) screening for patients that underwent breast conservative
therapy in the past (Mann et al., 2008).
Due to the strong enhancement of the breast
tissue immediately after surgery (which can last for more than a year), the
interpretation of breast MR images for residual disease is hampered (Orel et
al., 1997). Lee et al. concluded that the evaluation of MRI for residual
disease in patients with close or positive margins is limited due to overlap in
the appearances of benign and malignant lesions (Lee et al., 2004). Image
interpretation can also be hampered by post-radiation enhancement of the
breast, which is known to occur up to three months after the last irradiation
of the breast. Nonetheless, Morakkabati et al. demonstrated that the detection
and characterization of breast lesions can be performed with comparible
diagnostic accuracies in irradiated breasts (when compared with nonirradiated
breasts, Morakkabati et al., 2003).
Finally, the risk of local recurrence is
dependent on the age of the patient at the time of the diagnosis (Mann et al.,
2008). Even with additional booster radiation therapy, these patients still
have a life-time risk of developing breast cancer of probably more than 20%,
which is equal to the life-time risk for breast MRI screening for the high risk
patient, as discussed in paragraph 3.6. Therefore, annual MRI screening can be
considered for patients that underwent breast conservative surgery for primary
breast cancer, but large trials are needed to confirm this assumption.
3.6 Screening of the high risk patient
The first
non-randomised studies to determine the additional value of breast MRI to
conventional mammography in women who were BRCA1 or -2 gene mutation carriers,
or who had a lifetime risk of at least 20-25% for developing breast cancer were
published in the 1990s. Based on these studies initiated in the Netherlands,
the United Kingdom, the United States, Canada, Italy, and Germany, the American
Cancer Society (ACS) and European
Society of Breast Imaging (EUSOBI) recommended annual MR evaluation of the
breasts for all women with a lifetime risk for breast cancer of more than
20-25% (Saslow et al., 2007, Mann et al., 2008). These women include known BRCA
gene mutation carriers, first-degree untested relatives of a BRCA gene mutation
carrier, women with radiation to the chest wall between ages 10 and 30 years,
Li-Fraumeni syndrome and first degree relatives, and Cowden syndrome with first
degree relatives (Boetes, 2010).
3.7 Breast
implant imaging
Past publications have shown that breast MRI
can be an excellent modality to assess breast implant integrity. The
sensitivity of MRI for detecting implant rupture can be as high as 80 to 90%,
with a specificity of over 90% (Brown et al., 2000, Cher et al., 2001, Hölmich
et al., 2005). However, specific sequences have to be used to optimize the
visualisation of silicone and to provide concurrent suppression of water
signal. Depending on the reason the study was requested, these
prothesis-specific sequences can replace, or can be added to the previously
discussed dynamic, contrast-enhanced breast MR imaging protocol. It is the
authors’ opinion, however, that a more eloborate description on the technical
aspects and interpretation of images in breast implant imaging is beyond the
scope of this chapter. An instructive
pictorial essay on breast implant rupture was recently published by Colombo et
al. (Colombo et al., 2011).
3.8 MR
guided interventions
Despite the high sensitivity of breast MRI,
it’s specificity is relatively low. In practice, this leads to many
false-positive findings, which require additional tissue sampling to exclude
malignancy. In 2009, an interdisciplinary European committee established a
consensus on the uses and technique of MR-guided vacuum-assisted breast
biopsies (HeywangKöbrunner et al., 2009). Although an elaborate discussion on
the indications and techniques of MR guided breast interventions is beyond the
scope of this chapter, the authors wish to emphasize some essential
recommendations of this consensus meeting
Before performing any kind of MR guided breast
intervention, a full imaging work-up should be completed. It must be absolutely
certain that the culprit lesion can only be visualized by breast MRI. Patients
should not have any kind of contra-indication for MRI or contrast
administration. Relative contra-indications are lesions close to the chest wall
who are estimated to be unfeasible or unsafe, patients with coagulation
disorders, and patients with breast implants. When these criteria are met, MR
guided biopsy of a breast lesion should be performed using a vacuum-assisted
breast biopsy system (core needle biopsies are not recommended). Minimum probe
size should be 11 Gauge, and the average number of cores taken should be 24 or
more (or an equivalent volume if a larger probe is used). The intervention does
not stop with acquiring the samples: proper correlation between histopathologic
results and MR findings should be performed, preferably in a multidisciplinary
setting. If the correlation is uncertain, re-biopsy or short-term follow-up
should be considered (Heywang-Köbrunner et al., 2009).
