| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Division of Radiology, Magnetic Resonance Science Center, University of
California, Box 1290, 1 Irving St., Rm. AC-109, San Francisco, CA
94143-1290.
2
Present address: Department of Radiology, Hopitaux Universitaires de Geneve,
Hopital Cantonal, 24, rue Micheli-du-Crest, CH-1211 Geneve 14,
Switzerland.
3
Present address: Department of Radiology, University of Vienna AKH, Waehringer
Guertel, Vienna 1090, Austria.
4
Department of Surgery, University of California, San Francisco, CA
94143.
5
Department of Epidemiology and Biostatistics, University of California, San
Francisco, CA 94143.
Received September 2, 1999;
accepted after revision December 14, 1999.
Supported in part by grants from the National Institutes of Health and the
United States Army.
Abstract
 Top Abstract IntroductionMaterial and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Fifty-seven patients with suspicious mammographic or palpable findings underwent preoperative contrast-enhanced MR imaging of the breast using a three-time-point method of acquisition. Each lesion was prospectively analyzed by two independent radiologists for morphologic and visual dynamic enhancement characteristics. A classification and regression tree was used to examine the optimal order, cutoff points, and combination of imaging parameters to build a diagnostic rule separating benign from malignant lesions using histopathology findings as the standard of reference. Kappa statistics were used to determine observer variability.
RESULTS. Among 23 benign and 34 malignant lesions (12 invasive, three ductal carcinoma in situ, and 19 mixed cancer), margin morphology (p = 0.001) and enhancement pattern (p = 0.001) were the most significant MR imaging findings for lesion characterization. Focal mass lesions were classified as malignant when spiculated margins or both the washout enhancement pattern and "nonsmooth" margins were present. Interobserver agreement was almost perfect for washout pattern and substantial for margin assessment. In the limited population tested retrospectively, the diagnostic rule yielded a sensitivity and positive predictive value of 97% each and a specificity and negative predictive value of 96% each.
CONCLUSION. The washout enhancement pattern combined with lesion margin assessment on dynamic contrast-enhanced high-resolution MR imaging of the breast allows reproducible lesion characterization and may be a highly specific diagnostic tool.
|
TopAbstractIntroductionMaterial and MethodsResultsDiscussionReferences |
|---|
The rapid uptake and diffusion of low-molecular-weight contrast agents in breast tissue precludes the simultaneous acquisition of optimal dynamic and high-spatial-resolution MR images. Because both types of information—morphologic feature analysis and dynamic enhancement characteristics—seem to be useful for lesion characterization, this study evaluated a combination of morphologic and semidynamic data by acquiring three high-spatial-resolution MR sequences. The purpose was to develop diagnostic criteria to improve lesion characterization using this new technique. In addition, we assessed whether interobserver variability in the assessment of morphologic and kinetic lesion characteristics influenced the efficacy of the technique.
|
TopAbstractIntroductionMaterial and MethodsResultsDiscussionReferences |
|---|
Breast surgery was performed in 52 of the 57 patients 1-6 weeks after MR imaging and consisted of core biopsy (n = 21), excisional biopsy (n = 5), lumpectomy (n = 14), or excisional biopsy followed by mastectomy (n = 12). In two patients with malignant lesions and three patients with benign lesions, a core biopsy was postponed 8-38 weeks after MR imaging. Among 34 patients with histopathologically proven breast cancer, 19 (56%) had invasive ductal cancer with carcinoma in situ (DCIS), eight (23%) had invasive ductal cancer without DCIS, three (9%) had invasive lobular cancer, three (9%) had DCIS alone, and one (3%) had invasive tubular cancer. Twenty-one (62%) of 34 patients with breast cancer had stage pT1, three (9%) had stage pT2, eight (24%) had pT3, and two (6%) had pT4; 24 (71%) of these patients had stage pN0 and 10 (29%) were pN positive. No patient had distant metastasis at the time of diagnosis.
Histopathologic results of the 23 patients with benign breast lesions included proliferative (n = 3) or nonproliferative (n = 5) fibrocystic disease, fibroadenoma (n = 3), sclerosing adenosis (n = 2), apocrine metaplasia (n = 2), atypical ductal (n = 2) or atypical lobular (n = 4) hyperplasia, benign granular cell tumor (n = 1), and intraductal papilloma (n = 1).
