The Lactating Breast: Contrast-enhanced MR Imaging of Normal Tissue and Cancer 1

Transcription

1 Breast Imaging Leandro A. Espinosa, MS Bruce L. Daniel, MD Logi Vidarsson, MSEE Marowan Zakhour, MD Debra M. Ikeda, MD Robert J. Herfkens, MD Published online /radiol ; 237: Abbreviations: BI-RADS Breast Imaging Reporting and Data System DCIS ductal carcinoma in situ 3D three-dimensional 1 From the Department of, Stanford University Medical Center, 300 Pasteur Dr, Room H1307, Stanford, CA Received May 7, 2004; revision requested July 20; revision received November 22; accepted December 30. Address correspondence to B.L.D. ( Authors stated no financial relationship to disclose. The Lactating Breast: Contrast-enhanced MR Imaging of Normal Tissue and Cancer 1 PURPOSE: To retrospectively describe the magnetic resonance (MR) imaging characteristics of normal breast tissue and breast cancer in the setting of lactation. MATERIALS AND METHODS: The HIPAA-compliant study was exempt from institutional approval, and informed consent was not required. Unilateral MR imaging of 10 breasts was performed in seven lactating patients aged years. For the three patients in whom both breasts were imaged, each breast was imaged on a separate day. Nonenhanced T1-weighted and fat-saturated T2-weighted images and contrast material enhanced dynamic three-dimensional (3D) T1-weighted spiral gradient-echo images interleaved with T1-weighted high-spatial-resolution 3D gradient-echo images ( mm voxels) were obtained. Three readers in consensus assessed the glandular density, T2-weighted signal intensity, milk duct appearance, and contrast enhancement in normal and tumor-containing breast regions. The pharmacokinetic contrast enhancement parameters of tumors were compared with those of normal tissue by using Student t and Mann-Whitney tests. RESULTS: MR findings of normal breast tissue in the seven women included increased glandular density in six women, high T2-weighted signal intensity in six, dilated central ducts in seven, and rapid initial glandular contrast enhancement in seven. MR findings of invasive ductal carcinoma in five women, compared with findings of the normal glandular tissue, included lower T2-weighted signal intensity in five women, more avid and rapid contrast enhancement in five, and early contrast enhancement washout in four. One minute after contrast agent injection, tumor signal intensity increased significantly more than normal lactating tissue signal intensity (153% vs 60% from baseline, P.016). The median two-compartment model K 21 exchange rate in the tumors, sec 1, was significantly faster than the K 21 exchange rate in normal tissue, sec 1 (P.03). CONCLUSION: Normal lactating glands have increased density, high T2-weighted signal intensity, and rapid moderate contrast enhancement. Breast cancers are visible during lactation owing to their lower signal intensity and more intense initial contrast enhancement with early washout compared with normal breast tissue. RSNA, 2005 Author contributions: Guarantors of integrity of entire study, L.A.E., B.L.D., D.M.I.; study concepts and design, all authors; literature research, L.A.E., B.L.D., D.M.I., R.J.H.; clinical studies, L.A.E., B.L.D., D.M.I.; data acquisition, L.A.E., B.L.D., L.V., M.Z., R.J.H.; data analysis/interpretation, L.A.E., B.L.D., L.V., D.M.I., R.J.H.; statistical analysis, L.A.E., B.L.D., L.V., M.Z.; manuscript preparation, definition of intellectual content, revision/ review, and final version approval, all authors; manuscript editing, L.A.E., B.L.D., D.M.I., R.J.H. RSNA, 2005 Breast cancer is the most frequently diagnosed cancer during pregnancy (1,2). Pregnancyassociated breast cancer, defined as breast cancer diagnosed during pregnancy or up to 1 year after delivery, is associated with a poor prognosis and is often advanced at the time of diagnosis. Difficulties in evaluating the breast during pregnancy and lactation may delay the diagnosis (3); the sensitivity of conventional mammography is diminished owing to the increased parenchymal density, glandularity, and water content of the lactating breast (2,4). We hypothesized that the high sensitivity of contrast material enhanced magnetic resonance (MR) imaging in the diagnosis of breast cancer, even in dense breasts, might enable reliable detection of breast cancer in lactating women. However, a previous case report of contrast-enhanced MR imaging in lactating women (5) suggested that intravenous gadolinium-based contrast agents are rapidly taken up by lactating breast tissue and that this could limit the accuracy of MR imaging in the diagnosis of breast cancer in lactating women. No cancer was described in that report, however. Thus, the purpose of 429

