| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Clinical Observations |
1 Present address: Department of Diagnostic Radiology, Suedharz-Hospital
Nordhausen, Dr. Robert-Koch Str. 39, 99734 Nordhausen, Germany.
2 Institute of Diagnostic and Interventional Radiology,
Friedrich-Schiller-University Jena, 07747 Jena, Germany.
Received July 7, 2004;
accepted after revision October 18, 2004.
Address correspondence to A. Malich
(ansgar.malich{at}shk-ndh.de).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CONCLUSION. New MRI features, especially from T2-weighted images, are promising for more reliable and accurate interpretation of breast lesions. Prospective studies of these findings are required to define cut-off values and test clinical practicality.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To achieve unified and objective image interpretation, Fischer et al. [9] introduced a scoring system implementing morphologic and dynamic features to more accurately classify breast lesions. They aimed to point out the benefit of MR imaging in the preoperative planning of breast lesions and used a score for semiquantitative analysis. The signs and given scores are shown in Table 1.
|
When tested prospectively, initial use of this scoring system showed moderate results for sensitivity and specificity [15]. Using cut-off values of 6 and more for malignancies and 3 or less for benign lesions, a sensitivity and a specificity of 47.2% and 12.2%, respectively, was reached on exclusively MRI-detected suspicious breast lesions when applying the scoring suggested by Siegmann et al. [16]. Considering a sum of 5 points to be a sign of malignancy, the sensitivity increased to 66%. The authors concluded that "the used score system can separate lesions with a high positive biopsy rate (6-8 points) which rather demand an excisional biopsy after MRI-guided lesion localization than MR-guided core biopsy." Although the highly selected population and the inclusion of mainly small lesions in this study have to be considered as main reasons for low accuracy, the implementation of further signs (including benign signs) may be helpful to increase the diagnostic power of such a score system.
Further potentially helpful morphologic and dynamic signs of malignancy in breast MRI were evaluated in this study, including T2-weighted images. We supposed that there are further signs of breast lesions on MRI, especially on T2-weighted images, and that their prevalence in malignant versus benign lesions differs significantly. If they do, there is a discriminatory power of these features that could allow extended analysis of those lesions that are difficult to clarify using already established signs. After analyzing prevalence and potential clinical outcome of these new signs, their potential benefit was evaluated. The goal was to analyze the prevalence and potential clinical outcome of these new signs and, if they were potentially beneficial, suggest an extension of the reported MRI scoring.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To obtain consistent histologic data, the same pathologist was always used. Postsurgical surveillance was limited to patients from our women's hospital who were enrolled in the study, so that 1,035 patients examined by MRI mammographies during this time were included in our investigation retrospectively.
Only MRI examinations showing an enhancing mass (focal enhancement) were used (817 lesions of 1,035 patients), and only those MRI mammographies with histologic clarification were enrolled. Seven hundred and ninety three of these 817 lesions were histologically verified after surgical intervention; 514 were malignant and 279 were benign lesions. In the remaining 24 cases, no such verification was performed because patients did not give their assigned agreement, incongruencies between MRI findings and the mammography were present, or sonogram and palpation occurred, yielding to the decision not to histologically verify the lesions. That is why in these cases, an unsuspicious follow-up (mean, 2.33 years) was performed mammographically. Nevertheless, these cases were not used in the study due to the lack of histologic verification.
To achieve the most reliable dynamic data possible, potential influencing aspects were reported and caused the exclusion of the following types of cases: chemotherapy before MRI mammography, a core and surgical biopsy up to 3 months before MRI mammography on the enhancing mass or at this site, radiation therapy, problematic motion artifacts (to avoid misinterpretations), patients under hormonal treatment (hormone replacement therapy), whereas those who were administered thyroid hormone treatment were included; and patients with incomplete or difficult-to-interpret imaging, or other technical problems during imaging.