MR Imaging
MR imaging was performed on a Signa system (General Electric Medical
Systems, Milwaukee, WI) at 1.5 T. All patients were imaged in the prone
position in a dedicated double breast coil. The breast with the known palpable
or mammographic abnormality was imaged. No compression device was used. An IV
cannula was connected to an automatic injector. A transverse T1-weighted
spin-echo sequence was performed for localization purposes and was followed by
a sagittal fat-suppressed T2-weighted fast spin-echo sequence with the
following parameters: TR/TE, 4600/84; field of view, 18 cm; matrix size, 256
x 192; slice thickness, 3.0 mm with a 0.5-mm gap; and two signals
averaged. A three-dimensional sagittal fatsuppressed T1-weighted fast
gradient-recalled echo sequence was obtained before, at 2 min 30 sec, and at 7
min 30 sec after a bolus injection of 0.1 mmol/kg of gadopentetate dimeglumine
(Magnevist, Berlex Laboratories, NJ) with an acquisition time of 5 min for
each set of 60 images. The following parameters were used: TR/TE, 11.0/4.2;
field of view, 18 cm; matrix size, 256 x 192; slice thickness, 2.0 mm
with no gap; two signals averaged; and spectrally selected inversion
recovery-prepared fat suppression. Reformatted images in craniocaudal,
mediolateral, and anteroposterior projections (maximum intensity projection)
were created from unenhanced and early and late contrastenhanced
three-dimensional fast gradient-recalled echo data sets, each containing 60
sections.
Image Interpretation
Lesions were identified initially by a board-certified radiologist to
ensure that both radiologists performing image interpretation were
characterizing the same lesion. The radiologist was aware of the patient's
breast quadrant containing the clinical or mammographic abnormality. All
identified lesions were then evaluated independently by two radiologists for
the morphologic and dynamic parameters listed in
Table 1. One radiologist had
interpreted 700 breast MR examinations since 1993 and 2000 diagnostic
mammograms per year. The other radiologist had interpreted 350 breast MR
examinations since 1996 and 1500 screening or diagnostic mammograms a
year.
|
Morphologic parameters were lesion type (mass versus nonmass enhancement), mass margin, internal enhancement, and T1- and T2-weighted unenhanced visibility.
Dynamic parameters were enhancement degree and enhancement pattern. Lesion interpretation for enhancement degree and pattern was performed visually without the use of region-of-interest (ROI) methods or computer-guided parametric programs.
The degree of enhancement was defined in the most enhancing portion of the lesion identified on the first contrast-enhanced image. This was a qualitative assessment of the degree of enhancement in which the variable absent, low, mild, moderate, or strong was assessed in the area of maximum enhancement relative to vessel enhancement.
Enhancement pattern was visually assessed by comparing the signal intensity on the first and second dynamic high-spatial-resolution images acquired at 2 min 30 sec and 7 min 30 sec, respectively, and presented with identical window and level settings. By definition, any decline in signal intensity between 2 min 30 sec and 7 min 30 sec after contrast material injection was considered a "washout" enhancement pattern. "Plateau" enhancement was considered to be stabilized enhancement without change in signal intensity between 2 min 30 sec and 7 min 30 sec. "Progressive" enhancement was considered to be an increase in signal intensity between 2 min 30 sec and 7 min 30 sec. Each lesion was characterized by the strongest enhancement pattern seen over the entire lesion, in which the order of decreasing strength was washout, followed by plateau and then progressive.
Parameters with categoric ordered variables were margin, degree of enhancement, and enhancement pattern.
The parameter internal enhancement had nonordered categoric variables previously described in the literature [8, 9]. All morphologic characteristics were analyzed on the first contrast-enhanced images acquired at 2 min 30 sec. T1 and T2 unenhanced visibility were parameters with dichotomous variables and were analyzed as potential lesion localizers for secondary MR imaging-guided core biopsies. These parameters were assessed after lesion identification on contrast-enhanced images.
Before starting image interpretation of study cases, the two radiologists reviewed together 20 breast MR imaging teaching cases representing various types of morphologic lesion features. Both radiologists were trained to detect the washout enhancement pattern in any location of the lesion (periphery, center, or entire lesion) whenever a slight decline in signal intensity became visually evident.
Both interpreters were unaware of pathology results. The greatest lesion size in three dimensions and the lesion location in the breast were assessed on reformatted projections to help the correlation between surgery, pathology, and MR imaging findings. If more than one lesion was identified, each lesion was assessed separately.