4. Image
interpretation
According
to the Breast Imaging Reporting and Data System (BI-RADS), the interpretation
of breast MR images should start with the analysis of the type of enhancement
observed.
Three categories of enhancement can be
observed: focal, mass-, and nonmass-like enhancement (Figure 5, Molleran et
al., 2010).
Subsequently, shapes and margins of the lesions
should be assessed in the case of masslike enhancement. In the case of
nonmass-like enhancement, it should be assessed whether this enhancement
pattern is linear, ductal, regional, or
segmental. In addition, the reader should assess if the nonmass-like
enhancement is clumped, in other words beaded or cobblestonelike.
Fig. 5. Examples of focus (A), mass (B), and
segmental (clumped) nonmass-like enhancement (C).
Finally, the enhancement characteristics of the
lesion should be assessed by looking at both the internal enhancement
characteristics and the signal intensity time curves. Internal enhancement
characteristics can be described as homogeneous, heterogeneous, rim
enhancement, or dark internal septations (American College of Radiology, 2003).
Lesions can demonstrate slow, intermediate, or rapid contrast enhancement in
the initial enhancement phase. In general, this initial enhancement phase can
be followed by three different types of enhancement curves in the delayed
phase: persistent enhancement, plateau phase, or wash-out. The enhancement
characteristics of lesions can be indicative for their benign or malignant
character.
By combining the findings of these different
analyses, the radiologist estimates the likelihood of a lesion being benign or
malignant. This estimation can be expressed in the final conclusion of the
report as the BI-RADS classification, and should be the basis for management
recommendations (i.e. biopsy or follow-up).
4.1 Focal, mass-, and nonmass-like
enhancement
Focal
enhancement can be described as small (less than 5 mm) area of enhancement that
cannot be specified otherwise. A mass is a lesion that is visible in three
dimensions and which occupies a space. Masses can be round, oval, lobulated, or
irregular, and may have smooth, irregular, or spiculated margins. Nonmass-like
enhancement is an area of enhancement that does not belong to a three
dimensional mass or that has no distinct mass characteristics (American College
of Radiology, 2003, Erguvan-Dogan et al., 2006). Nonmass-like enhancement
patterns can be divided in linear, ductal, segmental, and regional enhancement
(Figures 5 and 6).
Fig. 6. Proper terminology (according to the
BI-RADS lexicon) for enhancement patterns, shapes, margins, and nonmass-like
enhancement distributions.
Linear nonmass-like enhancement is defined
according to the BI-RADS lexicon of the American College of Radiology as
‘enhancement in a line that is not definitely in a duct’. Ductal enhancement
can be defined as ‘enhancement in a line that points towards the nipple, and
may have branching, conforming to a duct’. Segmental enhancement can be defined
as ‘a triangular region or cone of enhancement, with the apex pointing towards
the nipple’. Finally, regional enhancement can be defined as ‘enhancement in a
large volume of tissue not conforming to
a ductal distribution’ (American College of Radiology, 2003).
Jansen et al. recently investigated the
pathology and kinetics of mass, nonmass, and focal enhancement in a
retrospective study using dynamic contrast-enhanced breast MRI. They analyzed a
total of 852 breast lesions (histologically proven) in 697 patients. Of the
lesions demonstrating mass-like enhancement (n=552), 71.7% proved to be
malignant. Of the lesions demonstrating nonmass-like enhancement (n=261), 81.2%
proved to be malignant. The remaining lesions demonstrated focal enhancement
(n=30), which were usually benign (76.9%). Malignant mass- and nonmass-like
enhancing lesions differed significantly in their pathology (p<0.0001), with
mass-like enhancing lesions usually consisting of invasive ductal carcinoma and
nonmass-like enhancement usually consisting of ductal carcinoma in situ.
