Radiology
John M. Lewin, MD Pamela K. Isaacs, DO Virginia Vance, RN Fred J. Larke, MS
Index terms: Breast radiography, comparative studies, 00.119 Breast radiography, technology, 00.119 Published online before print 10.1148/radiol.2291021276 Radiology 2003; 229:261–268 Abbreviations: DCIS ⫽ ductal carcinoma in situ DSM ⫽ digital subtraction mammography 1
From the Department of Radiology, University of Colorado Health Sciences Center, Denver; and Department of Radiology, University of Colorado Hospital Breast Center, Rm 3140A/Mailstop F724, 1635 N Ursula St, Aurora, CO 80010. From the 2001 RSNA scientific assembly. Received October 3, 2002; revision requested November 27; final revision received March 17, 2003; accepted April 7. Supported by a grant from the Susan G. Komen Breast Cancer Foundation. Address correspondence to J.M.L. (e-mail:
[email protected]).
Dual-Energy Contrastenhanced Digital Subtraction Mammography: Feasibility1 A technique for demonstrating breast cancers, dual-energy contrast agent– enhanced digital subtraction mammography, was performed in 26 subjects with mammographic or clinical findings that warranted biopsy. The technique consists of highenergy and low-energy digital mammography after administration of iodinated contrast agent. Weighted subtraction of the logarithmic transform of these images is then performed to obtain an image that preferentially shows iodine. Of the 26 subjects, 13 had invasive cancers. Eleven of these tumors enhanced strongly, one enhanced moderately, and one enhanced weakly. The duct in one patient with ductal carcinoma in situ was weakly enhancing. In the other 12 patients, benign tissue enhanced diffusely in two and weakly focally in two. These results indicate that the technique is feasible and worthy of further study. ©
Author contributions: Guarantor of integrity of entire study, J.M.L.; study concepts and design, J.M.L.; literature research, J.M.L.; clinical studies, J.M.L., V.V., P.K.I.; experimental studies, J.M.L., F.J.L.; data acquisition, J.M.L., V.V.; data analysis/interpretation, J.M.L.; statistical analysis, J.M.L.; manuscript preparation, definition of intellectual content, and editing, J.M.L.; manuscript revision/review and final version approval, all authors ©
RSNA, 2003
RSNA, 2003
At mammography, 10%–20% of breast cancers (1,2), including at least 9% of those already palpable (3), are not depicted. It has been hoped that full-field digital mammography, because of superior contrast resolution, will prove superior to standard screen-film mammography in the detection of breast cancers. This hope, unfortunately, was not borne out in one clinical study (4). Techniques are sought that make use of the technical properties of full-field digital mammography to find more cancers. One proposed idea is to use an intravenous iodinated contrast agent to attempt to enhance cancers that would otherwise be
mammographically occult. Digital subtraction angiography of the breast with fluoroscopic equipment was performed for evaluation of breast tumors almost 20 years ago (5) but did not prove to be clinically useful. Breast cancers have been shown to enhance with iodinated contrast agents at computed tomography (CT) (6,7) and with gadolinium-based contrast agents at magnetic resonance (MR) imaging (8 –10). There are difficulties inherent in this approach. The contrast resolution of fullfield digital mammography, although superior to that of screen-film mammography, is far less than that of CT or MR imaging. Additionally, the breast compression used for mammography creates an external pressure that is greater than the venous pressure. Thus, venous outflow is restricted, which potentially decreases delivery of the contrast agent to the tissues. Suboptimal tissue enhancement while the breast is under compression precludes the use of compression if the technique requires subtraction of precontrast from postcontrast images. This is done in breast MR imaging because the position of the breast cannot be reproduced adequately if compression is released and then reapplied. Since June 2000, we have studied techniques for performing contrast agent– enhanced digital subtraction mammography (DSM). Dual-energy subtraction is used in one of these techniques. Thus, the purpose of our study was to evaluate a technique we developed, dual-energy contrast-enhanced DSM, for use in demonstrating breast cancers.
