Dr. Oh received her medical degree from the Mayo Medical School in Rochester, MN in 1998. She is a third-year Radiology Resident at the University of Utah Hospital, Salt Lake City, UT. Dr. Oh plans to complete a Women's Imaging fellowship after her residency.

Magnetic resonance imaging (MRI) of the breast is emerging as an important tool in women's imaging. In the patient with breast implants, MRI provides an essential method for visualizing implants and their complications. In the patient with breast cancer, MRI is rapidly becoming a practical adjunct to mammography and ultrasonography. The imperfections of mammography have encouraged the growth and evolution of MRI indications.

Women's health is increasingly recognized as an important part of preventative medicine, with breast cancer detection and treatment at the forefront. With approximately 180,000 women diagnosed with breast cancer annually, it is the leading cause of death among women aged 40 to 55 years. ¹ Mammography has been the traditional screening tool for breast cancer, supplemented with ultrasonography for problem solving. It remains, however, an imperfect screening tool. Therefore, magnetic resonance imaging (MRI) is becoming increasingly popular for use in the clinical setting, with indications ranging from detection of implant failure to breast cancer evaluation.

Imaging breast implants

Breast implants are utilized for many different reasons, from self-image to breast reconstruction following mastectomy. Mammography does not provide the sensitivity needed for adequate visualization of implants; however, MRI does provide a viable alternative for more accurate evaluation. This is particularly important for the evaluation of patients with breast pain, irregularity, or concern over potential health-related complications associated with silicone leakage. Though no direct link has been proven between systemic disease and breast implant rupture, many patients would rather remove the prostheses than take that risk. However, many insurance companies will not pay for the removal of implants unless there is a documented complication, such as rupture. Accurate documentation of the breast implant is required in such cases, and MRI is the most effective modality in the diagnosis of implant rupture. ^2,3

Morgan et al ⁴ prospectively evaluated 345 breast implants in 185 women, and found that 100 implants showed evidence of rupture on MRI. Of these, 73 were subsequently removed surgically and correlated with pathologic findings of rupture. Similar to other reports in the literature, MRI demonstrated a sensitivity of 84% and a specificity of 92% for the detection of rupture in the removed implants. ^2,5

Imaging techniques

Ideally, a bilateral breast surface coil is used for comparison of sides in a patient with two implants. The patient is placed prone in the magnet during the study. In general, a typical protocol will include a T2-weighted fast-spin echo (FSE) sequence followed by a fat-suppressed inversion recovery sequence in the axial, coronal, or sagittal planes. Silicone-only sequences can be performed with the use of an inversion recovery FSE sequence with water saturation, or a three-point Dixon technique. ⁶ Selective silicone imaging can also be achieved with T1-weighted sequences utilizing silicone suppression (taking advantage of the difference in T1 relaxation time of silicone).

The normal implant

Most implants are single lumen, with an outer silicone elastomer shell filled with silicone gel. The less common double-lumen implant has an inner lumen containing silicone gel, surrounded by an outer lumen with saline. Both types of outer shells may be a smooth or textured silicone elastomer. Implants are placed in the sub-glandular or subpectoral positions. Reactive fluid around the textured implant is considered normal. ²

The implant contour should be smooth, but often with folds and invaginations. These are normal radial folds if they are at the margins of the implant and are thicker than the linear low-signal lines created by intracapsular rupture. Radial folds are generally thicker because of the two opposing layers of silicone elastomer, in comparison to the single layer seen in intracapsular rupture (figure 1). Additionally, radial folds frequently are perpendicular to the implant surface. Most patients develop a fibrous capsule surrounding the implant, caused by a foreign-body response, which appears as a rim of T2-hypointense tissue surrounding the silicone or saline implant. ³

The ruptured implant

There are two types of rupture, intracapsular and extracapsular. Intracapsular rupture is more common, with a breach of the outer elastomer without leakage beyond the fibrous capsule. Thin, hypointense lines within the high signal implant on T2-weighted images characterize the free-floating ruptured elastomer shell (figure 2). Silicone is seen on both sides of the elastomer shell as it collapses within the fibrous capsule. This is the "linguine sign" as originally reported by Gorczyca et al, ⁷ and is the most reliable sign for intracapsular rupture.

Extracapsular rupture may be seen on mammography, ultrasonography, or MRI. It is defined as leakage of silicone outside of the fibrous capsule. Extracapsular silicone is bright on silicone-only sequences or T2-weighted images with water saturation, and dark on T1-weighted silicone suppressed sequences (figure 3). Silicone can be seen anywhere in the adjacent breast parenchyma, axillary lymph nodes, or even within the pectoral muscle. ⁸ Siliconomas are the result of a host foreign body reaction, and can be seen on T2-weighted fat-suppressed MRI as hypointense collections.