2 our study was to retrospectively describe the MR imaging characteristics of normal breast tissue and breast cancer in the setting of lactation. MATERIALS AND METHODS Patients Between October 1998 and February 2004, 2475 diagnostic unilateral gadolinium-enhanced breast MR imaging examinations were performed. Only unilateral MR imaging is performed at our institution because this protocol maximizes the temporal and spatial resolution that can be obtained during the dynamic and high-spatial-resolution portions of the examination. When patients require bilateral imaging, they are imaged on separate days (usually consecutive) to allow sufficient time for contrast agent elimination between the acquisitions. We (L.A.E., B.L.D., and D.M.I.) retrospectively searched our breast MR imaging database to identify those examinations that had been performed while the women were lactating. To be included in the study, the woman had to be currently breast-feeding or have been breast-feeding within a month before the study. Our institution s human subjects panel exempted us from requiring approval for this retrospective study, and informed patient consent was not required. Our study was Health Insurance Portability and Accountability Act compliant. We identified seven patients who had undergone unilateral (n 4) or bilateral (n 3) breast MR imaging while they were lactating (Table 1). The patients were aged years (mean, 36 years). At the time of MR imaging, the women had been breast-feeding for a period ranging from 5 days to 22 months (mean, 18 months). All patients were currently breast-feeding or had stopped breastfeeding within 11 days before their MR imaging examination. Five of the seven patients underwent MR imaging for staging of a biopsyproved palpable breast cancer. One patient underwent MR imaging for evaluation of breast angiosarcoma that was suspected at skin biopsy performed during her recent pregnancy. Subsequent surgical biopsy and 18-month clinical follow-up failed to demonstrate malignancy in this patient. Another patient underwent screening MR imaging owing to a strong family history of breast cancer. At the time of this writing, clinical and imaging follow-up examinations were pending for this patient. Imaging Examinations MR imaging was performed while the patient was prone and by using a dedicated four-coil phased-array breast coil system (MR Imaging Devices, Waukesha, Wis) and a conventional 1.5-T unit (EchoSpeed; GE Medical Systems, Milwaukee, Wis). Initially, the body coil was used to obtain bilateral transverse T1-weighted large-field-of-view spinecho MR images (300/8 [repetition time msec/echo time msec], 5-mm section thickness, 5-mm section spacing, 36-cm field of view, acquisition matrix of , 90 flip angle). For the remainder of the examination, the breast coil was used to obtain unilateral 20-cm field of view sagittal images, which included nonenhanced T2-weighted images, nonenhanced three-dimensional (3D) T1-weighted images, rapid dynamic T1-weighted images obtained during intravenous bolus administration of gadolinium-based contrast agent (gadopentetate dimeglumine, Magnevist; Berlex Laboratories, Wayne, NJ), high-spatialresolution contrast-enhanced 3D T1- weighted images, and delayed-enhancement rapid dynamic T1-weighted images, as described by Agoston et al (6). Specific MR imaging parameters were as follows: For nonenhanced T2-weighted imaging, a fast spin-echo sequence with chemical fat saturation (4000/102.8, 32 section locations without intersection gaps, 3 4-mm section spacing, acquisition matrix of , echo train length of 12) was performed. For nonenhanced 3D T1- weighted imaging, a 3D spoiled gradientecho sequence with water-selective spectral-spatial excitation and a magnetization transfer contrast pulse (28.2/7.7, 32 section locations, 3 4-mm section thickness, acquisition matrix of , 50 flip angle) (7) was performed. For rapid dynamic spiral T1-weighted imaging, a spoiled gradient-echo sequence with water-selective excitation, a magnetization transfer contrast pulse, and 3D stack of spiral spatial encodings (8) (38/12, 40 flip angle, 20 spiral interleaves, 10 positive k z and four negative k z phase encodes with zero filling to yield 32 section locations with mm section spacing, imaging time of 10.7 seconds, with imaging repeated 20 times for a total wash-in dynamic series duration of 214 seconds) was performed. Forty seconds after the start of the wash-in dynamic series, a power injector was used to intravenously inject 0.1 mmol of gadopentetate dimeglumine per kilogram of body weight as a bolus at 2.0 ml/sec through an antecubital vein. A 20-mL saline flush followed. For high-spatial-resolution 3D T1-weighted imaging, a 3D spoiled gradient-echo sequence with water-selective spectral-spatial excitation and a magnetization transfer contrast pulse (30.012/8.596, 64 section locations, mm section spacing, acquisition matrix of , 50 flip angle, elliptic-centric phase encoding) was performed. Delayed-enhancement rapid dynamic T1-weighted imaging was performed by using the parameters used to perform rapid dynamic 3D spiral T1- weighted imaging during bolus gadopentetate dimeglumine administration, with the exception that the acquisition was repeated 26 times to yield a washout imaging duration of 277 seconds. After the MR imaging examinations, the patients were instructed to pump and discard their breast milk for 24 hours before reinitiating breast-feeding. This is the standard recommendation at our institution, although relatively recent evidence suggests that the amount of intravenous gadolinium-based contrast agent excreted into breast milk is minimal (9). When bilateral MR imaging was performed, each breast was imaged on consecutive days to allow complete contrast agent clearance from each breast between the examinations. In one patient, however, 14 days elapsed between the two examinations (Table 1). Image Interpretation Three