Histopathologic Correlation
Histopathologic correlation between MR images and histopathologic sections
was performed according to standard histopathologic procedures at our
institution after the performance of a research breast MR imaging examination.
All specimens were anatomically oriented during surgery to distinguish the
mediolateral, inferosuperior, and anteroposterior borders. A drawing that
showed the origin of the specimen in the breast was provided during surgery.
Specimens were serially sectioned medially to laterally in a sagittal
orientation. The histopathologic slides were numbered according to their
position in the specimen. To help the correlation between surgery, pathology,
and MR imaging findings, the largest histopathologic size of all lesions was
given in millimeters and was compared with the largest lesion size in three
dimensions assessed on reformatted projections. For lesions proven by core
biopsy only, lesion location on the MR image was verified on the mammogram or
sonogram used to guide the biopsy. For data analysis, only histologically
proven lesions were considered. Therefore, seven incidentally enhancing
lesions with no mammographic or sonographic correlate and no subsequent
histopathologic proof were not included in the data analysis.
Data Analysis and Statistics
All lesions were classified as benign or malignant (invasive carcinoma and
DCIS) using histopathology results. Each imaging parameter was analyzed
separately for its association to the benign or malignant nature of the lesion
using the chisquare test. Parameters with a p value of less than 0.05
were considered significant. Classification and regression tree analysis
[13,14,15]
was performed to develop diagnostic criteria, using the classification and
regression tree software (CART [classification and regression analysis];
Salford Systems, San Jose, CA). This program was designed to test the optimal
threshold values and combination and order of parameters yielding the highest
sensitivity and specificity for the characterization of suspicious breast
lesions. Our goal was to establish diagnostic criteria for breast lesion
characterization testing the best combination and threshold values for
different imaging parameters. For each node in the diagnostic algorithm, a
parameter is chosen that best splits the patient population into benign or
malignant cases. Each branch from the node is composed of cases that are more
similar to each other than to the node from which they stem. The choice of the
best splitting parameter depends on the improvement scores calculated with
CART [13]. Improvement scores
are unitless variables that provide a ranking of the strength of a parameter
for each node. The higher the improvement score of a parameter, the better its
performance in differentiating benign from malignant lesions. Review of
surrogate and competitor variables was performed to determine if any variables
were masked [13]. Sensitivity
and specificity acquired from a 90% cross-validation sample of the study
population were compared with the results provided by the first image
interpretation of one radiologist for the entire study population. Intra- and
interobserver variability for the imaging parameters were assessed using
nonweighted kappa statistics
[16]. The intraobserver
analysis of variables was performed using the two interpretative results from
one radiologist (separated by a period of 4 months and unaware of the
patient's name), whereas interobserver variability was performed by comparing
the interpretative results of both radiologists. We attached the following
qualitative terms to the kappa values: 0.0-0.2 = slight, 0.2-0.4 = fair,
0.4-0.6 = moderate, 0.6-0.8 = substantial, 0.8-1.0 = almost perfect agreement
[16].
To test the influence of inter- and intraobserver variability on the final diagnosis, we entered separately into the diagnostic algorithm both interpretative results of one and the interpretative result of the other radiologist.
|
TopAbstractIntroductionMaterial and MethodsResultsDiscussionReferences |
|---|
0.001)
than did the parameters of internal enhancement (p 0.032) and T1
(p 0.011) and T2 unenhanced material visibility (p
0.013).
Table 2 shows the inter- and
intraobserver variability of each imaging parameter with corresponding kappa
values. Variables with categoric values, such as progressive, plateau, and
washout, were assessed for complete concordance and for concordance of
dichotomous values, such as washout versus no washout. Pattern of enhancement
and margin had substantial interobserver agreement, whereas internal
enhancement and degree of enhancement showed only fair agreement. When the
reproducibility of the variable washout versus no washout was assessed, the
interobserver agreement improved to almost perfect (
= 0.8). T1 and T2
lesion visibility had moderate interobserver agreement.
|
Malignant Lesions
All 34 malignant lesions were focal masses, and none corresponded to areas
of regional enhancement. The mean size of the 34 malignant lesions was 12.3 mm
(±6 mm [SD]). All lesions enhanced with a strong or moderate degree of
enhancement. The washout enhancement pattern appeared in 29 (85%) of 34
malignant lesions. Figures
1A,1B,1C
and
2A,2B,2C
show two examples of focal lesions with subtle (Fig.
1A,1B,1C)
or more evident (Fig.
