| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Clinical Observations |
1 St. Michael's Hospital, University of Toronto, Medical Imaging, 60 Bond St.,
Toronto, ON, Canada M5B 1W8.
2 Mount Sinai Hospital, University of Toronto, Toronto, ON, Canada.
Received June 18, 2004; accepted after revision October 15, 2004.
Address correspondence to D. Muradali
(muradalid{at}smh.toronto.on.ca).
OBJECTIVE. The purpose of this study was to describe the sonographic appearance and histopathologic basis of those malignancies of the breast that show nonenhancement on MRI.
CONCLUSION. Breast malignancies with suspicious features on sonography may not be detected on MRI due to nonenhancement of the lesions. In such cases, we suggest the use of tissue sampling to differentiate between benign and malignant breast lesions.
MRI has been shown to be highly sensitive in the detection of breast malignancies with reported sensitivities ranging from 83–100% [1–6]. This observation initially led imagers to presume that nonenhancing breast lesions on MRI were benign and did not warrant biopsies [7, 8]. However, previous studies have shown that dynamic enhanced MRI may not detect all breast cancers, with nondetection of up to 8.4% of invasive cancers [2, 3, 9, 10].
Teifke et al. showed that nondetection of cancers on enhanced MRI was due to either technical reasons or factors related to the intrinsic properties of the tumor [10]. Technical reasons were responsible for 17.9% of these undetected invasive cancers and included faulty contrast infusion, nonimaging of the tumors located outside the field of view, and motion artifacts leading to nonvisualization of the tumor on subtraction images. Tumor size, growth pattern, and strong enhancement of surrounding glandular parenchyma were considered responsible for the remaining 82.1% of invasive cancers that were not detected with MRI.
While the visibility and interpretation of breast lesions on MRI are highly dependant on contrast uptake by the lesion, detection and evaluation of lesions by sonography is independent of enhancement, and is based solely on a morphologic assessment. Therefore, it is possible that a subset of breast malignancies will be occult on MRI, but visible on sonography. Furthermore, there has been minimal literature describing the sonographic features of breast malignancies that are occult on MRI. The purpose of this study was to describe the sonographic appearance and histopathologic basis of those malignancies of the breast that show no enhancement on MRI.
Materials and Methods
Study Population
The study population was obtained from a subset of 100 consecutive patients
(median age, 58 years; range, 34–94 years) with 104 pathologically
proven malignancies of the breast, who presented to our department over an
18-month period (June 2001 to November 2002). None of these patients had in
situ carcinoma without an invasive component. The spectrum of breast
malignancies in these 100 patients is given in
Table 1. For each of these
patients, the index lesion initially was detected on either mammography and/or
sonography, and the indication for breast MRI was to assess for additional
tumor foci before surgery. A total of 104 MRI scans were reviewed from these
patients.
|
MRI protocol was standardized in all cases. Diagnostic enhanced breast MRI was performed at 1.5T (SIGNA LX Echo speed, GE Healthcare) using a phased-array, single dedicated breast surface coil with the patient in the prone position. The protocol included an axial localizer sequence followed by sagittal and axial fast spin-echo T2-weighted sequences. This was followed by pre- and postgadolinium-enhanced sagittal 3D fast spoiled gradient echo (FSPGR) sequences. Gadopentetate dimeglumine was administered IV at a dose of 0.16 mmol/kg body weight, followed by 20 mL saline flush (0.9%). Two consecutive enhanced sequences were obtained immediately following gadolinium injection. Scans were obtained at a slice thickness of 3 mm with no intersection gap. Field of view used was 16 + with a matrix of 512 x 256. Postprocessing included subtraction of the enhanced images from the unenhanced images for studying contrast enhancement. A semiquantitative analysis of the signal-intensity-to-time was performed with the region of interest technique.
In all patients, high-quality mammograms (obtained with different equipment and films due to different referral centers) were available. In addition, all patients underwent physician-supervised breast sonography with either an ATL HDI 5000 (8 MHz linear transducer, ATL Ultrasound) or a GE Logic 700 scanner (7.5 MHz linear transducer, GE Healtcare).