Due to this strict study design, 641 cases were evaluated (Table 2). Two radiologists in consensus evaluated these cases retrospectively and were blinded to the histopathologic outcome.
|
A predefined technical protocol of MR imaging was used in all cases, differing somewhat from that given by Fischer et al. [9]. For example, T2-weighted imaging was included, and a 2D instead of a 3D acquisition technique was used. All MR images were obtained by a 1.5-T imager (Gyroscan ACS II, Philips) using a double-breast coil with the subject in a prone position. For the dynamic study, multislice 2D fast-field echo T1-weighted sequence images were obtained (TR/TE, 97/5.0; matrix, 205 x 256; flip angle, 80°; slice thickness, 4.0 mm; gap, 0.4 mm; field of view, 350 x 350 mm; axial orientation, 24 slices covering both breasts). After a native scan was obtained, 0.1 mmol/kg body weight of gadopentetate dimeglumine (Magnevist, Schering) was administered IV as a rapid bolus within 10 sec, followed by a 20 mL saline flush (30 mL saline flush if a dorsal hand vein was used). After bolus injection and saline administration, dynamic scanning was continued with the same sequence parameters and under identical tuning conditions at 1-min intervals for a total of 8 min. After the dynamic scan, the first coronal or axial T1-weighted scan was repeated approximately 10 min after contrast administration, followed by axial T2-weighted turbo spin-echo images in identical slice positions. Precontrast images were subtracted from postcontrast dynamic images.
The following signs were evaluated in addition to the suggested signs of MRI mammography scoring:
T2-weighted images—Only the supposed vital (enhancing) parts of the lesion were evaluated on T2-weighted imaging in comparison with normal breast tissue.
Blooming sign—Blooming effect describes a fast enhancing lesion with sharply shaped borders 1 min after bolus injection, becoming unsharp 7 min after contrast media administration, with the lesion fast enhancing during the first 2 min after bolus injection. Typical examples of a blooming sign and a nonblooming lesion are shown in Figures 1A, 1B and 2. T1-weighted subtraction images were used for analysis (Figs. 1A and 1B).
|
|
|
Hook sign—A hook sign symbolizes a hooklike spiculated dendrite coming from the lesion's center, leading to the pectoral muscle, determined on T2-weighted images (Fig. 2).
Nipple line—The integrity of the nipple line was analyzed using native T1-weighted images.
Edema—Using T2-weighted images, the presence of perifocal, unilateral, or bilateral edema in the enhancing lesion was analyzed (Figs. 3 and 4).
|
|
|
Adjacent vessels—Vessels leading to a lesion that was prominent in size were looked for on subtracted T1-weighted images.
The prevalence of these signs and the signs already scored by Fischer et al. [9] (Table 1) were calculated in benign and malignant lesions. In recent publications, the sensitivity and specificity of this score was estimated to vary from 36% to 48% [15, 16].
To adapt the new morphologic and dynamic signs to the already described ones, we used the following procedure. The prevalences were compared, and a factor created as follows: the prevalence (occurrence rate) of a special feature in malignant findings was divided by the prevalence of that same feature in benign lesions.
In case of a ratio of 1.5-3 (the evaluated MRI feature was at least 1.5 to a maximum of 3 times more common in malignancies) the analyzed feature was scored with 1 point, as is the case for ill-defined borders and plateau phenomenon, scored with 1 point by Fischer et al. [9]). This range from 1.5 to 3 was selected because these values fit best with the values of the signs scored by Fischer et al. A feature weighted with 1 point in the scoring system of Fischer et al. was ill-defined margins of contrast-enhancing lesions. In our study, this finding occurred in 84% of the malignancies and 49.5% of the benign findings, suggesting almost a doubling of the prevalence in malignant lesions using this phenomenon exclusively (ratio, 1.7). That is why all signs having similar ratios of prevalence were scored with 1 point.
Wash-out is a feature weighted with 2 points in the scoring system of Fischer et al. [9]. In our study, prevalence of this sign as the strongest given dynamic feature of malignancy was 70% in malignant lesions and 22.2% in benign findings, suggesting that this sign is about 3 times more common in malignancies versus benign findings (ratio, 3.2). That is why all other signs that were more common in malignancies to a comparable extent received 2 points as well. This means that a special lesion at least 3 times and a maximum of 4.5 times more common in malignant versus benign findings was rated with 2 points.