Materials and Methods Patients From November 2000 to January 2002, patients with mammographically occult or mammographically visible lesions that 261
Radiology
TABLE 1 Malignant Lesions at Dual-Energy Contrast-enhanced Digital Subtraction Mammography Patient No. 1 2 3 5 6 10 11 12 14 15 21 22 23 24
Pathologic Finding
Size (mm)*
Invasive lobular carcinoma Invasive lobular carcinoma, DCIS Invasive ductal carcinoma Invasive ductal carcinoma, DCIS Invasive ductal carcinoma Invasive ductal carcinoma, DCIS Tubular carcinoma DCIS Invasive ductal carcinoma, DCIS Invasive ductal carcinoma, DCIS Invasive ductal carcinoma Invasive ductal carcinoma Invasive ductal carcinoma, DCIS Invasive ductal carcinoma
94 12 20 30 11 43# 5 31# 12 18 20 20 28 23
Palpable
Detected at Mammography
Yes Yes Yes Yes No Yes No No No No Yes No No No
No No No Yes Yes Yes No Yes No Yes Yes Yes Yes Yes
Mammographic Finding
Subjective Enhancement Score†
Contrastto-Noise Ratio‡
No finding No finding No finding Mass Mass Microcalcifications No finding Microcalcifications No finding Mass Mass Architectural distortion Microcalcifications, mass Mass
Moderate Strong Strong Weak 㛳 Strong Moderate Strong Possible Weak** Strong Strong Strong Moderate Strong
NA§ 3.4 4.1 1.9㛳 2.5 2.8 4.1 1.2 1.2** 2.2 2.8 2.6 2.5 2.4
* Longest dimension determined at pathologic examination, unless specified otherwise. † Rated by blinded observer. ‡ Between the lesion and adjacent tissue. § NA ⫽ not applicable. Contrast-to-noise ratio could not be measured because of the infiltrating nature of the tumor with no discernible mass. 㛳 Only about 60 mL of contrast agent was injected because an intravenous tubing connection was lost. # Maximal extent of calcification, in ductal orientation. ** Injection rate of 2 mL/sec.
required biopsy or that were believed subjectively by an experienced breast imager (J.M.L.) to have a 50% or greater probability of being malignant were offered enrollment in this study. Women with contraindications to receiving an iodinated contrast agent were excluded. Each woman gave written informed consent. The protocol and consent form were approved by the Colorado Multiple Institution Review Board. Twenty-six women (age range, 33– 84 years; mean age, 51 years) were enrolled. A 27th patient, who enrolled in the study on the basis of adenocarcinoma of unknown origin in a supraclavicular lymph node, was later excluded because the enrollment criteria were not met: No primary lesion was found. All enrolled subjects completed the study. Twenty-four subjects were enrolled on the basis of lesions that were mammographically detected (n ⫽ 16), mammographically occult but palpable (n ⫽ 5), or both mammographically visible and palpable (n ⫽ 3). Two additional subjects had lesions that were neither mammographically visible nor palpable. The lesion in patient 6 was found incidentally during ultrasonography (US) for a nearby palpable abnormality. The lesion in patient 14 was found at MR imaging performed because of cancer in an ipsilateral axillary lymph node. The diagnosis was established at core biopsy in all but one patient. In that patient, biopsy of a palpable lesion that was 262
䡠
Radiology
䡠
October 2003
Figure 1. X-ray spectra calculated for high- and low-energy beams. Each curve is scaled to represent exposure through a 4.5-cm-thick 50% glandular–50% fat breast. High-energy parameters include 44 kVp, rhodium anode, 0.025-mm-thick rhodium and 8-mm-thick aluminum filters, and 200 mAs. Low-energy parameters include 30 kVp, molybdenum anode, 0.03-mm-thick molybdenum filter, and 140 mAs. The k edge of iodine, at 33.2 keV, is marked by a dashed line. (Modeling program courtesy of General Electric Corporate Research and Development, Niskayuna, NY.)
