| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Radiology, Memorial Sloan-Kettering Cancer Center, Cornell
University Medical College, 1275 York Ave., New York, NY 10021.
2
Department of Physics, Memorial Sloan-Kettering Cancer Center, Cornell
University Medical College, New York, NY 10021.
Received December 27, 2000;
accepted after revision March 16, 2001.
Presented at the annual meeting of the American Roentgen Ray Society,
Washington, DC, May 2000.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. In 82 patients, 101 masses, including 68 infiltrating carcinomas and 33 benign lesions, were interpreted independently by four radiologists and described by BI-RADS terminology with respect to mass shape and margin and BI-RADS final assessment categories. The enhancement pattern of the mass was also reported. In addition, two radiologists interpreted each case twice to evaluate intraobserver variability. The final case set for analysis was the 68 infiltrating carcinomas.
RESULTS. Most of the infiltrating carcinomas were described as irregular, spiculated, and heterogeneously enhancing masses. The final impression of the 68 carcinomas was BI-RADS category 5 (highly suggestive of malignancy) in 41 (61%), category 4 (suspicious abnormality) in 24 (35%), and category 3 (probably benign) in three (4%). Enhancement pattern was heterogeneous in 40 (59%), homogeneous in 14 (21%), and rim in 14 (21%). Interobserver agreement was moderate for mass margin, shape, enhancement, and final assessment category.
CONCLUSION. This study suggests that the mammographic BI-RADS lexicon with some modifications may be applied to describe the features of infiltrating carcinoma seen on breast MR imaging.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A standardized terminology has been developed to facilitate the interpretation and communication among physicians in the description of morphologic findings on mammography. The Breast Imaging Reporting and Data System (BI-RADS) [6], in the form of a lexicon, is widely used today and is now the standard for reporting mammograms. The use of morphologic criteria has been shown to improve specificity when used to describe breast findings on MR imaging [7, 8]. For example, smooth or lobulated margins on MR imaging show a 97-100% predictive value for benignity, whereas the presence of rim enhancement shows a 79-92% predictive value for malignancy [7]. Despite these encouraging results, the usefulness of morphologic criteria to describe MR imaging—detected lesions has been limited by the lack of a definitive classification scheme. Work in this area is currently under investigation by an international team of breast MR imaging investigators led by Schnall and Ikeda [9] to develop a lexicon for breast MR imaging reporting that would allow description of all types of enhancement identified on MR imaging.
In the absence of a current specific MR imaging lexicon, our study was undertaken to assess whether the descriptive terminology and final assessment categories of the BI-RADS lexicon [6] could be used in a more limited setting when describing breast masses detected on MR imaging. The BI-RADS terminology was chosen because it is established and familiar to breast imagers and is now the standard for mammographic reporting. We chose to limit our evaluation to masses in a pilot project to test the feasibility of applying BI-RADS terminology to MR imaging; masses seen on MR imaging should be similar to masses seen on mammography. Because the mammographic BI-RADS terminology does not contain a descriptor for the range of enhancement patterns found on MR imaging, we added terms to describe enhancement patterns associated with the masses, because the pattern of enhancement is a prominent morphologic feature.
We function at a cancer center that has a great number of new cancer cases each year, and most of our cases are malignant lesions. Because our patient population is skewed, we chose to focus on infiltrating carcinomas, both ductal and lobular, to see whether BI-RADS could help us not only characterize these lesions, but also assign an appropriate final assessment category so that biopsy would be performed. Also of use to the breast imager is an analysis of the morphologic characteristics of infiltrating carcinomas found on MR imaging.
Above all, a lexicon should be able to help ensure that lesions requiring biopsy are recognized. The BI-RADS lexicon uses descriptors to attach a level of suspicion to a lesion and its need for a biopsy. We wanted to evaluate whether this same approach could work for masses seen on MR imaging. In addition, our study assessed inter- and intraobserver variabilities in the use of the BI-RADS descriptors and final assessment categories.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Of the 101 masses that occurred in 82 patients, 68 (67%) were invasive carcinomas and constitute the basis of this study. The remaining 33 masses (33%) were benign. When retrospectively identifying the masses for inclusion in this study, we did not know whether the masses proved to be benign or malignant at pathology.