Features that were more than 4.5 times common in malignancies versus benign lesions were strong signs of malignancy and, therefore, were given 3 points (not recommended in the scoring system of Fischer et al. [9]). As a further extension to the already published diagnostic system, strong signs of benignity were also considered and scored as follows. If a special feature was more common in benign findings (range, 1.5-3 times), it was weighted with -1 as a weak sign of benignity. If a sign was at least 3 times more common in benign lesions (up to 4.5 times) it was rated with -2 points as a stronger sign of benignity; lesions that were even more common in benign findings received -3 points. Both the number of cases and the ratios and suggested points are shown in Tables 2 and 3, respectively.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In contrast, vital parts of malignant lesions were seldomly hyperintense (2%), in contrast with benign lesions (20%). The ratio of both prevalences was about -10, which is why hyperintensity of the vital parts in a lesion on T2-weighted images strongly suggested a benign histology (-3 rating).
Lymph Node Status (T2-Weighted Imaging)
The lack of axillary visible lymph nodes and the existence of subtle lymph
nodes were almost as common in malignant lesions (83%) as in benign lesions
(93%), and therefore were not a relevant parameter in the discrimination of
lesions.
In contrast, larger nodes (>10 mm) were more common in malignant (17%) versus benign cases (7%). The relation between prevalence of benign and malignant lesions was measured as 2.4, graded as a minor sign of malignancy (+1).
Edema
Lack of edema was more common in benign versus malignant lesions (91% vs
45.3%). The ratio of prevalence was -2, classified as a benign sign (-1).
The existence of bilateral or diffuse edema was, in contrast to the aforementioned sign, more common in malignant versus benign cases (14% vs 7%) and therefore considered a weak sign of malignancy (+1).
The existence of uni- and perifocal edema was found in 41% of malignant cases and only in 12% of benign lesions (ratio, 3.4). It was given 2 points as a sign in favor of malignancy.
Septations Within Lesions
Most benign and malignant lesions had no inner septations (80% and 98%,
respectively). Documenting the lack of septations is an unspecific sign.
Enhancing septations occurred in 1% of the malignant and 2% of the benign
lesions (ratio, 2; minor sign of benignity, -1 point). Nonenhancing lesions
were more common in benign lesions (18%) versus 1% of the malignant lesions
(ratio, -18), strongly suggesting a benign lesion (-3 points).
Blooming Sign
Blooming sign occurred in 63% of the evaluated malignant and 14% of the
benign lesions (ratio 4.5; 2 points as a stronger sign of malignity).
Hook Sign
Hook sign occurred in 33% of the malignant and 5% of the benign lesions.
The ratio of prevalence was 6.6, suggesting it was a strong sign of malignancy
(+3 points).
Nipple Line
The interruption of the nipple line seldomly occurred (overall prevalence,
11.9%, 76/641 cases) and occurred mainly in malignant lesions (14%) versus
benign ones (7%) (ratio, 2.0). It was a weak sign of malignancy, rated as
+1.
Skin Thickening in an Untreated Breast
Although uncommon in general (prevalence, 7.3%, 47/641 patients), skin
thickening was more common in malignant than benign cases (9% vs 3%) (ratio,
3.0) and as a sign of malignancy was estimated with a +1 rating.
Adjacent Vessels
Adjacent vessels occurred in 63% of the malignant and 28% of the benign
lesions (ratio, 2.3). It was a minor sign of malignancy, estimated with a +1
rating.
A summary of the new and well-established signs and their valuations are shown in Tables 2 and 3, respectively.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Some typical signs, such as plateau phenomenon and wash-out and fast initial wash-in effect have been reported as highly sensitive signs, whereas only a few publications report the prevalence and accuracy of other specific MRI features of malignant breast lesions, such as blooming phenomenon [20].
In this retrospective study, further signs of malignancy in breast MRI were analyzed. The highest potential to differentiate benign from malignant lesions was found for hyperintense T2-weighted signal intensity of the vital tumor parts, and septations without enhancement, both typical of benign lesions; and unifocal edema and hook sign, both strongly indicating malignancy.