visible as a hypoechoic focus at US was canceled when the lesion disappeared at both palpation and US. Findings at 1-year follow-up physical examination and mammography showed no sign of disease. In all patients with cancer, core biopsy was followed by surgical excision. All patients with benign biopsy results underwent clinical and mammographic
follow-up at 6 months or later with no evidence of malignancy. Image Acquisition The examination for each patient included low- and high-energy exposures during a single breast compression in the mediolateral oblique projection. The breast Lewin et al
Radiology
TABLE 2 Benign Lesions at Dual-Energy Contrast-enhanced Digital Subtraction Mammography Patient No. 4 7 8 9 13 16 17 18 19 20 25 26
Pathologic Finding Focal fibrosis Atypical lobular hyperplasia Apocrine metaplasia Fibroadenoma, ductal hyperplasia (usual type) Ductal hyperplasia (usual type), apocrine metaplasia Ductal hyperplasia (usual type), fibroadenoma Focal fibrosis Atypical ductal hyperplasia Fibrocystic change Sclerosing adenosis No biopsy㛳 Atypical lobular hyperplasia
Palpable
Detected at Mammography
Mammographic Finding
Subjective Enhancement Score†
Contrastto-Noise Ratio‡
No No No No
Yes Yes Yes Yes
Density Microcalcifications Density Microcalcifications
None None None None
0.0 ⫺0.2 ⫺0.1 0.1
11
No
Yes
Mass
None
0.2
8
No
Yes
Microcalcifications
None§
0.2
16 10 20 6 20 35
No No Yes No Yes No
Yes Yes No Yes No Yes
Microcalcifications Microcalcifications No mass Microcalcifications No mass Microcalcifications
None§ Weak Weak None None None
0.3 0.7 0.6 ⫺0.1 0.2 0.1
Size (mm)* 22 6 6 11
* Longest dimension determined at mammography if depicted or palpation if mammographically occult. † Rated by blinded observer. ‡ Between the lesion and adjacent tissue. § Mild diffuse enhancement. 㛳 Biopsy was canceled because of the disappearance of the palpable and sonographic finding. No evidence of disease at 12-month follow-up.
Figure 2. Patient 2. Invasive lobular carcinoma (12-mm diameter). Metal bead marks the palpable abnormality. (a) Mediolateral oblique mammogram is normal. (b) Precontrast dual-energy DSM image shows elimination of normal breast parenchyma. (c) Enhanced dual-energy DSM image shows the cancer as a round enhancing mass (arrow) in the superior part of breast. Volume 229
䡠
Number 1
Digital Subtraction Mammography
䡠
263
Radiology
was then released from compression, and the contrast agent was administered. After a delay of about 150 seconds, the breast was compressed again, and the low- and high-energy exposures were repeated. From these images, pre- and postcontrast dual-energy images were created. Images were acquired with a full-field digital mammography unit (Senographe 2000D; GE Medical Systems, Milwaukee, Wis). This unit allows a choice of either molybdenum or rhodium target materials and internal molybdenum (0.03-mm) or rhodium (0.025-mm) filters. To determine reasonable high- and low-energy techniques, an iodine-containing phantom (Nuclear Associates, Hicksville, NY) was covered with BR-12 breast-equivalent material of varying thicknesses and imaged at selected x-ray beam voltages (low-energy beam, 22–33 kVp; highenergy beam, 44 – 49 kVp). Only the rhodium target–rhodium filter combination was tested for the high-energy beam because this combination allows the hardest beam with the least tube heating. For the high-energy beam, an aluminum filter with varying thicknesses (range, 1– 8 mm) was placed in the beam. For the low-energy beam, the combinations of molybdenum target with molybdenum filter and rhodium target with rhodium filter were tested. The techniques for low- and high-energy acquisitions were chosen to maximize and minimize, respectively, the ratio of the attenuation of breast-equivalent material to that of iodine on the phantom images. Results of these preliminary studies showed that a high-energy technique of 44 kVp with 8 mm of aluminum would work well. Results were equivalent with several low-energy techniques, presumably because of the dominance of the k␣ and k peaks of the target material, which are in the range of 18 –22 keV, which is far from the k edge of iodine (33.2 keV). For the present study, most low-energy images were acquired at 30 kVp with a molybdenum target–molybdenum filter combination (patients 1–9) or at 33 kVp with a rhodium target–rhodium filter combination (patients 10 –26). Mean time between exposures was 30 seconds (range, 24 –50 seconds). During this time, the first image was acquired and displayed, the techniques were changed, and the additional aluminum filter was placed in the x-ray beam. Energy spectra calculated for the beams are shown in Figure 1. Half-value layers for the lowand high-energy beams were 0.39 and 3.07 mm of aluminum, respectively. For 264
䡠
Radiology
䡠
October 2003
Figure 3. Patient 6. Invasive ductal carcinoma (11-mm diameter). (a) Mediolateral oblique mammogram shows possible spiculated mass (arrow). (b) Dual-energy enhanced DSM image shows the cancer as an enhancing mass with definite spiculations (arrow).
a 50% glandular–50% fat breast with compressed thickness of 4.5 cm, the dose was calculated to be 0.7 mGy above that needed for conventional mammography.
injection rate of 2 mL/sec. The breast was not compressed during injection. All 26 subjects tolerated the contrast agent administration well without complaint.