Mean patient age was 54 years (range, 24-88 years). Of the 68 invasive carcinomas evaluated, 35 (51%) were seen on mammography, 16 (24%) were clinically and mammographically occult, 12 (18%) were detected both clinically and mammographically, and the remaining five (7%) were detected clinically.
Histology of the 68 masses found invasive ductal carcinoma in 44 (65%), invasive lobular carcinoma in 12 (18%), mixed invasive lobular and ductal carcinoma in five (7%), poorly differentiated invasive carcinoma in six (9%), and spindle cell carcinoma in one (1%).
All examinations were performed on a 1.5-T magnet (Signa; General Electric Medical Systems, Milwaukee, WI) after patients gave informed consent. Unilateral examinations using a dedicated breast coil were performed with the patient in the prone position and the breast uncompressed. After an axial localizing sequence, sagittal T1-weighted MR images using a three-dimensional fat-suppressed gradient-echo pulse sequence were acquired before and after IV injection of gadolinium dimeglumine (Magnevist; Berlex, Wayne, NJ). A contrast dose of 0.1 mmol/kg of body weight was used and was delivered in 10 sec with a hand injection. At the beginning of the study, the imaging parameters included a fast-spoiled gradient-echo technique with a TR/TE of 25/5 and a flip angle of 30° (n = 34). The time of acquisition was 2-3 min. This imaging sequence was then replaced by an enhanced fast gradient-echo sequence using the following parameters: minimum TR/TE, 6.5/1.6; flip angle, 10°; inversion time, 30 msec (n = 34); slice thickness with no gap, 2-3 mm; matrix, 256 x 192; number of excitations, 2; and field of view, 18-22 cm. Sagittal T2-weighted images were also obtained using a fast spin-echo sequence with fat saturation and the following parameters: TR/TE, 4000/120; echo train, 12; slice thickness, 3-4 mm; section gap, 1 mm; matrix, 256 x 192; number of excitations, 2; and field of view, 18-22 cm. The time of acquisition was 90 sec-3 min. After acquisition, subtraction of the unenhanced images from the contrast-enhanced images was performed on a workstation (Advantage Windows; General Electric Medical Systems) for better lesion conspicuity. All lesions were documented and reviewed on hard-copy film.
Four radiologists (A, B, C, and D) with different levels of experience independently interpreted the examinations. Three radiologists (observers A, B, and C) specialized in breast imaging and one radiologist (observer D) specialized in body MR imaging. Before participating in this study, radiologist D was trained in the BI-RADS lexicon by radiologist A. Radiologists A and D had daily experience in interpreting breast MR imaging examinations. Radiologists B and C were mammographers who did not interpret breast MR imaging examinations, although they used the BI-RADS lexicon on a daily basis. Radiologist D collected all the cases for analysis and marked the lesions on the hard-copy film. When radiologist D participated in the interpretation, approximately 3 months had passed since she had collected the cases for the study.
The observers had the entire hard-copy film examination available when interpreting the studies. All observers were kept unaware of lesion pathology. The hard-copy radiograph was marked with a grease pencil to indicate the pertinent index mass to be described. The observers in the study reviewed a mixture of cases containing both benign and malignant masses so that all BI-RADS final assessment categories would be used and not all lesions would be recommended for biopsy. The mixed cohort of 101 masses included both malignant (n = 68) and benign masses (n = 33). Although the observers interpreted both the benign and malignant masses, the case group for this study consisted of only the 68 malignant masses.
All observers were unaware of the pathologic results and clinical information. Two radiologists (A and B) interpreted each case twice to evaluate intraobserver variability. The second interpretation occurred approximately 6 months after the first. Each radiologist interpreted the findings from the same hard-copy radiograph and recorded their findings on a data sheet that had boxes to check for margin, shape, enhancement pattern, and final assessment category.