Similarly, benign mass- and nonmass-like enhancing lesions differed
significantly in their pathology (p<0.002), with the former usually
consisting of fibroadenomas and the latter usually presenting fibrocystic
changes. Finally, the predominant pathology of focal enhancing lesions was
fibrocystic changes (Jansen et al., 2011).
4.2 Morphologic descriptors in masslike-
and nonmass-like enhancement
Margins of masses can be described as smooth
(or sharp), irregular, or spiculated. Similar to mammography, some morphologic
features of a lesion are more associated with malignancy than others (Liberman
et al., 1998). Past studies showed that spiculated margins, irregular shapes, and linear/ductal
nonmass-like enhancement had the highest positive predictive values for
malignancy (Nunes et al., 1997, 2001). However, these studies included patients
with mammographic or palpable findings, creating a potential bias in the study
population.
Therefore, Liberman et al. performed a
retrospective review of 100 consecutive solitary MR imaging-detected lesions.
For mass-like enhancement, margins and shape were evaluated. With respect to
lesion margins, spiculated margins had the highest positive predictive value
for malignancy (80%), much higher than irregular (22%) and smooth (17%)
margins. With respect to lesion shapes, irregular shapes had the highest
positive predictive value for malignancy (32%), lobular shapes had a positive
predictive value for malignancy of only 13% (Liberman et al., 2002).
In the same study, the pattern of nonmass-like
enhancement was evaluated. With respect to linear or ductal enhancement,
clumped enhancement (or beadlike enhancement) had a positive predictive value
for malignancy of 31%. Smooth linear enhancement was not observed in malignant
lesions. Clumped regional enhancement had a positive predictive value of 67%,
whereas clumped segmental enhancement had a positive predictive value of 67%
too (Liberman et al., 2002).
In addition, Siegmann et al. looked at lesion
size as a additional descriptor for the assessment of malignancy. They showed
in a study of 51 lesions (in 45 patients) that lesions with a diameter of more
than 10 mm have a higher positive predictive value (45.5%) than lesions smaller
than 10 mm (27.6%, Siegmann et al., 2002).
To summarize, features that have the highest
positive predictive value for malignancy are spiculated (ill-defined) margins
and irregular shapes (based on morphology alone and in the case of masslike
enhancement). For nonmass-like enhancement, features that have the highest
positive predicitive value are clumped linear, segmental or regional
enhancement. Lesions larger than 10 mm have a higher positive predictive value
for being malignant than lesions < 10 mm
(Tse et al., 2007).
4.3 Kinetic analysis of the signal
intensity time curves
Lesion enhancement is described as homogeneous,
heterogeneous, rim enhancement, or enhancement with dark internal septations
(American College of Radiology, 2003, Figure 7).
In a landmark paper by Kuhl et al., the value
of signal intensity time curves was evaluated with respect to the differential
diagnosis of enhancing breast lesions. A total of 266 breast lesions (101
malignant, 165 benign) were examined using a dynamic contrast-enhanced breast
imaging protocol. The relative enhancement of breast lesions was assessed by
drawing a region-of-interest in the lesion itself. The enhancement was then
calculated according to the following formula:
Relative
signal enhancement (%) = (SIpost – SIpre) / SIpre
x 100
In this formula, SIpre and SIpost
represent pre-contrast and post-contrast signal intensities, respectively. By
calculating the signal intensity time curves, it was demonstrated that
enhancement patterns can be divided into two phases: early enhancement (from
contrast administration to approximately two minutes post-contrast, or when the
curve starts to change), followed by the delayed enhancement.
Fig. 7. Proper terminology
(according to the BI-RADS lexicon for lesions enhancement patterns)
homogeneous, heterogeneous, rim enhancement, and enhancement with dark internal
septa.
For the early enhancement phase, it was assumed
that benign lesions had a (slow) enhancement of 60% or less. Indeterminate
lesions were assumed to have an (intermediate) enhancement of more than 60%,
but less than 80%. Finally, malignant lesions were assumed to have a (strong)
enhancement of more than 80%. For these assumptions, the diagnostic accuracies
in this study were: sensitivity 91%, specificity 37%, positive predictive value
47%, negative predictive value 87%, diagnostic accuracy 58%. Mean peak
enhancement was significantly higher for malignant lesions than for benign lesions:
mean enhancement 104% versus 72%, p<0.001 (Kuhl et al., 1999b).