Contrast Agent Administration
Image Processing and Evaluation
The contrast agent (iohexol [100 mL, with 350 mg of iodine per milliliter], Omnipaque 350; Amersham Health, Princeton, NJ) was administered with a mechanical power injector (Vistron CT; Medrad, Indianola, Pa) via a short 20gauge intravenous catheter placed in an antecubital or forearm vein, at a rate between 4 and 5 mL/sec. The injection rate was governed by a subjective assessment of the quality of the venous access. In patient 5, contrast agent administration had to be stopped twice because of a loss of the connection between the tubing and the intravenous catheter. Because of the time required to reestablish the connection in this case, only 60 mL of contrast agent was administered during 3 minutes. In patient 14, rather than subject the patient to placement of a second intravenous catheter, we used an intravenous catheter that was previously placed in a hand vein for MR imaging, with an
Custom image processing software was written (J.M.L.) in interactive data language (IDL; Research Systems, Boulder, Colo). To create a dual-energy image, weighted subtraction of the logarithm of the low-energy image from that of the high-energy image was performed. The weighting factor for the precontrast images was chosen as the value that eliminated the visibility of changes in the thickness of breast-equivalent material on a phantom made of overlapping layers of the material. The optimal weighting factor is dependent on the x-ray beam voltage, target, and filter of the image pair but not on the tube current and exposure time. A weighting factor of 0.20 was used with a low-energy technique of 30 kVp with molybdenum target and molybdenum filter and was 0.26 with 33 kVp with rhodium target and rhodium filter. A 4 ⫻ 4 boxcar filter was applied to each image for noise reduction. Lewin et al
Radiology
Figure 4. Patient 10. Invasive ductal carcinoma and DCIS. (a) Mediolateral oblique mammogram shows grouped microcalcifications in the breast (arrows) and in a lymph node (arrowhead). Enhancement is barely perceptible on postcontrast (b) low-energy and (c) high-energy images. (d) Subtracted dualenergy enhanced DSM image shows the invasive component as enhancing lesions (black arrows), but there is no definite enhancement around grouped calcifications in the posterior breast (white arrow). The malignant lymph node (arrowhead) also enhanced.
Contrast-to-noise ratios were calculated. A region of interest was drawn manually (J.M.L.) around the lesion for comparison with the surrounding tissue within a 9-mm radius of the margin, excluding the 3 mm immediately adjacent to the region of interest. The average of three measurements is reported. Subjective judgement of lesion enhancement— with the scale of strong, moderate, weak, possible, and none—was performed (P.K.I.) with the clinical full-field digital mammography review workstation with dual 2.0 ⫻ 2.5-megapixel monitors. The reader was not familiar with the cases, was blinded to the pathologic result, and did not have access to the raw images or the standard mammograms.
Results Tables 1 and 2 give the details for each lesion, the protocol used, the injection rate used, and the results of dual-energy enhanced DSM. Thirteen subjects had invasive carcinomas, and one subject had ductal carcinoma in situ (DCIS). Three of the cancers were palpable but mammographically occult, seven were mammographically visible but not palpable, two were both palpable and mammographically visible, and two were neither palpable nor mammographically visible. Benign lesions included focal fibrosis and a variety of usual and atypical proliferative changes. The patients are separated into benign and malignant categories in these tables. The patient number indicates the order in which each subject was enrolled. Eleven of the invasive cancers enhanced strongly and had contrast-tonoise ratios above 2.0 (Figs 2, 3). The other two invasive cancers enhanced weakly. The subjective judgement of lesion enhancement generally agreed with the contrast-to-noise ratio. In patient 12, with pure DCIS, the duct was faintly enhancing. In patient 10, with both invasive and in situ carciVolume 229
䡠
Number 1
Digital Subtraction Mammography
䡠
265
Radiology Figure 5. Patient 18. Benign tissue at remote lumpectomy site. Results of biopsy of new calcifications in the area showed atypical ductal hyperplasia. (a) Mediolateral oblique mammogram shows new grouped calcifications (arrow). (b) Precontrast dual-energy DSM subtraction image shows subtraction of breast tissue. (c) Enhanced dual-energy DSM image shows faint enhancement of the area around the lumpectomy site (arrowheads).
noma, strong enhancement of the invasive component was seen but not enhancement around the calcifications associated with the DCIS (Fig 4). In two patients with benign findings, areas of focal non-masslike enhancement were seen. These findings are potentially false-positive. In patient 18, the lesion was in an area of atypical ductal hyperplasia that surrounded the site of previous lumpectomy (Fig 5), while that in patient 19 was in an area of fibrocystic change. In two other patients with benign findings, diffuse enhancement of normal breast tissue was seen. In patient 11, with cancer, diffuse enhancement of the normal parenchyma was seen in addition to enhancement of the cancer as a 5-mm-diameter mass (Fig 6).