2A,2B,2C)
washout enhancement patterns. Figure
3A,3B,3C
provides an example of a slightly progressive enhancement pattern. The five
malignant lesions without washout corresponded to invasive ductal cancer with
(n = 4) or without (n = 1) surrounding DCIS (Fig.
3A,3B,3C).
All three cases of DCIS alone presented as focal masses with washout. The most
frequent types of margins were irregular (Fig.
1A,1B,1C)
and spiculated (Figs.
2A,2B,2C
and
3A,3B,3C).
None of the 34 malignant lesions had smooth margins or nonenhancing septation.
The comparison between the contrast-enhanced high-resolution MR images and the
unenhanced MR images allowed the retrospective identification of more than two
thirds of malignant lesions on T1-weighted unenhanced images and T2-weighted
images.
|
|
|
|
|
|
|
|
|
Benign Lesions
Among the 23 benign lesions, five corresponded with areas of regional
enhancement and 18 with focal masses. The mean size of the benign lesions was
15.0 ± 4 mm. All types of margins were found
(Table 1); the largest group
had irregular margins (n = 9, 39%) (Fig.
4A,4B,4C).
One lesion with spiculated margins showed proliferative fibrocystic disease at
histopathology. Most benign lesions showed a moderate (Fig.
4A,4B,4C)
or strong (Fig.
5A,5B,5C)
degree of enhancement. Although the progressive or plateau enhancement pattern
appeared in 95% of benign lesions (Fig.
4A,4B,4C),
washout enhancement was possible and corresponded to one case each of atypical
ductal hyperplasia, intraductal papilloma, and benign granular cell tumor
(Fig.
5A,5B,5C).
The retrospective identification of benign lesions on unenhanced sequences was
significantly lower than that of malignant lesions (p
0.013).
|
|
|
|
|
|
Diagnostic Algorithm
Table 3 shows the
improvement scores for each parameter in the nodes of the diagnostic
algorithm. The best improvement score in characterizing benign or malignant
lesions was obtained by the parameter enhancement pattern with the parameter
washout, followed by lesion margin. The diagnostic algorithm, determined by
the classification and regression tree analysis, is presented in
Figure 6. The tree diagram
shows how breast lesions can be characterized using the washout enhancement
pattern and margin characteristics. The breakdown of margin features in benign
or malignant lesions depends on whether washout took place: for lesions with
washout, only smooth margins are classified as benign; for lesions without
washout, all margins, with the exception of spiculated margins, are classified
as benign. A lesion with irregular or lobulated borders is classified as
benign in the absence of washout and as malignant in the presence of
washout.
|
|
Highest sensitivity and specificity were obtained in three nodes. The first node, on the basis of the presence or absence of washout, split the data set into 29 malignant and three benign lesions showing washout and 20 benign and five malignant lesions without washout. Therefore, the first node yielded a sensitivity of 85% (29/34) and specificity of 87% (20/23).
The second node in the 32 lesions with washout and the third node in the 25 lesions without washout were both best split into benign and malignant lesions using lesion margin. Because 91% (29/32) of the lesions with washout were malignant and did not reveal smooth margins, versus three benign lesions with washout and smooth or "unable to assess" margins, the best cutoff point for lesion characterization in the second node was smooth versus nonsmooth margins. The histopathology of those three benign lesions with washout corresponded to one intraductal papilloma and one benign granular cell tumor with smooth margins and one atypical ductal hyperplasia with margins that were unable to be assessed because of clumped regional enhancement.
In the third node the optimal cutoff for lesion characterization was spiculated versus nonspiculated. The four spiculated malignant lesions without washout corresponded to three invasive ductal carcinomas associated with DCIS and one invasive lobular cancer.
The three nodes of the diagnostic algorithm yielded 33 true-positive results, 22 true-negatives, one false-positive (proliferative fibrocystic disease, no washout, and spiculated margins) and one false-negative result (invasive ductal carcinoma, no washout, and irregular margins), yielding a sensitivity and positive predictive value of 97% (33/34) and a specificity and negative predictive value of 96% (22/23). No additional parameters yielded nodes that could further improve the performance of dynamic high-resolution MR imaging in this study population.
Summarizing the diagnostic algorithm, the following diagnostic rules can be formulated (Fig. 6). A lesion is classified as malignant in the presence of spiculated margins or nonsmooth margins and washout. A lesion is classified as benign in the absence of spiculated margins or washout; however, a lesion with smooth margins and washout is still classified as benign.