authors (L.A.E., B.L.D., and D.M.I.) with, respectively, 0, 6, and 6 years of experience reading breast MR images reviewed the breast imaging studies to reach a consensus regarding the findings. The breast density seen on the T1- weighted spin-echo images was subjectively judged to be increased if glandular tissue accounted for more than 90% of the breast volume. The presence or absence of dilated high-signal-intensity ducts was assessed on both the waterspecific T1-weighted 3D spoiled gradientecho (with water-selective excitation and a magnetization transfer contrast pulse) images and the fat-saturated T2-weighted fast spin-echo images. The overall appearance of glandular tissue on the fat-saturated T2-weighted images was categorized as high, moderate, or low signal intensity and as uniform or heterogeneous compared with the appearance of the adjacent pectoral muscle. The contrast enhancement of the normal glandular tissue was categorized by using terminology similar to the 430 November 2005 Espinosa et al

3 TABLE 1 Patient Data Patient No./ Age (y) Breast Side* No. of Days Breast Feeding No. of Days Breast-Feeding Stopped Indication for MR Imaging 1/38 Left 5 0 Tumor staging 2/33 Right Tumor staging 3/42 Left 7 0 Possible cutaneous angiosarcoma 4/27 Left, right (1) , 10 Tumor staging 5/38 Left, right (14) 660 0, 0 Strong family history of breast cancer 6/36 Left Tumor staging 7/36 Left, right (1) 210 7, 8 Tumor staging * Numbers in parentheses are numbers of days between left and right breast examinations. Number of days woman had been breast feeding at time of MR imaging examination. Number of days since woman stopped breast feeding to time of MR imaging examination. When two numbers are given, the first and second numbers correspond, respectively, to left and right breast feeding. Breast Imaging Reporting and Data System (BI-RADS) MR imaging lexicon for describing abnormal enhancing lesions. Potential patterns of enhancement were homogeneous, heterogeneous, or stippled. The extent of normal glandular enhancement was categorized as diffuse, geographic, or focal. The degree of enhancement was categorized as mild, strong, or moderate. Ductal or periductal enhancement was judged to be absent, present, or marked (ie, affecting the majority of the major central ducts). A single ovoid region of interest containing entirely normal glandular tissue was generated on the T1-weighted 3D spiral MR images, and for all lesions, dynamic time signal intensity data were obtained and subjectively and quantitatively analyzed. For each patient, data from a single region of interest containing normal glandular tissue were analyzed. In patients in whom bilateral MR images were obtained, this normal tissue region of interest was in the contralateral breast (ie, breast without cancer). In patients with cancer, the size of the normal tissue region of interest was similar to the size of the region circumscribed around the tumor (described in following text). All abnormal lesions were identified on contrast-enhanced water-specific T1- weighted 3D spoiled gradient-echo MR images. The abnormal lesions were examined for morphologic features (by using 2003 BI-RADS MR imaging lexicon) on nonenhanced T1- and T2-weighted MR images and on nonenhanced and contrastenhanced water-specific T1-weighted 3D spoiled gradient-echo images. The lesions were examined for dynamic features on T1-weighted spiral images. The three readers examined the lesions in consensus. T2-weighted imaging lesion signal intensity was rated as low, equal, or high compared with the signal intensity of the surrounding glandular tissue. In addition to characterizing the index lesion in each patient according to 2003 BI-RADS MR imaging lexicon, we determined the presence or absence of additional lesions on MR images and rated the overall signal intensity on the contrast-enhanced T1-weighted high-spatialresolution images as high, equal, or low compared with the overall signal intensity of the surrounding glandular tissue. The abnormal lesions were also identified on the T1-weighted spiral images. Manually fitted regions of interest were circumscribed along the perimeter of the enhancing lesion, and curves of average signal intensity as a function of time were constructed by using computer software (Functool, Advantage Windows; GE Medical Systems), as described by Daniel et al (10). The curves were subjectively classified by the readers in consensus as type 1, 2, 3, 4, or 5 curves according to the classification system of Daniel et al (10): A type 1 curve indicated essentially no lesion enhancement; a type 2 curve, slow sustained enhancement; a type 3 curve, rapid initial enhancement and sustained late enhancement; a type 4 curve, rapid initial enhancement and stable late enhancement; and a type 5 curve, rapid initial enhancement and decreasing late enhancement. These curves were quantitatively analyzed for empiric and pharmacokinetic parameters (described in Statistical Analyses section). The corresponding conventional mammographic and ultrasonographic (US) results were reviewed by one author (D.M.I.). Written reports on the conventional mammographic and US examinations performed in five patients at outside institutions and on the original conventional mammographic and US examinations performed in one patient at our