For the delayed phase, three different type of
signal intensity curves were defined. A type I curve was characterized by a
persistent increase in signal intensity over time. A type II curve was
characterized by a plateau in signal intensity values over time. Finally, a
type III curve was characterized by a so-called washout, i.e. the signal
intensity decreases in time after the initial upslope in the early enhancement
phase (Figure 8).
For benign
lesions, a type I curve was observed in 83.0% of the cases. A type II curve was
observed in 11.5% of the cases, whereas a type III curve was hardly seen in
benign lesions:
5.5% of the
cases. For malignant lesions, a type III curve was most frequently observed:
57.4% of the cases. A type II curve was observed in 33.6% of the cases, whereas
a type I curve was infrequently seen in these cases: 8.9%. The assessment of
the signal intensity time curves had an excellent interreader agreement with a Kappa-value
of 0.849, p<0.001 (Kuhl et al., 1999b).
enhanced breast MRI.
In the past, Jansen et al. demonstrated that
analysis of the signal intensity time curve can help distinguish between benign
and malignant mass lesions effectively, but the analysis is not that useful in
discriminating between benign and malignant nonmass-like lesions. Although
their pilotstudy only consisted of a total of 108 breast lesions with 70
observed masses, 44 of which were malignant and 26 benign. There were 38
nonmass-like lesions observed, of which 31 were malignant and 7 benign. Despite
these relatively small numbers, they showed that analysis of the signal
intensity time curve was helpful in distinguishing between benign and malignant
masses on MRI. However, it could not be used to accurately distinguish between
benign and malignant nonmass-like lesions. Therefore, they suggested that
analysis of the signal intensity time curves of nonmass-like enhancement is not
very useful and that morphology analysis should be favored (Jansen et al.,
2008).
In summary, it is advised by the BI-RADS MRI
lexicon that the signal intensity curve of a lesion should be described
qualitatively. A proper region-of-interest should at least contain 3 pixels and
if this enhancement of the lesion is heterogeneous, the most suspicious
enhancement curve should be mentioned in the final report. Initial enhancement
can be slow, moderate, or rapid, while the delayed enhancement can show a
persistent, plateau, or wash-out curve (American College of Radiology, 2003). A
strong early enhancement is suggestive of malignancy, whereas a slow signal
intensity increase is suggestive of a benign entity. More importantly, type I
signal intensity curves are suggestive of benign breast lesions, whereas type
III curves are suggestive of malignancy. The indeterminate type II curve scan
be observed in both benign and malignant breast lesions, albeit slightly more
suggestive of malignancy (in a ratio of 2:3, Kuhl et al., 1999b).
It should
be emphasized that kinetic analysis of contrast enhancement is no substitute
for morphology analysis. It should be used as an aid in further narrowing the
differential diagnosis. With this respect, several recommendations can be made:
First, it is recommended to perform the kinetic
analysis after morphologic analysis of a lesion. When the morphology is highly
suggestive of malignancy, kinetic analysis should be skipped, and the lesion
should be biopsied. Kinetic analysis should be performed in lesions with
indeterminate or benign morphologies.
Second, lesions with a type III enhancement
curve should always be biopsied, even if morphology is suggestive of a benign
lesion. In contrast, the absence of a clear wash-out phase in the signal
intensity time curve cannot rule out malignancy.
Third, when lesion morphology is indeterminate
and a type I curve is observed, follow-up of the lesion might be considered to
reduce false-positive biopsy findings.
4.4 What the clinicians need to know:
report organization
The pre-surgical planning and post-surgical
treatment is dependent not only on tumor type, but also on it’s corresponding
TNM-classification. The most recent TNM-classification, edition 7, was recently
published in 2010. (Edge et al., 2010). For a proper TNMclassification, several
issues need to be adressed in the final report of any breast MRI.
For a proper T-classification of breast cancer,
the maximum diameter of the culprit mass should be mentioned in the report,
including any suspicious nonmass-like enhancement that can be associated with
an extensive intraductal component. In addition, the relationship of the tumor
to the skin, pectoral muscle and thoracic wall must be accurately described.