The cross-validation sample estimates of sensitivity and specificity were 91% and 83%, respectively, predicting the sensitivity and specificity that could be obtained if the diagnostic criteria were applied to a prospective set of data.
When the diagnostic algorithm was applied to the second interpretation of one radiologist, intraobserver variability did not change lesion characterization and yielded an unchanged sensitivity of 97% and a specificity of 96%. When the diagnostic algorithm was applied to the interpretation of the second radiologist, the interobserver variability in the assessment of washout and margin changed the final diagnosis in six (11%) of 57 patients.
|
TopAbstractIntroductionMaterial and MethodsResultsDiscussionReferences |
|---|
This study provides optimized diagnostic criteria for lesion characterization. Among four morphologic and two dynamic enhancement imaging parameters, margin and enhancement pattern were the optimal discriminators for breast lesion characterization. The combination of both types of information in the presented diagnostic algorithm achieved an optimized specificity and negative predictive value of 96% while maintaining sensitivity and positive predictive value at 97%. Although this study was not prospective, results of the cross-validation predicted the prospective performance of those diagnostic criteria [13]. The 6% decline in sensitivity and 13% decline in specificity between the study population and the cross-validation sample might represent a more realistic measure of those criteria in a clinical setting.
To our knowledge, this is the first report on the diagnostic value of visually assessed washout. The study supports the potential value of washout suggested by Sherif et al. [12] and Kuhl et al. [11]. However, the measurement period of washout in our study and in the study of Kuhl et al. was shortened to 6.5-7.5 min after contrast enhancement versus 1 hr in the study of Sherif et al. High-spatial-resolution images allowed us to observe that washout was not always peripheral but often presented as being centrally or heterogeneously distributed, which may better represent the spectrum of tumor biology. We therefore defined "washout" as being identified anywhere in the lesion, not just at the periphery (as in the study by Sherif et al.), which may be one reason that our study yielded a higher sensitivity than the study of Sherif et al. As shown in this study, one advantage of visual versus user-defined ROI measurements of washout is excellent intra- and interobserver reproducibility. Indeed, ROI measurements are strongly operator- and technique-dependent [17, 19, 20] and show low to fair interobserver reproducibility [20]. The reproducibility of parametric programs measuring contrast enhancement has been reported to be as high as visual assessment [11, 21] and superior to ROI measurements [22] but parametric programs require sophisticated image-analysis software. The high reproducibility of enhancement pattern by an observer with no previous experience in visually assessing enhancement pattern suggests that visual washout assessment is easy to learn. We recommend including visual washout assessment in the interpretation of enhancing lesions when parametric semiautomated programs are not available.
The substantial agreement on margin assessment in this study differs from the fair interobserver agreement reported by Mussurakis et al. [20]. More defined categories and the use of observer training might explain the difference in reproducibility in our study.
The interpretation model presented here is not inconsistent with that of Nunes et al. [7, 10]. Both models used margins for the characterization of mass enhancement. However, the high-spatial-resolution acquisition technique used in two studies by Nunes et al. [7, 10] did not provide dynamic information. Therefore, dynamic parameters were not part of the interpretation model for high-spatial-resolution MR imaging. Using the proposed three-time-point acquisition, visual assessment of washout was possible and therefore was included in the data analysis. The value of washout in our diagnostic algorithm was to dictate different diagnostic strategies for the interpretation of lesion margin in those lesions showing washout versus those not showing washout. The final diagnosis depended heavily on margin assessment, with no other features leading to improvement in the performance. Our interpretation rules, therefore, resulted in significantly lower parameters than the interpretation model proposed by Nunes et al., which, in addition to margin, included septation and enhancement degree.
A number of characteristics of our study population may limit the generalizability of our findings. We evaluated a relatively small group of 57 patients. Although this interpretation model included nonfocal enhancement, characterization of those lesions could not be achieved because of the small proportion of this lesion type. All five lesions corresponding to nonfocal enhancement in this study—one of which revealed washout enhancement and corresponded to atypical ductal hyperplasia—were benign. However, nonfocal enhancement has been described in 23-38% of patients with DCIS [10, 23, 24] and therefore raises the question of possible malignancy. The 10% proportion of DCIS in our patient population was lower than the expected 20-40% reported in the mammography literature [25, 26] and is likely attributed to a higher proportion of palpable masses in our population of patients referred for MR imaging. Therefore, caution must be exercised in attempting to extend our findings to patients with DCIS. The patient population referred for MR imaging at our institution differs from a patient population undergoing tissue sampling for mammographic abnormalities by a greater percentage of palpable lesions and invasive cancer. However, because DCIS accounts for as much as two thirds of false-negative results reported with dynamic MR imaging [27], our study is likely to overestimate the sensitivity of MR imaging in patients with mammographically identified lesions alone.