institution were reviewed. The mammograms were reviewed for overall breast density according to the standard BI-RADS classification, the presence or absence of any masses and/or suspicious microcalcifications, and the distribution (eg, unifocal, multifocal, multicentric, or extensive) of all suspicious findings. The US images were reviewed for the presence or absence of hypoechoic masses and the distribution of all suspicious masses. Statistical Analyses Quantitative contrast enhancement parameters were calculated from the dynamic T1-weighted imaging region of interest data as follows: The baseline signal intensity was determined by averaging the signal intensity in the region on images obtained at four consecutive time points before the start of the breast contrast enhancement. The 1-minute enhancement percentage was calculated as the ratio of the region signal intensity divided by the baseline signal intensity for each region. The dynamic enhancement was fitted to a two-compartment pharmacokinetic model by using a numeric least-squares based Broyden-Fletcher-Goldfarb-Shanno quasi-newton method and computer software (Matlab; Mathworks, Natick, Mass) (5). Two parameters, amplitude and K 21 or K ep exchange rate, which previously have been shown to be predictive of malignancy (10), were evaluated. In this model the K 21 or K ep exchange rate is defined as the rate at which contrast agent flows between the plasma and extracellular space of the tumor in response to the difference in contrast agent concentration between the plasma and the extracellular space of the tumor. The amplitude reflects the relative volume of the Volume 237 Number 2 MR Imaging of Lactating Breast Tissue 431

4 Figure 1. Prone MR images obtained in 33-year-old lactating woman with invasive ductal carcinoma. (a) Transverse T1-weighted spin-echo image (300/8) shows diffuse increase in glandular volume that is, increased breast density. (b) Sagittal fat-suppressed T2-weighted fast spin-echo image (4000/103; echo train length, 12) shows dilated high-signal-intensity central milk ducts (short arrows) and diffuse high signal intensity of all glandular tissue except the invasive ductal carcinoma (long arrow), which has relatively low signal intensity. (c) Sagittal contrast-enhanced high-spatial-resolution water-selective T1-weighted spoiled gradient-echo image (30/8.6, 40 flip angle) shows linear enhancement of the central milk ducts (short arrows), moderate heterogeneous enhancement of the normal lactating glands, and strong enhancement of the tumor (long arrow), which has suspicious rim enhancement features. TABLE 2 Characteristics of Normal Lactating Breast Tissue on Nonenhanced MR Images Patient No. Breast Density T1-weighted Imaging T2-weighted Imaging Dilated High-Signal-Intensity Ducts Gland Uniformity Gland Signal Intensity Dilated High-Signal-Intensity Ducts 1 Increased No Uniform High Yes 2 Increased No Uniform High Yes 3 Increased No Heterogeneous Moderate Yes 4 Increased Yes* Uniform High Yes 5 Scattered fibroglandular No Uniform High Yes 6 Increased No Uniform High Yes 7 Increased No Uniform High Yes * Single duct near site of recent biopsy. plasma and tumor extracellular space compartments. The parameter values for the seven regions of normal glandular tissue (one per patient) were compared with the parameter values for the five regions containing tumor by using parametric (two-tailed Student t test) and nonparametric (Mann- Whitney test) unpaired statistical analyses and computer software (for parametric analysis: Excel for Mac, version X, 2003, Microsoft, Redmond, Wash; for nonparametric analysis: SPSS, version , for Windows, 2001, SPSS, Chicago, Ill). P.05 indicated a significant difference. RESULTS Normal Lactating Breast Tissue Conventional mammography revealed diffuse increased breast density in the six patients in whom conventional mammography was performed. MR imaging revealed increased breast density due to confluent glandular hypertrophy in the posterior and peripheral regions of the breast in the same six patients (Fig 1a, Table 2). In contrast, patient 5, who had been lactating for 22 months, had less dense breast tissue with patches of glands intermixed with a higher proportion of fat tissue at MR imaging; no mammographic results were available for this patient. All seven patients had multiple dilated ducts with high T2-weighted imaging signal intensity, which was consistent with milk (Fig 1b, Table 2). With the exception of a single duct in one patient, none of the ducts had intrinsic high signal intensity on the nonenhanced T1-weighted images. The close proximity of the one high-signal-intensity duct to a recent biopsy site suggested that the high signal intensity might have been due to the presence of methemoglobin rather than milk. Five of the seven patients had linear enhancement of one or more retroareolar central ducts on the contrast-enhanced T1-weighted high-spatial-resolution 3D water-specific spoiled gradient-echo images (Table 3). High central duct enhancement was seen in only one patient (Fig 1c). Six of the seven patients had uniformly high glandular tissue signal intensity on the T2-weighted MR images (Fig 1a, Table 2). All patients glands were diffusely enhanced at contrast-enhanced high-spatialresolution water-specific T1-weighted 3D spoiled gradient-echo imaging. Strong enhancement was noted in only one of the seven patients. The seven patients demon- 432 November 2005 Espinosa et al