Imaging breast cancer

MRI is rapidly becoming incorporated in the field of breast cancer imaging. As the most sensitive modality for imaging the breast, many potential uses have been, and continue to be, studied.

Imaging techniques

Given that breast MRI is still evolving, there is no universally accepted imaging protocol. In general, high field strength is thought to be advantageous, and specialized breast coils are required to increase temporal and spatial resolution. Bilateral breast coils are used for a comprehensive view of the breasts, to permit comparison of both sides. Single breast coils may also be used to gain better spatial resolution and an increased signal-to-noise ratio. T1-weighted images are obtained for visualization of the anatomy, and T2-weighted images for characterization of cysts and fibroadenomas. Fat-saturation following dynamic contrast administration is required to clearly delineate enhancing lesions from fat on post-gadolinium images. Fat-saturation may be active (prior to contrast administration) with a saturation pulse, or passive, using subtraction of pre- and postcontrast images. ⁹ Both methods have their drawbacks. Active fat-saturation can result in incomplete fat suppression. Passive techniques are susceptible to motion artifacts between the pre- and postcontrast images, causing the subtraction to be inaccurate. Other technical issues include the choice of two-dimensional versus three-dimensional volume acquisition, field of view, and the tradeoff in signal-to-noise ratios with volume averaging.

Spatial resolution

Morphology is best visualized with high-resolution MRI, and can help distinguish benign from malignant lesions, ^10-12 with the margins of a mass most helpful in differentiating malignant from benign. Spiculation or irregular margins, similar to mammography, suggest malignancy (figure 4). Peripheral rim enhancement is also associated with cancer, with a positive predictive value (PPV) as high as 93% for rim enhancement occurring with spiculated or irregular margins. ^10-12 However, only 36% to 65% of cancers have this feature, therefore the negative predictive value (NPV) is not as high. ^10-12 Clumped and/or linear enhancement is highly predictive of ductal carcinoma in situ (DCIS) (figure 5). ⁹ Features of benign disease include smooth or lobulated borders, absence of enhancement, nonenhancing internal septations, and low T2 signal. ^10-12 Focal masses with nonenhancing internal septations are benign in >90% of cases, the most common lesion being a fibroadenoma. ¹⁰ Even more fundamental than lesion morphology, Harms et al ¹³ initially used contrast in conjuction with a special fat supression technique to improve spatial resolution (RODEO [rotating delivery of excitation off resonance]). This allowed clear delineation of enhancing tumors from adjacent fat. A lesion was considered positive if its signal intensity was higher than that of breast parenchyma on static postcontrast images. Although the initial study does not address lesion morphology, they report a sensitivity of 94% and specificity of 37%. Nunes et al ¹⁴ recently published an update on a tree-shaped interpretation model based on architectural findings, developed to increase the sensitivity and specificity of breast MRI. Confirming their earlier work, they report a sensitivity of 96%, specificity of 75% to 80%, and NPV of 96% for their model. ¹⁴

Temporal resolution

Intravenous gadolinium enhancement provides a sensitive, but not specific, tool for the evaluation of breast lesions. Therefore, multiple methods have been developed to increase specificity. Kaiser et al ¹⁵ were the first to suggest rapid sequential imaging with limited spatial resolution. The group further proposed a quantification, to normalize enhancement with respect to the baseline signal intensity. In their model, enhancement is based on the equation: [(SIpost ­ SIpre)/SIpre] * 100. This equation has become the standard formula for measuring lesion enhancement. Kaiser et al ¹⁵ found that malignant lesions doubled their signal intensity within the first minute after contrast; however, DCIS may not enhance as rapidly and therefore can create false negative results. ^16,17 Gilles et al ¹⁸ had a similar approach, suggesting that early enhancement indicates malignancy. Their approach found a sensitivity of 95% and specificity of 53%. These theories were based on the fact that malignancies enhance before normal glandular tissue, which is typically more delayed. However, Kuhl et al ¹⁷ have since shown that normal glandular tissue can enhance during the menstrual cycle, with 73% of enhancing lesions resolving completely during follow-up. Some of these lesions even demonstrated initial enhancement rates beyond the malignancy threshold value (>80% increase in the first minute). ¹⁷ Boetes et al ¹⁶ used the single-section technique with high temporal resolution to image, and found a sensitivity of 95% with a specificity of 86%. This group further associated a centripetal pattern of enhancement with malignant lesions, while centrifugal was more likely benign. ¹⁶

To further clarify enhancement characteristics, Kuhl et al ¹⁹ observed that beyond the initial early enhancement, the time-signal curve could be an important criterion in differentiating malignant from benign lesions. This group classified lesions according to the curve types. Type Ia curves are continuously increasing over the dynamic period, type Ib curves increase and flatten late postcontrast. Type II is an initial upstroke followed by a plateau. Type III has an initial upstroke with an abrupt cut off and signal washout in the intermediate post-contrast period (ie, 2 to 3 minutes). ¹⁹ Type III curves have a likelihood of breast cancer in 87% of cases, whereas type I curves have a likelihood of only 6%. ¹⁹ Type II curves are seen in both malignant and benign lesions, but at a ratio of 3:2 malignant to benign. ¹⁹ These curves help to distinguish normal enhancing glandular tissue from malignancy, since normal parenchyma does not typically demonstrate the washout phenomenon after initial enhancement. ¹⁷ Kuhl et al ¹⁹ also concluded, however, that type I curves can be seen in lobular or scirrhous ductal cancer and the lack of a washout phenomenon cannot be used to exclude cancer.