Masses were classified according to the BI-RADS descriptors for features including mass margins (circumscribed, microlobulated, indistinct or ill-defined, spiculated, and obscured), mass shape (round, oval, lobulated, irregular, and architectural distortion), and final level of suspicion (BI-RADS categories 1-5). Only one descriptor per category was used.
Because the BI-RADS lexicon was developed for mammographic lesions, there was no descriptor for the contrast enhancement pattern in the lesion. Taking this information into account, a separate enhancement pattern descriptor was added to the evaluation and classified as heterogeneous, homogeneous, rim enhancement, enhancing internal septations, and nonenhancing internal septations (Table 1).
|
If present, associated linear or ductal enhancement was noted; however, the radiologists were instructed to describe only the indicated focal mass lesion. Other associated features, such as edema, skin or nipple retraction, skin thickening, fluid-filled ducts, abnormal signal void, lymphadenopathy, chest wall or skin invasion, presence of cysts, presence of hematoma or blood, architectural distortion due to postoperative changes, and artifact due to surgical clips were also documented. Associated features were not independently evaluated in this study, although the presence of any suspicious associated features might have influenced the radiologists' final assessment category.
A consensus result was obtained using the agreement of more than two radiologists for a particular feature of each lesion. Inter- and intraobserver variabilities for each descriptive category were determined using the kappa statistic. The kappa statistic represents the proportion of agreement beyond that expected by chance and is the most widely accepted measure of agreement for data arising from nominal or ordinal scales. Although no absolute scale exists, previous reports suggest that kappa values greater than 0.80 represent almost perfect agreement; 0.61-0.80, substantial agreement; 0.41-0.60, moderate agreement; 0.21-0.40, fair agreement; and 0.20 or less, slight agreement [10].
Statistical analysis for significance between enhancement pattern and tumor size was performed using the two-tailed Fisher's exact test using statistics software (EPI-INFO; Centers for Disease Control, Atlanta, GA).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
The shape of the 68 infiltrating carcinomas was described as irregular in 39 (57%) (Fig. 3), lobulated in 16 (24%), round in eight (12%), and oval in five (7%). No carcinomas were defined as an architectural distortion.
|
The enhancement pattern of the 68 infiltrating carcinomas was heterogeneous in 40 (59%), homogeneous in 14 (21%) (Fig. 4), and peripheral rim in 14 (21%) (Fig. 5). No enhancement patterns were described as enhancing internal or nonenhancing internal septations. The mean size was 1.1 cm (range, 0.5-2.0 cm) for the 14 homogeneously enhancing masses, 2.2 cm (range, 0.4-6.2 cm) for the heterogeneously enhancing masses, and 2.2 cm for the rim-enhancing masses (range, 1.1-3.8 cm). Carcinomas less than or equal to 1 cm were significantly more common in the homogeneously enhancing group than in the heterogeneously or rimenhancing groups (p = 0.007) as determined by the two-tailed Fisher's exact test.
|
|
Most of the infiltrating carcinomas were described as irregular, spiculated, and heterogeneously enhancing masses (Fig. 6). Among the different histologic groups, we found that the most common features of the 43 ductal carcinomas were irregular shape in 24 (56%), spiculated margins in 20 (47%), and heterogeneous enhancement in 26 (60%). Associated ductal carcinoma in situ was found in 21 (49%) of the 43 pathologic specimens, although the association of the focal mass with other forms of enhancement, particularly linear enhancement, was found in only five (12%). In 39 masses for which pathologic measurement was available, the mean tumor size was 2.1 cm (range, 0.2-8.2 cm).
|
Among the 12 pure lobular carcinomas, the most prevalent features were irregular shape in 10 (83%), spiculated margins in eight (67%), heterogeneous enhancement in six (50%), and homogeneous enhancement in six (50%) (Fig. 7).
|
The most prevalent features in the 17 mixed lobular and ductal carcinomas were irregular shape in 12 (71%), spiculated margins in 10 (59%), and heterogeneous enhancement in eight (47%).