The low frequency of fibroadenomas or proliferative changes in our study and in other MR imaging study populations supports a general referral bias [28]. It also reflects the fact that patients with mammographic or sonographic features suggesting probably benign lesions, such as fibroadenomas, are often treated on the basis of mammographic or sonographic features alone. Mammographic treatment by periodic surveillance rather than MR imaging or tissue diagnosis of probably benign lesions in part accounted for the high pretest probability of malignancy in our study population (60%). The primary role of MR imaging at our institution has been as a complementary tool to mammography, sonography, and palpation—a tool that is usually recommended if imaging findings are inconclusive or further staging is desired before tissue diagnosis. From a cost-effective point of view, this is likely to be the role of MR imaging in the future.
Lesion characterization is important for tumor staging. Patients with breast cancer often have multifocal or multicentric disease. Diagnostic criteria that improve lesion characterization might contribute to the diagnosis of incidentally enhancing lesions. We were unable to examine the incidentally enhancing lesions in our population because of the lack of histopathologic sampling, difficulties in obtaining MR imaging follow-up examinations, and unavailable MR imaging-guided biopsies.
In conclusion, the morphologic characteristic of the margin and an enhancement pattern of washout emerged as the most useful MR imaging parameters for the characterization of suspicious breast lesions using a three-time-point method of acquisition. Both parameters are reproducible and easy to learn. In patients with suspicious breast lesions, the following diagnostic MR imaging criteria are proposed for lesion characterization: lesions with spiculated margins and lesions with washout and nonsmooth margins are classified as malignant; lesions without washout or spiculated margins and lesions with washout and smooth margins are classified as benign.
|
TopAbstractIntroductionMaterial and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  N. H. G. M. Peters, I. H. M. Borel Rinkes, N. P. A. Zuithoff, W. P. T. M. Mali, K. G. M. Moons, and P. H. M. Peeters Meta-Analysis of MR Imaging in the Diagnosis of Breast Lesions Radiology, December 1, 2007; 246(1): 116 - 124. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Kuhl The Current Status of Breast MR Imaging * Part I. Choice of Technique, Image Interpretation, Diagnostic Accuracy, and Transfer to Clinical Practice Radiology, August 1, 2007; 244(2): 356 - 378. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. C. Williams, W. B. DeMartini, S. C. Partridge, S. Peacock, and C. D. Lehman Breast MR Imaging: Computer-aided Evaluation Program for Discriminating Benign from Malignant Lesions Radiology, July 1, 2007; 244(1): 94 - 103. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Pediconi, C. Catalano, A. Roselli, S. Padula, F. Altomari, E. Moriconi, A. M. Pronio, M. A. Kirchin, and R. Passariello Contrast-enhanced MR Mammography for Evaluation of the Contralateral Breast in Patients with Diagnosed Unilateral Breast Cancer or High-Risk Lesions Radiology, June 1, 2007; 243(3): 670 - 680. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H Nishida, T Kinoshita, N Yashiro, Y Ikeda, and T O'Uchi MR findings of granulocytic sarcoma of the breasts. Br. J. Radiol., September 1, 2006; 79(945): e112 - e115. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Tozaki, T. Igarashi, and K. Fukuda Breast MRI using the VIBE sequence: clustered ring enhancement in the differential diagnosis of lesions showing non-masslike enhancement. Am. J. Roentgenol., August 1, 2006; 187(2): 313 - 321. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. K. Kuhl, P. Jost, N. Morakkabati, O. Zivanovic, H. H. Schild, and J. Gieseke Contrast-enhanced MR Imaging of the Breast at 3.0 and 1.5 T in the Same Patients: Initial Experience Radiology, June 1, 2006; 239(3): 666 - 676. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Bartella, L. Liberman, E. A. Morris, and D. D. Dershaw Nonpalpable mammographically occult invasive breast cancers detected by MRI. Am. J. Roentgenol., March 1, 2006; 186(3): 865 - 870. [Abstract] [Full Text] [PDF]
|
|