In contrast with the findings of Stelling [21] and Orel et al. [22], T2-weighted information in our study, including the density of the vital tumor, hook sign, and the perifocal edema, seemed to be a promising parameter in the differentiation of malignant from benign lesions, confirming the findings of Kuhl et al. [23] and Helbich [24].
The total number of studies dealing with T2-weighted-images in MRI mammography, however, is low. The majority of these studies focus on the description of edema, which is, according to our findings as well, a strong sign of malignancy if it occurs perifocally.
In our study, T2-weighted signal intensity of the vital tumor compartments was analyzed. This signal intensity is mainly influenced by the degree of fatty and liquid tissue and the degree of fibrotic compartments [23]. This suggests a hyperintensity of T2-weighted images in only very rare malignant findings such as mucinous cancers [23], whereas the majority of malignant lesions are hypointense on T2-weighted images, corresponding to our findings.
A low signal intensity of vital tumor compartments suggests a collagen-rich fibrotic tissue with a low water and fat content [23] that is more common in malignant versus benign findings. Compared with the findings of Kuhl et al. [23], in our study the number of hypo- or isointense malignant findings on T2-weighted imaging was slightly higher (87% vs 91.9%), which might be caused by the analysis of the vital tumor parts in our study exclusively. This procedure allows the discrimination of necrotic components, which can mimic a falsely high intensity of T2-weighted images as reported by Kawashima et al. [25]. Furthermore, used TR and TE differs between studies from Kuhl et al. [23] and ours (4,000 msec vs 2,450 msec; 300 msec vs 110 msec, respectively) that may cause a different interpretation of T2-weighted signal intensity in borderline cases, as suggested by Wielopolski et al. [26].
A possible explanation of the hook sign could be the inclusion of the Cooper ligaments in the malignant process. Intraductal processes might cause a hooklike imaging by the development of fibrotic tissue. The potential to induce desmoplastic processes is, however, varying and especially high in invasive cancers [27]. From that point of view, this sign might be of special use in invasive malignancies, whereas noninfiltrating cancers usually do not show this sign. However, it should be noted that scars after surgical treatment, biopsy, and minimally invasive diagnosis and treatment might mimic a hook sign by fibrotic, inflammatory, or reparative processes [28, 29]. This is why it is essential to address whether a breast has already received any kind of treatment. In our study, only cases not treated and not invasively diagnosed before the MRI were included. A variety of mechanisms can cause edema, e.g., an imbalance of filtration and reabsorption in association with trauma, inflammation, heart insufficiency, and thrombosis, among others.
It can also be assumed that unifocal edema is more common in malignant lesions than bilateral edema. This might be attributed to the recent finding that vessels induced by tumor neoangiogenesis have a less intact basal membrane, which causes a higher penetration of mainly hydrophile, low-weight substances [30], resulting in perifocal edema. Furthermore, proliferating cells produce in a high number endothelium growth and vascular permeability factor; in that way, they increase permeability of surrounding capillary structures [31].
Blooming phenomenon was found to be as typical of malignancies as wash-out. The interpretation of the blooming sign is, at least in part, subjective and requires further objective analysis, as suggested by Fischer et al. [20].
The study and the analysis of the findings were adapted to the score suggested by Fischer et al [9]. This is why a further correction of the suggested score by Fischer et al. is neither required nor useful.
There are some further critical aspects and limitations of the study that have to be addressed. First, because of the heterogeneity of study protocols used in breast MRI, it cannot be assumed that our findings are per se reproducible on other systems, because data are strongly associated with the technology used. Related prospective studies are necessary to verify our findings. A gold standard of how to perform an MRI of the breast is necessary to achieve comparable results.
The interobserver variability of the evaluated signs has not been tested yet, which might further limit the clinical value. The interpretation of used characteristics remains, at least in part, subjective. That is why a computer-assisted analysis would increase objectivity and may simplify MRI analysis. Such technology is under investigation.