The BI-RADS final impression of the 68 infiltrating carcinomas was category 5 (highly suggestive of malignancy) in 41 (60%), category 4 (suspicious abnormality) in 24 (35%), and category 3 (probably benign) in three (4%) (Table 1).
The three lesions that received final BI-RADS category 3 (probably benign) showed homogeneous enhancement in two (67%); circumscribed margins in three (100%); and oval, lobular, or round shape in three (100%). These findings were not evident on physical examination or mammography. Surgery was performed in all three of these patients as a result of clinical suspicion or suspicious associated findings on MR imaging in the same breast. One patient had a history of ipsilateral infiltrating lobular carcinoma, and the mass was removed at mastectomy, yielding a separate focus of infiltrating lobular carcinoma (Fig. 7). A second patient presented with occult breast carcinoma and ipsilateral axillary metastasis and underwent mastectomy, revealing that the mass corresponded to the primary infiltrating ductal carcinoma. The third patient had another palpable focal mass with suspicious features within the same quadrant, and a wide surgical excision was performed for both masses, yielding infiltrating ductal carcinoma. The mean size of the BI-RADS category 3 lesions was 6 mm (range, 4-9 mm).
Rim enhancement is considered a suspicious morphologic feature [7, 8], and this finding appeared in 14 (21%) of the 68 infiltrating carcinomas (Fig. 5), with the following distribution: 10 (23%) of the 43 ductal carcinomas, three (60%) of the five mixed lobular and ductal carcinomas, and one adenocarcinoma with papillary features. No pure lobular carcinomas in this series exhibited rim enhancement.
Interobserver agreement among the four radiologists was moderate for mass
margins (
= 0.46 ± 0.03), mass shape ( = 0.41 ± 0.03),
enhancement pattern ( = 0.56 ± 0.03), and final assessment
category ( = 0.45 ± 0.03). Intraobserver agreement for two of the
radiologists was substantial for mass shape ( = 0.70). Moderate
intraobserver agreement was found for enhancement pattern ( = 0.58), mass
margin ( = 0.50), and final assessment category ( = 0.47).
Interobserver variability with respect to the imaging sequence used did not
differ.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To describe enhancement in focal infiltrating carcinomas, the terms "homogeneous enhancement," "heterogeneous enhancement," and "rim enhancement" were useful. Not surprisingly, no carcinoma was described as having nonenhancing internal septations, a descriptor associated with fibroadenomas. More importantly, the descriptor "enhancing septations," usually associated with malignancy, was not used by a single radiologist for any of the 68 cases of carcinoma. In our experience, this represented a rare pattern of enhancement for cancer.
With respect to margins, Nunes et al. [7] found that no malignant masses had "smooth" borders. However, this was not our experience. Mammographically circumscribed masses can be malignant, particularly the specific histologic subtypes of medullary, colloid, and papillary carcinoma. We found that 13 (19%) of 68 infiltrating carcinomas were described as circumscribed. This finding contrasts with mammography, in which 5-6% of malignant masses have been described as circumscribed [11, 12].
The greater number of circumscribed carcinomas seen on MR imaging in our study may be a result of a number of factors. First, the administration of IV contrast material allows sharp definition between the enhancing mass and the surrounding nonenhancing breast, thus allowing sharper definition of the mass with accentuation of the borders, which make the mass appear smoothly marginated to the observer. Second, although the imaging sequence was designed to provide high resolution (the spatial resolution that we used was 0.78 x 1.04 mm in-plane resolution), the resolution may not have been adequate to show surrounding small spicules. Spicules may be caused by fibrotic desmoplastic reaction and therefore may not enhance. Also, the technique used in this study was selected to allow imaging of the entire breast with acceptable spatial resolution to image in 2-3 min, the peak in the contrast enhancement [13]. Third, when the window settings are selected for the lesion and too much contrast material is used, the edges of the mass may blur, making it appear more circumscribed. These manipulations could not only blur the mass margin but could also change the enhancement characteristics of a mass, making a heterogeneously enhancing mass appear more homogeneous. Nevertheless, in our study, most of the lesions that were described as circumscribed had other more worrisome descriptors, such as heterogeneous or rim enhancement, that raised the radiologists' suspicion for malignancy.