Discussion In this small study, dual-energy enhanced DSM demonstrated most of the 266
䡠
Radiology
䡠
October 2003
known cancers, including those that were visible at standard mammography and those that were not (as well as a few false-positive findings). The study was not intended to be a definitive test of enhanced DSM but rather a demonstration of its feasibility and potential. To improve our chances of success, we optimized the x-ray beam energy, incorporated power injection at a high rate, and eliminated breast compression during injection by means of dual-energy subtraction. Dual energy allowed full compression during imaging, which increased the morphologic definition of the lesion. Additionally, compared with temporal subtraction, dual-energy subtraction increases the options for imaging. With temporal subtraction, for example, multiple views cannot be obtained with a single injection. Because dual energy does not require the matching of pre- and postcontrast views, images can be obtained in
multiple projections, which allows localization of any enhancing lesion. Enhanced DSM is similar in concept to enhanced breast MR imaging and could potentially be applicable in situations in which MR imaging is currently used. Such situations include detection of a primary breast cancer in a woman with a positive axillary lymph node and determination of the extent of disease in cases of known cancer, as well as problem solving in cases of mammographic findings that were not depicted in additional mammograms or US scans. It remains to be seen whether the sensitivity to cancer is as high for enhanced DSM as it is for MR imaging, which has been shown to have a very high sensitivity (8 –10). Both techniques make use of the same property of tumor angiogenesis, which causes cancers to take up contrast agent faster and to a greater degree than do normal tissue or benign masses because of denser capillaries that are also Lewin et al
Radiology Figure 6. Patient 11. Mammographically and clinically occult tubular carcinoma (5-mm diameter). (a) Mediolateral oblique mammogram is normal. Metal bead marks a palpable abnormality with the pathologic finding of fibroadenoma. (b) Precontrast dual-energy DSM subtraction image shows subtraction of breast tissue. (c) Enhanced dual-energy DSM image shows enhancement of the cancer (arrow) and diffuse enhancement of normal tissue.
abnormally “leaky” (11). Because of its higher contrast resolution, MR imaging is probably more sensitive to contrast enhancement than is enhanced DSM, but the degree to which that translates into higher sensitivity for cancer detection is unknown. One drawback of MR imaging is that its high sensitivity to contrast agent uptake causes it to be plagued by numerous false-positive foci of enhancement. MR imaging also has relatively limited sensitivity to DCIS, which is depicted as microcalcifications at mammography (12). Enhanced DSM was 83% specific in this study, and the low-energy source image showed microcalcifications, which could be used in the diagnosis of DCIS. Findings with MR imaging suggest that morphologic features (ie, shape and margin) help differentiate benign from malignant enhancing areas (13). Because enhanced DSM allows a higher spatial Volume 229
䡠
Number 1
resolution than that with MR imaging, differentiation of benign from malignant morphologic features at enhanced DSM should be easier. Enhancement kinetics, also used for differentiating benign from malignant lesions at MR imaging (14), can be determined at enhanced DSM with serial imaging. Because wholebreast images can be acquired more rapidly than with most MR imaging sequences, kinetic information could be determined with greater precision. Unlike MR imaging, however, each image has a penalty of additional radiation. Enhanced DSM should be less expensive than MR imaging because, although a digital mammography unit is much more expensive than a standard mammography unit, it is much less expensive to purchase and operate than is an MR imager. The expansion of existing mammographic core biopsy and preoperative