Enhancement of the pectoral muscle or skin is one of the most reliable signs
for the assesment of tumor invasion in these structures. Although inflammatory
breast cancer is clinical diagnosis, it can be suggested in MRI when strong
enhancement of the breast is observed, together with diffuse skin thickening
and enhancement.
Many authors have tried to developed accurate
criteria for the assessment of axillary lymph node status on MRI. In a study of
65 patients, Kvistad et al. demonstrated a significant correlation between flow
kinetics and axillary lymph node status (Kvistad et al., 2000). Murray et al.
demonstrated a correlation between nodal enhancement and nodal area and
axillary lymph node status in a study encompassing 47 patients (Murray et al.,
2002). More recently, Mortellaro et al. stated in their study of 56 patients
that the presence of any axillary lymph node without a fatty hilum and the
number of nodes without a fatty hilum correlated significantly with axillary
lymph node positivity for metastases (Mortellaro et al., 2009). In summary,
study results on MRI of axillary lymph node status vary in study design, study
population, and outcome. Until now, there are no reliable criteria for the
evaluation of axillary lymph node positivity. However, it is the authors’
opinion that analysis of the axillae is an important part of the total breast
MRI evaluation. Patients with suspicious axillary lymph nodes on MRI should be
considered for (re)evaluation with (second look) ultrasound.
With respect to a proper M-classification, it
should be emphasized that other imaging modalities, such as (PET-)CT, need to
be performed. However, extramammary findings on breast MRI should be noted and
reported. In a retrospective review of 1535 breast MRI examinations, Rinaldi et
al. observed 285 patients with extramammary (incidental) findings. Most
incidental findings occured in the liver (51.9%). Other sites were lung
(11.2%), bone (7%), and mediastinum (4.2%). Pleural or pericardial effusions
were observed in 15.4% of the cases. Of all these incidental findings, 20.4%
proved to be malignant (Rinaldi et al. 2011). Therefore, the occurence of
extramammary findings is a non-negligible phenomenon.
Finally,
the radiologst should construct a comprehensible report of all findings
observed on breast MRI. By analyzing morphology, enhancement, and signal
intensity time curves, the probability of malignancy should be estimated. The
maximum diameter of suspicious lesions should be provided, together with their
location within the breast and their relationship with the skin, pectoral
muscle, or thoracic wall. Together with an assessment of the axillary lymph
node morphology and incidental extra-mammary findings, the
radiologist should finish the report with the
appropriate BI-RADS classification and possible
management recommendations (Americal College of Radiology, 2003):
|
BI-RADS 1:
|
Additional imaging is needed (i.e. failure of equipment, severe
artefacts)
|
|
BI-RADS 1:
|
Normal, there is nothing to comment on
|
|
BI-RADS 2:
|
Benign findings
|
|
BI-RADS 3:
|
Probably
benign findings; the probability of malignancy is less than 2%.
|
|
|
Short-term follow-up is recommended
|
|
BI-RADS 4:
|
Suspicious
findings; the probability of malignancy is 2-95%. Biopsy should
|
|
|
be considered
|
|
BI-RADS 5:
|
Highly suggestive of malignancy;
the probability of malignancy is higher
|
|
|
than 95%. Appropriate action should be taken
|
|
BI-RADS 6:
|
Proven malignancy (through histopathologic results)
|
In conclusion, dynamic, contrast-enhanced
breast MRI can be a powerful adjuvant imaging modality for the detection of
breast cancer. It can be of help when inconclusive findings are encountered on
conventional imaging or in the case of an unknown primary cancer. The
evaluation of neoadjuvant chemotherapy respons can be evaluated with breast
MRI, and it can aid in the assessment of the postoperative breast. Breast MRI
is advised in screening certain populations with high risk of developing breast
cancer, breast implants can be accurately analyzed with MRI, and it can aid in
MR guided breast interventions. One of the most important indications of breast
MRI is preoperative planning, and it’s superiority compared to other breast
imaging modalities to evaluate disease extent, multifocality or
multicentricity, and the presence of (occult) contralateral malignancy.
However, due to it’s limited specificity, falsepositive findings are frequently
observed. Therefore, patient selection should be performed with care and the
proper indications for breast MRI should be observed.