5 TABLE 3 Characteristics of Normal Lactating Breast Tissue on Contrast-enhanced MR Images Patient No. Enhancement Pattern Enhancing Central Ducts Dynamic Curve Type 1-Minute Signal Intensity Increase (%)* Amplitude Exchange Rate (K 21 sec 1 ) 1 Heterogeneous, diffuse, moderate Not marked Heterogeneous, diffuse, moderate Marked Stippled, diffuse, moderate Not marked Heterogeneous, diffuse, moderate Not marked Homogeneous, diffuse, strong Absent Heterogeneous, diffuse, mild Not marked Heterogeneous, diffuse, mild Absent * Increase above baseline before contrast material injection (eg, 74% increase is equivalent to 1.74 times baseline). Figure 2. Time signal intensity curves of dynamic contrast enhancement in (a) normal lactating glands and (b) invasive ductal carcinomas in lactating women. (a) Unlike the normal glands in nonlactating women, the normal glands in these lactating women were rapidly enhanced initially. With the exception of patient 1, in whom early contrast enhancement washout was noted, all of the patients had slow but progressive delayed enhancement. (b) All of the invasive ductal carcinomas had rapid enhancement followed by early enhancement washout. The overall degree of enhancement of the invasive carcinomas was significantly higher than that of the normal lactating glandular tissue. strated rapid initial contrast enhancement, a type 3 or type 5 average signal intensity curve (Fig 2a), and a 1-minute change in signal intensity of 38% (1.38 times baseline signal intensity value) to 89% (1.89 times baseline) (average change, 60% [1.60 times baseline]). Two-compartment pharmacokinetic modeling yielded K 21 exchange rates of sec 1 (median, sec 1 ) in six of the seven patients and amplitudes of (mean, 1.25) (Table 3). Breast Cancer in Lactating Patients Histopathologic analysis revealed breast cancer in five patients (Table 4). All five breast cancers were invasive ductal carcinomas and moderately or poorly differentiated. In one of the two patients who underwent mastectomy at our institution, multifocal microscopic (2-mm or smaller) invasive carcinomas were present throughout the quadrant that contained the primary tumor. A single invasive tumor was noted in the other patient. Histopathologic measurements of the extent of tumor were not available for three patients: two patients who underwent neoadjuvant chemotherapy between MR imaging and definitive surgery and one patient in whom definitive surgery was performed at another institution. The invasive cancer in four of the five patients was associated with ductal carcinoma in situ (DCIS) (Table 4), which was predominantly high grade. In one of these four patients, a 4.8-cm DCIS was in the same quadrant that contained the primary invasive tumor and multifocal microscopic foci of invasion. In another patient, a DCIS larger than 4 cm was present throughout the breast. In two patients, histopathologic measurements of the extent of DCIS were not available. Conventional mammography revealed a focal mass corresponding to the invasive tumor in only one of the five patients with breast cancer (Table 4). However, suspicious microcalcifications were present in all five patients. In three of the five patients, multifocal or extensive carcinoma was suggested by the presence of suspicious microcalcifications, which corresponded to extensive DCIS in the two patients who underwent mastectomy. US revealed hypoechoic masses including the one mass seen at mammography or ill-defined hypoechoic regions corresponding to the invasive tumors in all five patients with breast cancer (Table 4). Multifocal cancer was suspected in one patient at US owing to multiple solid hypoechoic masses that were not seen at mammography. This patient underwent neoadjuvant chemotherapy before surgery; thus, histopathologic confirmation of the suspected initial multifocal disease was not possible. At nonenhanced T2-weighted MR imaging, the tumors in all five patients had low signal intensity compared with the Volume 237 Number 2 MR Imaging of Lactating Breast Tissue 433

6 TABLE 4 Imaging and Pathologic Characteristics of Breast Cancers in Five Lactating Patients Characteristic Patient 1* Patient 2 Patient 4 Patient 6 Patient 7 Histopathologic Type IDC Grade 2 IDC Grade 3 IDC Grade 2 IDC Grade 3 IDC Size (cm) NA NA NA Multifocal invasive NA NA No Yes NA tumor Associated DCIS Yes No Yes, comedo Yes, intermediate to Yes, comedo high grade DCIS size NA 4.0 cm 4.8 NA Mammography Mass size (cm) No mass No mass No mass 1.5 No mass Size of suspicious Focal cluster 7.0 Entire quadrant 4.0 Entire quadrant calcifications (cm) Multifocal or No No Yes Yes Yes extensive tumor US Hypoechoic lesion Mass, unknown size , 3.1, 1.2 size (cm) Multifocal or No No No No Yes extensive tumor MR imaging Primary focus Primary focus size (cm) Additional abnormality Heterogeneously enhancing mass Enhancing mass with heterogeneous rim Enhancing mass with heterogeneous rim Homogeneously enhancing mass Heterogeneously enhancement over half of breast NA Numerous heterogeneously enhancing masses over half of breast Multiple homogeneously enhancing masses # Patchy enhancement of entire breast Quadrant of segmental enhancement** No