All studies were reviewed retrospectively by two breast imagers with an expertise in breast MRI. At the time of MRI review, mammographic and sonographic findings and pathology reports from all surgical and core biopsies were available. Patients with biopsy-proven malignant breast lesions, which showed absence of contrast uptake on MRI, were entered into the study.
Two patients with breast malignancies that were not detected on MRI due to technical reasons were excluded from the study. In one patient, the tumor nodule was not detected on MRI due to susceptibility artifacts caused by adjacent surgical clips and, in another patient, the malignancy was not visible as it was located at the edge of the field of view.
The study population was composed of eight patients (median age, 68 years; range, 48–89 years) with a total of nine nonenhancing breast malignancies on MRI. Seven of the eight patients presented with a solitary breast malignancy (seven of nine lesions). One of the eight patients presented with multifocal breast carcinoma with four malignant nodules detected on sonography, three of which showed enhancement on MRI. All four nodules were marked with guidewires before surgery, using sonographic guidance. At pathology, a fifth malignant nodule was identified, which was not detected on either sonography or MRI. Therefore, in this patient, two malignant nodules (two of nine lesions) showed no enhancement on MRI, and one of these nodules also was not detected on sonography.
Sonography Review
Images from physician-supervised breast sonography from all eight study
patients were reviewed by two breast imagers in a consensus reading.
Sonography studies were assessed in conjunction with the corresponding
mammograms.
On sonography, each lesion was evaluated based on size, shape (ellipsoid, round, irregular), margins (smooth, macrolobulated, microlobulated, spiculated, angulated, ill-defined), posterior sonographic artifact (shadowing, through transmission, no posterior artifact), echogenicity (hypoechoic, isoechoic, echogenic), presence of calcifications, and ductal extension. Lesions were classified as benign, indeterminate, or malignant based on the criteria described by Stavros et al. [11].
Histopathologic Correlation
Histopathologic correlation was determined based on surgical pathology
specimens in eight of the nine lesions (lumpectomy, six of nine; mastectomy,
two of nine) and tissue from sonography-guided core biopsy in one of the nine
lesions. For lumpectomies, lesions were localized with guidewires under either
sonographic (five of six) or mammographic (one of six) guidance. The position
of all localizing wires was confirmed on craniocaudal (CC) and mediolateral
(ML) mammographic projections before surgery. Specimen radiographs were
reviewed to ensure that the targeted lesion was resected.
Two patients underwent a unilateral mastectomy. In both cases, a solitary lesion was identified on preoperative imaging. One patient who underwent core biopsy was subsequently treated with chemotherapy based on the histologic results and did not undergo surgical excision.
Tumor grade, size, histologic subtype, and the presence of invasion were documented. The diameter of all malignant nodules that were resected surgically (eight of the nine) was measured at histopathology. Pathologic-imaging correlation was performed in conjunction with both pathologists and breast imagers with regard to lesion location and size to ensure that the imaged lesion was evaluated histologically.
Results
Of the nine malignant breast lesions that did not show enhancement on MRI, six of the nine lesions were detected on sonography (Table 2). Five of the six lesions showed frankly malignant features (Figs. 1A, 1B, 2A, 2B, 2C, 2D, 3, 4, 5), while one appeared indeterminate (Fig. 6A, 6B). The most common sonographic appearance of these lesions was that of a solid (six of six), hypoechoic (five of six) nodule associated with spiculated margins (three of six) and posterior acoustic shadowing (five of six). All these lesions were detected on mammography. The mammographic features of the lesions that were detected on sonography included spiculated mass (two of six) (Fig. 2A, 2B, 2C, 2D), asymmetric density (three of six), architectural distortion (two of six) and high-density oval mass with obscured margins (one of six). Histopathology of these six nodules revealed invasive ductal carcinoma (IDC) (two of six), tubular carcinoma (one of six), invasive lobular carcinoma (ILC) (one of six), mixed invasive ductal and lobular carcinoma (one of six), and follicular lymphoma (one of six).
|
|
|
|
|
|
|
|
|
|
|
|
Three breast malignancies, which did not enhance on MRI, also were not detected on sonography. While these nodules were occult on MRI and sonography, they were visible mammographically as a spiculated mass (one of three) (Fig. 7), pleomorphic calcifications (one of three), architectural distortion (one of three), and a new asymmetric density compared with prior studies (two of three) (Table 3). Histopathology of these sonographically occult lesions showed IDC (one of three), ILC (one of three), and mixed invasive ductal and lobular carcinoma (one of three).
|
|
As previously described, one of the eight patients presented with multifocal breast cancer, with two lesions (two of nine) that did not enhance on MRI. One of these lesions was detected on sonography, and the other was occult on sonography. At pathology, both lesions showed mixed invasive ductal and lobular carcinoma. On mammography, both of these lesions appeared as areas of asymmetric density associated with architectural distortion.