Yet another reason for the high percentage of circumscribed malignant masses compared with mammography in our study may be the timing of the contrast-enhanced images. Several sets of contrast-enhanced images were acquired, and only those showing the strongest enhancement, usually the first acquisition after contrast injection in 2 min, were used for evaluation of morphologic features of the masses. This method was chosen because on the latter images, as the contrast material starts to wash out of the carcinoma and the background glandular tissue gains more enhancement, the margins of the carcinoma are difficult to assess and appear indistinct or obscured. The trade-off between high resolution and dynamic acquisition is a well-known compromise [1,2,3,4,5, 13].
The lack of kinetic data likely hampered our ability to classify suspicious lesions. However, on the basis of our results, it appears that morphologic criteria used alone were usually able to predict malignancy in 65 (96%) of the 68 cancers. Only three carcinomas were described as probably benign (BI-RADS category 3), and the addition of kinetic data most likely would have made a difference in interpretation and final assessment category in these three. In lesions that are suspicious for malignancy based on morphologic criteria (BI-RADS category 4 or category 5), the need to biopsy is based on suspicious morphology. Nevertheless, the addition of kinetic data undoubtedly would add important additional information that would aid in categorization, particularly for those lesions that are possibly benign. Practically speaking, we will plan to analyze the morphologic information so that if a lesion exhibits suspicious characteristics (i.e., spiculated margin or rim enhancement), we will recommend biopsy regardless of the kinetic data. However, if a lesion exhibits benign characteristics, such as in a circumscribed margin, kinetic information becomes invaluable in making the final decision to biopsy.
Of the three carcinomas that were given a BI-RADS category 3 final assessment, two of the three showed homogeneous enhancement; all three showed circumscribed margins; and all three had a round, oval, or lobular shape. The size of the three masses was small (4, 5, and 9 mm). All the descriptors used in these three cases are associated with a benign finding (category 2) or probably benign finding (category 3) in the final assessment categories. If these small masses had not been biopsied, a 6-month follow-up would have been performed. As with mammography, if a change were noted at that time, biopsy would be performed. The two smallest masses were likely so small that at the resolution we used, the enhancement appeared uniform and the margins smooth. Therefore, the issue for these two lesions was limitation of technique. On the other hand, the 9-mm lesion was large enough so that the morphologic features were clearer. This lesion was described in consensus as oval, circumscribed, and heterogeneous in enhancement. Perhaps, because of the heterogeneous enhancement, it should have been classified as BI-RADS category 4 instead of category 3. Its classification was more a result of the radiologist's lack of realization of the significance of heterogeneous enhancement than a lack of sufficient image resolution.
Of note, none of the 68 carcinomas was assigned a final assessment of category 2 by any of the radiologists. Also, most—65 (96%) of 68 carcinomas—were assigned a BI-RADS category 4 or category 5 (suspicious abnormality or highly suggestive of malignancy, respectively), necessitating biopsy, even by those radiologists who had no extensive experience interpreting breast MR images. It is logical that mammographers who are accustomed to evaluating breast lesions using BI-RADS terms on mammography can readily translate those skills to MR imaging. The assignment of a final assessment category, as in mammography, indicates to the referring physician what appropriate step should be taken next and what information should be included in the report. We also make sure that the breast MR imaging is interpreted in conjunction with all other pertinent breast imaging studies, such as mammogram and sonogram, to render a recommendation. The breast imager is in the best position to synthesize the results from all studies and arrive at a final assessment.