additional abnormality Note. Patients 3 and 5 were healthy lactating patients without breast cancer. IDC invasive ductal carcinoma, NA not available. * Histopathologic analysis of findings in patient 1 was performed at an outside institution, so complete pathologic data were not available. Size and presence or absence of multifocal invasive tumor unknown. Only incisional biopsy or core-needle biopsy was performed before neoadjuvant chemotherapy. Size of region with calcifications was not reported. Patient 7 had three hypoechoic lesions at US. All five patients had multifocal extensive tumors at MR imaging. At precontrast T2-weighted imaging, all five patients had low-signal-intensity cancerous lesions. At postcontrast MR imaging, all patients except patient 4 had high-signal-intensity cancerous lesions. Patient 4 had a cancerous lesion that was isointense relative to the surrounding normal glandular tissue. # Patient 2 had enhancing 1.6-, 1.0-, and 0.7-cm masses. ** Six-centimeter quadrant of enhancement. normal lactating glandular tissue (Fig 1b, Table 4). The tumor signal intensity was only minimally higher than the signal intensity of the underlying muscles. No specific architectural features were noted on the T2-weighted images. At contrast-enhanced T1-weighted MR imaging, all invasive tumors in all five patients were visible as focal enhancing regions. Four invasive tumors manifested as focal areas with high signal intensity at contrast-enhanced high-spatial-resolution 3D water-specific spoiled gradientecho imaging. In the fifth tumor, focal rim enhancement and mass effect in the chest wall distinguished the location of the tumor, even though the overall signal intensity was similar to that of the avidly enhancing heterogeneous background tissue that contained DCIS. In this breast, the rapidly enhancing tumor was much brighter than the surrounding tissue on the very early (eg, seconds after contrast material injection) dynamic images. The invasive tumors had variable morphologic features on the 3D high-spatial-resolution contrast-enhanced water-specific spoiled gradientecho images. Rim enhancement was present in two of the five tumors. Spiculation was not observed. The enhancement pattern was heterogeneous in four tumors and homogeneous in one. At dynamic T1-weighted imaging, all tumors were rapidly enhancing and had early contrast enhancement washout (type 5 curves) (Fig 2b). The 1-minute change in signal intensity ranged from 88% (1.88 times baseline signal intensity value) to 239% (3.39 times baseline) (average change, 153% [2.53 times baseline]) and was significantly higher than that in the normal lactating tissue (P.016, unpaired two-tailed Student t test) (Fig 3a). Two-compartment pharmacokinetic modeling yielded K 21 exchange rates of sec 1 (median, sec 1 ) in four of the five patients; these values were significantly higher than those in the regions of normal lactating tissue in the seven patients (P.03, Mann-Whitney test) (Fig 3b). Amplitudes ranged from 0.92 to 2.41 (mean, 1.64) and were nearly identical to amplitudes in the normal lactating tissue (Fig 3c). In the two patients who underwent mastectomy, the primary invasive tumors corresponded to the sizes and shapes of the dominant focal masses seen at MR imaging. However, the multiple microscopic foci of invasion that were seen in one of the two patients were indistinguishable from the surrounding DCIS that was visible in the entire quad- 434 November 2005 Espinosa et al

7 Figure 3. Comparisons of dynamic enhancement patterns in normal lactating breast tissue ( ) and invasive breast cancer (E). (a) The invasive tumors enhanced significantly more strongly within the 1st minute after contrast agent injection (P.016). (b, c) Two-compartment pharmacokinetic modeling yielded higher K 21 exchange rates (b) (P.03) but similar amplitudes (c) in the cancers. rant surrounding her primary tumor (discussed in following text). MR Imaging Features of DCIS Four of five patients with breast cancer had DCIS in addition to their invasive tumors. In one of these four patients, mastectomy revealed extensive comedo DCIS throughout much of the breast. In another patient, mastectomy revealed comedo DCIS with multiple microscopic foci of invasion throughout the lower inner quadrant surrounding the index mass but sparing the other quadrants of the breast (Fig 4). In both patients, T2- weighted MR imaging revealed lower signal intensity throughout the DCIS compared with high signal intensity in the uninvolved lactating tissue in the same or contralateral breast. In both patients, the DCIS also demonstrated substantial heterogeneous contrast enhancement that was nearly equal to the degree of enhancement of their index tumors on the 2.5-minute postcontrast high-spatialresolution 3D water-specific spoiled gradient-echo images. Multifocal or extensive invasive carcinoma or DCIS was also suspected at MR imaging in the other three patients, including one patient with a large, confluent, heterogeneous, and abnormally enhancing region and two with multiple enhancing masses. As noted earlier, histopathologic measurements of the extent of disease were not available for these patients. However, in one patient, the malignant nature of the multiple enhancing masses was confirmed at follow-up MR imaging, which depicted a