Discussion
Our results show that some breast malignancies that are occult on enhanced MRI may be visualized on sonography. In our series, six of the nine lesions that did not show enhancement on MRI were detected with sonography. Five of these were invasive cancers and showed one or more malignant features on sonography (angulated/spiculated margins, posterior acoustic shadowing), while one nodule, diagnosed as follicular lymphoma, was indeterminate on sonography (ill-defined margins, through transmission). None of these lesions showed benign features on sonography. Therefore, once visible on sonography, this subset of malignancies seems to present as either a frankly malignant nodule or an indeterminate lesion.
Our results are similar to those of Teifke et al. [10], who described a false-negative rate of 12% for breast cancer detection, using enhanced MRI. However, there are some differences in the methodologies between their paper and our series. The false-negative lesions described by Teifke et al. included those malignancies that were not detected due to the technical reasons, and those lesions that were not visualized due to the histologic features of the tumor and/or the surrounding tissues. However, in our study we excluded the lesions that were not detected due to technical factors, and included only those breast malignancies that were occult on MRI due to nonenhancement of the tumor. In addition, Teifke et al. used a 1.0 T magnet, as opposed to a stronger 1.5 T magnet as used in our study. The use of a 1.0 T magnet by these authors would not account for the lack of detection of malignancies in their series. Heywang-Kobrunner et al. [12] reviewed the diagnostic parameters for enhanced breast MRI in a multicenter study, and showed that the differences between 1.0 T and 1.5 T magnets were not statistically significant. Despite these differences in the study design, the results of both our study and that by Teifke et al. clearly show the existence of MRI occult breast malignancies, and emphasize the value of other imaging techniques in patient assessment.
Three of the nine malignancies that were occult on MRI also were occult on sonography. However, each of these lesions was detected on mammography, with classic mammographic malignant features as shown in Table 3, (pleomorphic calcifications, spiculated mass, architectural distortion, new asymmetric density). Boetes et al. [9] also described four breast malignancies that were not visible on enhanced MRI and sonography but were detected on mammography. These results therefore would suggest that breast lesions, which show concerning features on mammography and/or sonography, should be considered suspicious for malignancy despite a lack of enhancement on MRI. In such a scenario, it would seem reasonable that these lesions be further assessed with percutaneous core biopsy under sonographic or mammographic guidance in the correct clinical context.
Previous studies have shown that MRI is the most sensitive method available for the detection of invasive breast cancer when compared with mammography and sonography [10, 13]. Teifke et al. [10] showed that if used independently, MRI was more sensitive for detecting breast carcinomas when compared with sonography or mammography, with reported sensitivities of 88.4%, 80.5%, and 73.7%, respectively. However, the highest sensitivity for lesion detection (95.5%) was achieved when the results of all three techniques were combined [10]. Our study was not designed to evaluate the sensitivity of breast cancer detection on any of these techniques, however, since carcinomas that were occult on enhanced MRI in our study were detected on either sonography and/or mammography, our data also would support a multitechnique approach when assessing a suspicious lesion detected on imaging.
A negative MRI therefore should not influence the management of lesions that appear to be of concern on sonography, mammography, or clinical examination. In such a scenario, recommendations should be based on the appearance of the lesion on sonography and mammography, and on the clinical findings. Furthermore, MRI should not be used in place of high-quality mammographic work-up of abnormalities detected at screening, or as a replacement for sonography as a tool for evaluating equivocal findings on mammography [14]. MRI should function as an adjunct to sonography and mammography in the management of suspicious breast lesions as opposed to a substitute for these techniques.