In an attempt to merge the issues of temporal versus spatial resolution, Kuhl et al utilizes the following guidelines ^19,20 : Initial evaluation is made for lesions with significant enhancement, and those with suspicious morphology are biopsied. If the morphology is indeterminate or benign, the time-signal curve is reviewed. Lesions with type III curves are sent for biopsy, while those with type I or II curves are followed up. If a lesion has mild enhancement, management decisions are based on lesion morphology to include lobular carcinoma and DCIS. In their guidelines, rim enhancement leads to biopsy, while lesions with nonenhancing internal septations are diagnosed as fibroadenoma. ²⁰ With this type of interpretation, integration of both temporal and spatial data may further aid in the diagnosis of malignancy.

Interpretation

A consistent terminology system would facilitate accurate breast MRI interpretation, and would provide standardized descriptors similar to the Breast Imaging and Reporting and Data System (BI-RADS) for mammography. To this end, the Breast MRI Lexicon Working Group is developing a system of terminology as new research comes to light. Both morphology and kinetic findings should be included in a breast MRI report, with a final recommendation to convey the level of suspicion to the referring clinician. ²¹ Morris, ²¹ in conjunction with the Breast MRI Lexicon Working Group, proposed a table of descriptors (Table). This is still a work in progress, but is an initial attempt to standardize breast MRI reporting.

Clinical breast MRI indications

Despite ongoing research, many controversies remain. Consistent acquisition techniques are not yet established, and once the images are acquired, a formal reporting system does not yet exist. The mammography BI-RADS reporting system is inadequate for the intricacies of MRI, as it does not include contrast-enhancement characteristics. Furthermore, once a lesion is identified, there is a lack of consensus on the exact criteria required to diagnose breast cancer. Given the many unresolved issues and wide variety of imaging techniques, breast MRI is usually performed only in specific clinical settings. Other than for breast implant evaluation, there is no data to support the use of MRI for the general screening population, especially given the cost differential between MRI and mammography.

The equivocal mammogram

Mammography is an imperfect screening tool. MRI of the breast can help increase the sensitivity of lesion detection. Although breast MRI is very sensitive in the search for malignancy, its reported specificity varies widely in the current literature, from 37% to 98%. ^{11,13-16,18,19,22} This lack of specificity is caused by the overlap in the appearance of enhancement in benign and malignant lesions. Additionally, normal breast parenchyma may enhance with a similar pattern to malignant lesions depending on hormonal fluctuations. ¹⁷ Higher specificities are associated with those protocols evaluating dynamic contrast uptake. ^{15,16,18,19,22} Lee et al ²³ studied 86 lesions where equivocal findings were present on mammography, and concluded that breast MRI can be a useful adjunct in managing the problematic mammogram. They acknowledged, however, that false negatives could occur in cases of DCIS, invasive lobular carcinoma, and invasive ductal carcinoma. ^17,18 Also, they concluded that MRI should be used as an adjunctive tool only, when mammography and sonography are inconclusive.

Breast cancer staging and preoperative planning

The imperfect sensitivities of mammography and ultrasound have led to an attempt to use MRI to more accurately stage breast cancer, with a secondary benefit of allowing for preoperative planning. MRI can delineate tumor margins, assess involvement of adjacent structures, and screen for multifocal lesions (figure 6). Accurate preoperative planning may mean the difference between breast conservation surgery and mastectomy. Breast MRI may reveal unsuspected multifocal, multicentric, or contralateral breast carcinoma. Fisher et al ²⁴ studied the use of MR in a group of 463 patients, each with a previously discovered breast abnormality diagnosed on mammography, ultrasound, or clinical examination. Therapy was altered correctly in 14.3% of patients with a known breast abnormality, because MR imaging showed more extensive disease than had been appreciated initially. These included multifocal, multicentric, or synchronous, contralateral carcinomas. The rate of false positive findings was low at 3.5% (16 of the 463 patients had a lesion biopsied unnecessarily). DCIS poses a known limitation, however, because the sensitivity of MRI is low for detection of DCIS; 21 of their 30 false-negative findings were DCIS on histology. ^24-26

Axillary adenopathy with an unknown primary cancer

Although the incidence is low, at 0.3% to 1.0% of operable breast cancers, axillary adenopathy with an occult primary cancer can present a significant problem for clinicians. ²⁷ Patients who have a negative clinical examination and negative mammogram are traditionally treated with mastectomy and axillary node dissection. ²⁷ Breast MRI has been shown to be very sensitive in the detection of occult breast cancer in patients with malignant axillary adenopathy. ²⁸ MRI can provide more accurate localization of the primary malignancy, and can offer information on staging and treatment planning.