needle localization techniques to include enhanced DSM would be straightforward, given the right equipment. Such procedures are difficult to perform with MR imaging guidance because of the geometry of the MR imager, the time required for imaging, and the need to work in a high-strength magnetic field. Dualenergy enhanced DSM would be especially well suited for biopsy, since positioning can be adjusted after injection. Such adjustment might be needed with a biopsy device with a small field of view, because there would be no way to ensure that a mammographically occult lesion was included on a precontrast scout image. To allow complete biopsy of an enhancing lesion, enhancement would need to last at least long enough for targeting. Many aspects of enhanced DSM remain to be optimized. Neither the optimal injection rate nor the optimal timing Digital Subtraction Mammography
䡠
267
Radiology
are known, but results in this study indicate acceptable parameters for both. The two weakly enhancing cancers were imaged with nonstandard technique, as a result of failure of a tubing connection in patient 5 and a slower injection rate in patient 14 because the intravenous catheter was placed in the hand. The x-ray energy spectra used in this study is not optimal for dual-energy iodine imaging. The ideal high-energy beam would have a narrow spectrum centered just above 33.2 keV. Similarly, the ideal low-energy beam would be narrow and centered just below 33.2 keV. The commercial digital mammography unit we used is designed to produce x-ray energies that are optimal for unenhanced mammography. These energies are 18 –23 keV, which is far below the k edge of iodine, at 33.2 keV. The spectrum output with either molybdenum or rhodium anodes has characteristic peaks in this range. To have an adequate proportion of the energy above 33.2 keV for the highenergy exposure, we use high x-ray beam voltage and filter the beam with a large thickness of aluminum. While much of the low-energy portion of the beam is filtered, the remaining high-energy portion, as a result of bremsstrahlung, is very broad and much of it is below 33.2 keV.
268
䡠
Radiology
䡠
October 2003
In the long term, a different anode material may be desirable to obtain a more optimal high-energy beam. Dual-energy enhanced DSM is capable of demonstrating cancers that are not visible at standard mammography. We believe further studies are warranted for its evaluation.
7.
8.
9. References 1. Bird RE. Low-cost screening mammography: report on finances and review of 21,716 consecutive cases. Radiology 1989; 171:87–90. 2. Linver MN, Paster SB, Rosenberg RD, Key CR, Stidley CA, King WV. Improvement in mammography interpretation skills in a community radiology practice after dedicated teaching courses: 2-year medical audit of 38,633 cases. Radiology 1992; 184:39 – 43. 3. Baker LH. Breast Cancer Detection Demonstration Project: five-year summary report. CA Cancer J Clin 1982; 32:194 –225. 4. Lewin JM, Hendrick RE, D’Orsi CJ, et al. Comparison of full-field digital mammography with screen-film mammography for cancer detection: results of 4,945 paired examinations. Radiology 2001; 218:873– 880. 5. Ackerman LV, Watt AC, Shetty P, et al. Breast lesions examined by digital angiography. Work in progress. Radiology 1985; 155:65– 68. 6. Chang CHJ, Nesbit DE, Fisher DR, et al. Computed tomographic mammography using a conventional body scanner. AJR Am J Roentgenol 1981; 138:553–558.
10.
11.
12.
13.
14.
Muller JWT, Van Waes PFG, Koehler PR. Computed tomography of breast lesions: comparison with x-ray mammography. J Comput Assist Tomogr 1983; 7:650 – 654. Gilles R, Guinebretiere JM, Lucidarme O, et al. Nonpalpable breast tumors: diagnosis with contrast-enhanced subtraction dynamic MR imaging. Radiology 1994; 191:625– 631. Nunes LW, Schnall MD, Orel SG, et al. Breast MR imaging: interpretation model. Radiology 1997; 202:833– 841. Harms SE, Flamig DP, Hesley KL, et al. MR imaging of the breast with rotating delivery of excitation off resonance: clinical experience with pathologic correlation. Radiology 1993; 187:493–501. Buadu LD, Murakami J, Murayama S, et al. Breast lesions: correlation of contrast medium enhancement patterns on MR images with histopathologic findings and tumor angiogenesis. Radiology 1996; 200:639 – 649. Boetes C, Strijk SP, Holland R, Barentsz JO, Van Der Sluis RF, Ruijs JH. False-negative MR imaging of malignant breast tumors. Eur Radiol 1997; 7:1231–1234. Frouge C, Guinebretiere JM, Contesso G, Di Paola R, Blery M. Correlation between contrast enhancement in dynamic magnetic resonance imaging of the breast and tumor angiogenesis. Invest Radiol 1994; 29:1043–1049. Liberman L, Morris EA, Lee MJ, et al. Breast lesions detected on MR imaging: features and positive predictive value. AJR Am J Roentgenol 2002; 179:171–178.
Lewin et al