Since there are multiple scanning protocols available for breast MRI, it is possible to speculate that the nondetection of malignant lesions observed in our study could be related the selected protocol, specifically the number of pulse sequences obtained after the injection of contrast medium. In our series, only two pulse sequences were performed after the injection of gadolinium, as opposed to four or five postcontrast sequences as used by other authors [10, 12, 15, 16]. However, Weinstein et al. [17] selected a protocol similar to the one used in our study, in that the entire breast was imaged twice in succession after the administration of contrast, and reported three cases of lobular carcinoma, which were not detected on MRI. In addition, Teifke et al. [10] selected a scanning protocol that entailed five enhanced sequences, and reported 20 cases of invasive carcinomas that showed no evidence of abnormal enhancement, and so were not detected. The results of these studies therefore would suggest that the nondetection of breast malignancies in our series was not related to the scanning protocol, but was secondary to the histopathology of the tumor nodules.
The physiologic basis for certain breast malignancies showing no enhancement on MRI is, to date, uncertain. In our series, two of the nine nonenhancing lesions were invasive lobular cancers, and another two of nine nonenhancing cancers showed mixed lobular and ductal features. The lack of contrast uptake by lobular carcinoma on MRI has been described previously by other authors. Boetes et al. [9] described a 40-mm invasive lobular cancer that showed nonenhancement on MRI, was occult on sonography, and was detected only on mammography. Weinstein et al. [17] also described a similar case of underestimation of lobular carcinoma by MRI. In that case, mammography and MRI showed a 1.5-cm and 1.0-cm mass, respectively, while pathology revealed diffuse infiltrative lobular carcinomas, with a span of 8 cm. In a series by Wurdinger et al. [18], four ILCs, which were detected on sonography, were missed on MRI due either to delayed enhancement or nonenhancement within the tumors. However, Qayyum et al. [15] described 13 cases of infiltrating lobular carcinoma, all of which were detected on enhanced MRI presenting either as a solitary mass, multiple enhancing foci, or enhancing septa. The enhancement pattern of these lesions on MRI correlated with the pathologic tumor morphology. Although poor enhancement was noted in one malignancy in this series, none of the tumors showed a complete lack of enhancement. Therefore, while in some cases lobular carcinoma may be visualized on MRI due to an abnormal pattern of enhancement, in other instances these malignancies may not be detected resulting in a false-negative study.
It is possible that the lack of contrast uptake by certain lobular carcinomas may be related to the pattern of cellular growth within these tumors. Lobular carcinomas have been shown to diffusely infiltrate into the adjacent breast parenchyma in a cell-by-cell fashion ("Indian File" growth pattern) [19]. It is possible that due to this single-file pattern of growth, preexisting parenchymal capillary networks may be sufficient to nourish the tumor cells via diffusion. Therefore, induction of a significant amount of neovascularity by these tumors may not be necessary for their growth. This paucity of surrounding angiogenic vessels could potentially account for the lack of enhancement observed on MRI.
Teifke et al. [10] suggested that infiltrating cancers associated with extensive intraductal component (EIC) and ductal carcinoma in situ (DCIS) might be difficult to detect on MRI. In our study, all three nonenhancing invasive ductal cancers were associated with DCIS, of which two were associated with EIC. However, there is only minimal literature describing enhancement patterns of invasive cancers associated with EIC and further studies are required to prove a definitive association between false-negative MRI and EIC.
While we acknowledge that our study is limited by the relatively small sample size, our results show that breast malignancies with suspicious features on sonography may not be detected on MRI due to nonenhancement of the lesions. While the clinical indications for breast MRI have yet to be precisely defined and universally accepted, the imager must be aware of the limitations of this technique. Furthermore, as the availability and clinical demand for breast MRI increases, one should not overestimate the ability of MRI to differentiate between benign and malignant lesions of the breast. Therefore, we suggest that breast lesions that appear to be of concern for malignancy on sonography due to their morphologic features be evaluated with tissue sampling in the correct clinical setting, despite a lack of enhancement on MRI.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Orel Who Should Have Breast Magnetic Resonance Imaging Evaluation? J. Clin. Oncol., February 10, 2008; 26(5): 703 - 711. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. G. ODLE Breast Ultrasound Radiol. Technol., January 1, 2007; 78(3): 222M - 242M. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