Screening for high-risk cohorts

Women who have the BRCA1 or BRCA2 gene or a strong personal or family history of breast cancer have an increased risk for the development of the disease. These women are a special group and are offered earlier education, clinical breast exams, and screening mammography than the general population. ²⁹ Concerned patients sometimes consider aggressive prophylactic treatments such as mastectomy. Breast MRI is a promising alternative to conventional evaluation, since its high sensitivity may allow it to be used as an effective screening tool in these women. Many of these patients are younger than the typical screening population, and interpretation can be difficult because of normal parenchymal enhancement, which is cyclical and related to hormonal fluctuations. ¹⁷ Additionally, incidental enhancing lesions, which are unlikely to be malignant, can be seen in up to 29% of imaged patients. ³⁰ In one study, the incidental lesions found were mostly seen in women who were younger, were premenopausal, or had dense breast parenchyma. ³⁰ Despite these challenges, early results indicate that breast MRI may be superior to mammography and ultrasound for the screening of women at high risk for hereditary cancer. ^31,32

Future directions of breast MRI

MRI biopsies

Breast MRI is emerging as a valuable adjunct to mammography and ultrasound, and can be used to detect primary breast cancer with a higher sensitivity as discussed above. What happens when a suspicious lesion is identified on MRI that was not seen on the diagnostic mammogram or ultrasound? Redirected ultrasound is one alternative, to permit an ultrasound-guided localization. A second alternative is MRI-guided localization using MRI-compatible stereotactic wires, with subsequent surgical resection. ³³ The pitfall with this approach is the inability to confirm lesion removal in the lumpectomy specimen. More promising is the core-needle biopsy, using MRI-compatible guns to allow tissue sampling. This, like core-needle biopsy under mammographic or ultrasound guidance, can be used to avoid the morbidity and costs of surgical resection. Kuhl et al ³⁴ reported a series of 78 lesions, with 99% of the lesions yielding sufficient material for histologic diagnosis, and a diagnostic accuracy of MRI-guided bore biopsy of 98%. Their technique includes a 14-gauge coaxial needle, with verification of placement on a T2-weighted sequence, possibly using a second contrast-enhanced series if the lesion was not adequately visualized. ³⁴ Difficulties included the "vanishing target" phenomenon, in which the core-biopsy needle obscures the lesion. ³⁵ MRI-compatible vacuum-assisted biopsy has also been found to be a reliable and accurate biopsy method. ^36-39

MR spectroscopy

MR spectroscopy is already used for evaluation of brain masses, to increase the specificity of MRI. Spectroscopy in the breast has been studied for the same purpose. In a study by Button et al, ⁴⁰ MR spectroscopy and relative regional blood volume mapping (rRBV) were used to evaluate patients who had rapidly enhancing lesions. They showed that of the 13 lesions with a prominent choline peak and "enhanced" rRBV, all were shown to be malignant tumors. The other 11 patients who had no choline or "enhanced" rRBV detected had pathologically benign lesions. Two fibroadenomas also demonstrated a detectable (but not prominent) choline peak. However, this group's research indicates that breast MR spectroscopy can be used to improve specificity.

MR elastography

At the 2001 Radiologic Society of North America meeting, Sinkus et al ⁴¹ addressed one of the newer techniques of breast cancer diagnosis. This group presented their work on dynamic MR elastography, used to assess the visco-elastic properties of tissue. The technique is performed following routine contrast-enhanced imaging, and illustrates different values of "stiffness" for different tissues. Discrepancies in breast cancer isotropic and anisotropic elastic properties potentially can be used to improve the specificity of breast MRI. ⁴¹

Conclusion

Breast MRI is becoming increasingly practical in the clinical setting, with advances in both technology and interpretation techniques. There are innumerable potential uses for breast MRI, with a few of the more extensively studied clinical indications discussed above. Other potential clinical applications include chemotherapy follow-up, radial scar evaluation, and postoperative residual tumor scanning. Research continues on new imaging techniques such as MR spectroscopy and elastography in the hope of increasing specificity. Many more clinical indications are expected to arise as our knowledge expands and we maximize the information provided by breast MRI.

Acknowledgments

The author would like to thank Dr. Maryellyn Gilfeather, Dr. Anne Kennedy, and Dr. Kevin Moore for their help in reviewing this paper.