progression in the size and number of suspicious rim-enhancing lesions despite interval neoadjuvant chemotherapy having been performed. DISCUSSION Figure 4. Comedo DCIS with multiple microscopic foci of invasion surrounding a 1.9-cm invasive ductal carcinoma in a 36-year-old woman. (a) Prone sagittal T2-weighted fast spin-echo image (4000/102.8, echo-train length of 12) shows low signal intensity throughout the lower inner quadrant of glandular tissue (short arrows), as well as within the main tumor mass (long arrow), compared with the high signal intensity of the uninvolved lactating glands in the upper region of the breast. (b) Prone sagittal high-spatial-resolution contrast-enhanced 3D waterselective spoiled gradient-echo image (30/8.6, 50 flip angle) shows segmental enhancement (short arrows) throughout the lower inner quadrant and the enhancing DCIS tumor (long arrow). (c) Corresponding time signal intensity curves of contrast enhancement at rapid dynamic T1-weighted MR imaging indicate that the invasive tumor was initially enhancing extremely fast and intensely, and a rapid washout followed (type 5 curve). The DCIS was enhancing more strongly than the uninvolved lactating tissue but at a similar time course (type 3 curve). Our results confirm the MR imaging appearances of normal lactating breast tissue, as previously described in two case reports (5,11). There is increased glandular volume, consistent with increased breast density, as has previously been noted at conventional mammography (2,4) and MR imaging (11). Unlike nonlactating breast tissue, lactating breast tissue has a uniformly long T2, which results in diffuse high signal intensity on T2-weighted MR images. This finding reflects the increased mobile water fraction within milk as it is being produced in the glands. The diffuse high signal intensity is consistent with the slightly lengthened T1 of lactating breast tissue reported by Talele et al (5). Unlike normal nonlactating breast tissue, which is mildly and progressively enhanced Volume 237 Number 2 MR Imaging of Lactating Breast Tissue 435

8 by intravenous contrast material, normal lactating glands show rapid enhancement followed by an early plateau of enhancement. This time course of the signal intensity of normal lactating glands overlaps with the qualitative enhancement characteristics of invasive malignancy in nonlactating women previously described by Talele et al (5) and has been attributed to increased vascular permeability. The degree of initial enhancement at 1 minute and the high median K 21 exchange rate of sec 1 noted in the normal lactating glands in this series were also within the range of quantitative dynamic enhancement criteria previously reported for invasive malignancy in nonlactating patients (10). A previously unreported (to our knowledge) feature of lactation is enhancement of the central retroareolar ducts, some degree of which was noted in five of the seven patients. Image resolution was inadequate to determine whether this central duct enhancement was due to enhancement of the duct walls and periductal tissue or to excretion of contrast agent into the breast milk (9) in the ductal lumen. The combination of high signal intensity at T2-weighted imaging and rapid contrast enhancement in normal lactating tissue resembles the MR imaging signal intensity characteristics of some fibroadenomas (12). This appearance has been attributed to an expanded extracellular space that contains myxoid matrix with a high water content (12). However, the enhancement characteristics of normal lactating breast tissue differ from those of fibroadenomas in one critical way: The degree of enhancement of normal lactating breast tissue is much lower than that reported for fibroadenomas. The average degree of enhancement was only 60% (1.6 times baseline value) in our series. This finding is not consistent with an expanded extracellular space in lactating breast tissue. Given the abnormal rapid dynamic contrast enhancement characteristics of normal lactating breast tissue observed in this series and described in a previous case report (5), the hypothesis that normal lactating tissue might obscure invasive carcinomas on contrast-enhanced MR images (5) seems plausible, although to our knowledge there are no published reports directly describing the MR imaging characteristics of breast cancer in lactating patients. Contrary to this hypothesis, all of the invasive carcinomas in this series were readily apparent on both the nonenhanced T2-weighted images and the contrastenhanced T1-weighted images, despite the surrounding lactating tissue. All tumors in our series were also consistently conspicuous as low-signal-intensity foci on the nonenhanced T2-weighted MR images. The tumors were more reliably visible on T2-weighted images than were the tumors in a group of nonlactating women (13), in whom T2 was not a reliable discriminator between tumors and normal tissue. This was not because the tumors in the lactating women had uniquely short T2 relaxation times; rather, it was because the typical intermediate T2-weighted signal intensity of these tumors was conspicuous against the high signal intensity of the noncancerous lactating glandular tissue that surrounded them. All tumors in this series were also visible as high-signal-intensity foci on the