The study concept was retrospective and the analysis was performed in a condensed time schedule. The clinical value of this extension of scoring is still unclear and has to be tested prospectively to determine a cut-off value that suggests a malignancy when reaching a value above this cut-off, also according to BI-RADS classification.
Examinations were blinded to the observers who had not seen the images or cases before. It is well known that interpretation of breast MRI critically depends on the experience of the radiologist. In our study, experienced radiologists (more than 1,000 breast MRI examinations) exclusively diagnosed the images using a predefined protocol. When used by less experienced radiologists, the incidence might change.
An objective quantification of the border sharpness during enhancement was not possible because of the retrospective study design. By implementing such a technique (setting various regions of interest) in future prospective studies, the aspect of subjectivity can be significantly reduced. Consequently, rather subtle effects that remain undetected might be quantifiable.
Time of analysis of images might increase by using the aforementioned further signs. Due to the study design, nonenhancing and, therefore, MRI mammographically occult lesions could not be detected. However, an increased classification and accuracy in the interpretation of enhancing equivocal lesions might be possible.
Finally, only enhancing lesions were included in the study. Consequently, this study does not solve the problems of falsely negative MRI procedures.
Even with these limitations, from the data captured in this study, the hypothesis has been affirmed: There are characteristics of breast lesions, especially on T2-weighted images, with a prevalence that differs between malignant and benign findings. Using these signs (e.g., hook sign, edema, septations, and lymph nodes), the accurate characterization of breast lesions on MRI can be improved. If there is a lesion with uncertain characteristics, it is helpful to take other features, such as hook sign, blooming, and T2-weighted features (especially septations and perifocal edema), into account.
In addition, the results spur the next study and hypothesis that implementation of these signs in a scoring system will increase the rather moderate sensitivity and specificity when using the already established score exclusively.
This article reports some new morphologic and dynamic signs in benign and malignant breast lesions showing a focal enhancement. MRI features such as blooming sign, hook sign, hypo- and isointensity of vital tumor parts on T2-weighted images, perifocal edema, and associated larger axillary lymph nodes might be helpful in the discrimination of breast lesions, because their prevalence differs significantly in benign and malignant findings. These aspects were taken into account and an extension of the recently published scoring of MRI features was suggested to improve objectivity and uniformity of breast lesion evaluation. Some aspects regarding this study are, however, still under evaluation. According to the results, it is recommendable to take T2-weighted images—that is, perifocal unilateral edema, a hooklike spiculation of the lesion with association to the pectoral muscle, enlarged axillary lymph nodes as signs of malignancy and septations as benign signs— into account when a diagnosis is not sufficiently possible using already well-established signs.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  T. A. Heusner, S. Kuemmel, L. Umutlu, A. Koeninger, L. S. Freudenberg, E. A.M. Hauth, K. R. Kimmig, M. Forsting, A. Bockisch, and G. Antoch Breast Cancer Staging in a Single Session: Whole-Body PET/CT Mammography J. Nucl. Med., August 1, 2008; 49(8): 1215 - 1222. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Moy, F. Ponzo, M. E. Noz, G. Q. Maguire Jr., A. D. Murphy-Walcott, A. E. Deans, M. T. Kitazono, L. Travascio, and E. L. Kramer Improving Specificity of Breast MRI Using Prone PET and Fused MRI and PET 3D Volume Datasets J. Nucl. Med., April 1, 2007; 48(4): 528 - 537. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. J. Macura, R. Ouwerkerk, M. A. Jacobs, and D. A. Bluemke Patterns of Enhancement on Breast MR Images: Interpretation and Imaging Pitfalls RadioGraphics, November 1, 2006; 26(6): 1719 - 1734. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. G. ODLE Breast MR Radiol. Technol., September 1, 2006; 78(1): 45M - 66M. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. R. Fischer, A. Malich, S. Wurdinger, J. Boettcher, M. Dietzel, and W. A. Kaiser The adjacent vessel on dynamic contrast-enhanced breast MRI. Am. J. Roentgenol., August 1, 2006; 187(2): W147 - W151. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