We found no significant differences among the most prevalent features of invasive cancer according to histologic subtypes. Both ductal and lobular subtypes showed irregular shape, spiculated borders, and heterogeneous enhancement as their prevalent features. Despite the small number of cases, no pure invasive lobular carcinomas presented with rim enhancement, supporting the theory that pure invasive lobular carcinomas will not show rim enhancement [14].
Most carcinomas showed heterogeneous enhancement, indicating nonuniformity of the tissue and varying rates of enhancement, depending on the area of the carcinoma. Twenty-one percent of carcinomas showed homogeneous enhancement. As carcinomas grow, they may become more pathologically heterogeneous, with areas of necrosis once the carcinoma outgrows the blood supply. When we compared the mean pathologic size of 65 (96%) of the 68 masses and the pattern of enhancement, we found that the mean size of carcinomas showing homogeneous enhancement was smaller than the mean size of carcinomas showing rim or heterogeneous enhancement (Table 1). Homogeneous enhancement was significantly more common in carcinomas measuring 1 cm or less than in larger carcinomas. Homogeneous enhancement was present in eight (44%) of 18 carcinomas measuring 1 cm or less versus six (12%) of 50 carcinomas greater than 1 cm (p <0.007).
Rim enhancement, a strong predictor of malignancy, was found in 20% of the 68 carcinomas in our series. It has been theorized that rim enhancement indicates a more vascular periphery in which growth of the tumor is occurring [15]. Rim enhancement would appear to be relatively straightforward in appearance on MR imaging and easily evaluated; however, in our study, rim enhancement did not perform significantly better than any of the other descriptors in terms of interobserver agreement.
There was moderate agreement among the four observers and moderate to good agreement after a second interpretation. This result is similar to results of agreement studies described by Baker et al. [16, 17], who used BI-RADS for mammography and description of lesions on sonography. Another study of the description of sonographically detected lesions using the descriptive terms described by Stavros et al. [18] found that interobserver variability ranged from slight to substantial for different descriptors. Therefore, the interobserver results from this study are within previously reported ranges.
The descriptors were useful in predicting patient treatment and the final assessment category. Although lesion description may vary from observer to observer, it is important that the final assessment category result in the correct action. How observers describe lesions may be less important than the final assessment category. Using mammography, Berg et al. [19] showed in 96 cases that lesion perception was more variable for masses than for calcifications, but lesion management was more consistent than description.
Another limitation of our study is that all four interpreting observers worked at the same institution, and one individual had trained the other three in mammography. Therefore, the results may reflect a higher agreement rate than would be expected if the study were carried out at multiple institutions. Further work is needed at other institutions to confirm our results. Additionally, this study was limited to focal breast masses, although nonmasslike enhancement is often found on MR imaging. A definitive lexicon that includes both morphologic and kinetic analyses is currently under development by the lexicon group of the International Working Group for Breast MRI [9, 20]. Additionally, a prospective study would be important to establish whether the final assessment category accurately reflects the need for biopsy.
This study suggests that the mammographic BI-RADS lexicon may be applied to breast MR imaging of focal carcinomas with some modifications. We found BI-RADS useful in the analysis of focal breast carcinomas, but further assessment of its usefulness by other institutions is needed.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  J. I. Wiener, K. J. Schilling, C. Adami, and N. A. Obuchowski Assessment of Suspected Breast Cancer by MRI: A Prospective Clinical Trial Using a Combined Kinetic and Morphologic Analysis Am. J. Roentgenol., March 1, 2005; 184(3): 878 - 886. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Liberman, E. A. Morris, M. J.-Y. Lee, J. B. Kaplan, L. R. LaTrenta, J. H. Menell, A. F. Abramson, S. M. Dashnaw, D. J. Ballon, and D. D. Dershaw Breast Lesions Detected on MR Imaging: Features and Positive Predictive Value Am. J. Roentgenol., July 1, 2002; 179(1): 171 - 178. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