T1-weighted contrast-enhanced images, despite the rapid enhancement of the surrounding lactating tissue. This finding was due to both the time course of enhancement in the tumors, which was even faster than the rapid time course of enhancement in the lactating tissue, and the overall greater degree of enhancement. Moreover, suspicious morphologic features, such as rim enhancement (14,15), were visible in some tumors, despite the surrounding enhancing lactating tissue. It is noteworthy that the tumors in all five women with cancer in this series were also visible to some degree at conventional mammography, US, or both. However, the observation that conventional mammography depicted primarily pleomorphic calcifications and demonstrated a mass in only one of the five patients with cancer suggests that MR imaging may be more reliable than conventional mammography in depicting the extent of invasive carcinomas in lactating patients. US, on the other hand, was helpful in identifying the invasive breast cancer masses that were not detected at mammography. It is possible that conventional mammography performed to identify in situ tumors in combination with US performed to identify invasive tumors may depict the full extent of cancer as well as MR imaging. Future studies are needed to determine the exact role of MR imaging, as compared with the role of conventional mammography combined with US, in the assessment of lactating patients known or suspected of having breast cancer. There were several limitations to this study. The described retrospectively identified patients all had extensive, palpable, moderately or poorly differentiated tumors at presentation. It is unclear how well MR imaging would depict more subtle forms of breast cancer such as infiltrating lobular carcinoma that enhance less avidly and do not manifest as confluent, space-occupying masses or how well it would depict mucinous cancer, which has intrinsically high signal intensity on T2- weighted images. Similarly, although confluent high-grade DCIS appeared distinct from the lactating tissue in two patients in this study, it is unclear how reliably other forms of DCIS in the lactating breast would be visualized at MR imaging. Moreover, in this retrospective study, we did not evaluate how long lactational changes can be seen on MR images after weaning, how these changes evolve over the duration of breast-feeding, or whether the duration of breast-feeding affects the diagnostic accuracy of MR imaging. In conclusion, we have described the MR imaging features of the lactating breast in seven patients. The observed MR imaging features associated with normal lactating glands were increased density and size of the breast, high signal intensity on T2-weighted images, and rapid, moderate uptake of contrast agent. MR imaging also depicted the breast cancer in five of the lactating patients. The observed MR imaging features of cancer in lactating breast tissue included signal intensity lower than that of the surrounding lactating glandular tissue on T2-weighted images, strong rapid initial contrast enhancement with early washout, irregular margins, and occasional rim enhancement. References 1. DiFronzo LA, O Connell TX. Breast cancer in pregnancy and lactation. Surg Clin North Am 1996;76: Liberman L, Giess CS, Dershaw DD, Deutch BM, Petrek JA. Imaging of pregnancyassociated breast cancer. 1994;191(1): Woo JC, Yu T, Hurd TC. Breast cancer in pregnancy: a literature review. Arch Surg 2003; 138: Ahn BY, Kim HH, Moon WK, et al. Pregnancyand lactation-associated breast cancer: mammographic and sonographic findings. J Ultrasound Med 2003;22: Talele AC, Slanetz PJ, Edmister WB, Yeh ED, Kopans DB. The lactating breast: MRI findings and literature review. Breast J 2003;9(3): Agoston AT, Daniel BL, Herfkens RJ, et al. Intensity-modulated parametric mapping for simultaneous display of rapid dynamic and high-spatial-resolution breast MR imaging data. RadioGraphics 2001;21: Leong CS, Daniel BL, Herfkens RJ, et al. Characterization of breast lesion morphology with delayed 3DSSMT: an adjunct to dynamic breast MRI. J Magn Reson Imaging 2000;11: Yen YF, Han KF, Daniel BL, et al. Dynamic breast MRI with spiral trajectories: 3D versus 2D. J Magn Reson Imaging 2000;11: Kubik-Huch RA, Gottstein-Aalame NM, Frenzel T, et al. Gadopentetate dimeglumine excretion into human breast milk during lactation. 2000;216: Daniel BL, Yen YF, Glover GH, et al. Breast disease: dynamic spiral MR imaging. 1998;209: Kaiser WA. The lactating breast in the magnetic resonance tomogram: spin-echo and fat-water images. Rofo 1987;146: Kuhl CK, Klaschik S, Mielcarek P, Gieseke J, Wardelmann E, Schild HH. Do T2-weighted pulse sequences help with the differential diagnosis of enhancing lesions in dynamic breast MRI? J Magn Reson Imaging 1999;9: Heywang SH, Bassermann R, Fenzl G, et al. MRI of the breast: histopathologic correlation. Eur J Radiol 1987;7: Wedegartner U, Bick U, Wortler K, Rummeny E, Bongartz G. Differentiation between benign and malignant findings on MR-mammography: usefulness of morphological criteria. Eur Radiol 2001;11: Nunes LW, Schnall MD, Siegelman E, et al. Diagnostic performance characteristics of architectural features revealed by high spatial-resolution MR imaging of the breast. AJR Am J Roentgenol 1997;169: November 2005 Espinosa et al