You are in: eMedicine Specialties > Radiology > BREAST Magnetic Resonance MammographyArticle Last Updated: Jan 26, 2007AUTHOR AND EDITOR INFORMATIONSection 1 of 11
  Author: Shih-Chang Wang, MD, Associate Professor, Chair, Department of Diagnostic Radiology, National University of Singapore Shih-Chang Wang is a member of the following medical societies: Radiological Society of North America Coauthor(s): Robyn L Birdwell, MD, Associate Professor of Radiology, Department of Radiology, Harvard Medical School; Consulting Staff, Brigham and Women's Hospital and Dana-Farber Cancer Institute Editors: John M Lewin, MD, Associate Clinical Professor, Department of Preventative Medicine and Biometrics, Director of Teleradiology, Co-director of Breast Imaging Section, Director of Breast Imaging Research, Department of Radiology, University of Colorado Health Sciences Center; Consulting Radiologist, Diversified Radiology of Colorado; Bernard D Coombs, MB, ChB, PhD, Consulting Staff, Department of Specialist Rehabilitation Services, Hutt Valley District Health Board, New Zealand; Edward Azavedo, MD, PhD, Director of Clinical Breast Imaging Services, Associate Professor, Department of Radiology, Karolinska University Hospital, Sweden; Robert M Krasny, MD, Consulting Staff, Department of Radiology, The Angeles Clinic and Research Institute; Lawrence M Davis, MD, Assistant Professor of Diagnostic Imaging (Clinical), Department of Diagnostic Imaging, Brown Medical School Author and Editor Disclosure Synonyms and related keywords: MRM, MR mammography, contrast-enhanced breast MRI, contrast-enhanced MRI of the breast, breast imaging, magnetic resonance imaging of the breast, breast magnetic resonance imaging, breast MRI, magnetic resonance mammography, MRI mammography, MRI, breast cancer BACKGROUNDSection 2 of 11
  Incidence of breast cancer The incidence of breast cancer is slowly increasing worldwide. In developed nations in the Organisation for Economic Co-operation and Development (OECD), women have a lifetime breast cancer risk of approximately 1 in 12. The age-related incidence has also gradually increased by 3-4% per annum since the 1950s. In these nations (eg, United States, Australia, Canada, United Kingdom, Sweden), the incidence of breast cancer steadily increases with age, from about age 40 years to about 80 years. The incidence of breast cancer is only 0.05 case per 1,000 women undergoing screening. With the introduction of mammographic breast screening programs, the incidence has increased in the last 10 years, but this change has been accompanied by a shift to smaller lesions and by a marked increase in the detection of ductal carcinoma in situ (DCIS). In combination with earlier detection, improved and adjuvant treatments have actually resulted in a decline in breast cancer mortality rates in some countries in the last decade; this is a reversal of the previous mortality rate that had been inexorably unchanged since the 1950s. Over the last 20 years, the historically low rate of breast cancer in many other countries has also risen markedly, particularly in developing Asian nations such as Singapore, Malaysia, and Taiwan. In Singapore, for example, the incidence has risen from 20 cases per 100,000 women in 1968-1972 to 46 cases per 100,000 women in 1993-1997 (Chia, 2000). Although the overall incidence of breast cancer in Singapore is half that of the United States, the peak incidence in Singapore is almost 190 cases per 100,000 women in the group aged 35-45 years. This rate is similar to that in the United States, in the same age group. In addition, hereditary breast cancer has been increasingly recognized with the discovery of the BRCA1 and BRCA2 genes. Women with these genes account for only about 5% of all those of breast cancer, but compared with the healthy population, they are more likely to develop cancer before the age of 50 years. In this age group, the benefits of routine screening mammography are debatable, and future developments in magnetic resonance mammography (MRM) may hold promise for improving the detection of early cancers in these women. A number of studies are now under way to investigate the benefits of a comprehensive multimodality for breast screening in this cohort (Brown, 2000; Kuhl, 2000; Tilanus-Linthorst, 2000; Mansfield, 1979). Limitations of conventional breast assessment Conventional breast assessment includes the combination of screen-film X-ray mammography, high-resolution breast ultrasonography (US), and clinical breast examination. This tried-and-true method leads to the detection of approximately 85-90% of breast malignancies, and it forms the foundation of modern strategies for breast cancer detection. The results of this assessment also serve as the final arbiter for the use of breast biopsy in breast screening programs worldwide. Despite the usefulness of this approach, almost every breast radiologist is familiar with cases in which conventional assessment fails to depict a breast malignancy accurately, as in 1 of the following scenarios:
In short, in a significant minority of patients, breast malignancy is not adequately assessed by using conventional imaging and physical examination. In each of the scenarios above, MRM has been shown to be a sensitive and effective method of detecting, diagnosing, and staging intramammary breast malignancy, even when conventional imaging results have been negative. The use of MRM may change the clinical management in these situations if unexpected abnormalities are detected. Because of this ability to depict malignancies that are otherwise not visible, MRM has been the subject of active research and development around the world, and its use for certain specific indications has been accepted. Development of breast MRI The breast was one of the first organs studied with MRI for the detection of cancer, albeit initially in vitro (Mansfield, 1979). The breast was also the first organ in which the detection of invasive tumor neovascularity was highlighted through the application of rapid serial imaging after an injection of contrast agent. With the clinical application of nonenhanced breast MRI in the early 1980s, the value of T1-weighted (T1W) and T2-weighted (T2W) spin-echo imaging rapidly became clear, through the analysis of characteristics such as lesion morphology, signal intensity, and tissue relaxation times. In addition, some significant limitations of nonenhanced breast MRI also became clear. For example, the T2 relaxation rates of benign tissues and malignant tissues overlap (McSweeney, 1984), and in situ cancers could not be reliably detected at all. Initially, imaging was limited to 2-dimensional (2D) spin-echo acquisitions with intersection gaps and a significant limitation in the number of sections, which limited the overall sensitivity of the technique for small lesions. It was the combination of rapid 2D gradient-echo (GRE) imaging with a dedicated breast coil, coupled with the bolus injection of gadolinium dimeglumine (Kaiser, 1989) that created the technique of dynamic contrast-enhanced breast MRI. This technique showed an extremely high sensitivity for breast malignancy, which in some cases exceeded that of conventional imaging. Although this technique was initially limited to a single section location, it was soon modified with newly developed multisection spoiled GRE sequences, with no loss of sensitivity. This still forms the foundation of modern breast MRI. Since then, numerous developments and refinements have improved the diagnostic performance of breast MRI. The technical developments that revolutionized breast MRI and propelled it into its current importance as an adjunct technique for the evaluation of breast disease includes the following:
The current use of contrast-enhanced breast MRI to detect and stage breast malignancy, with a dynamic serial high-temporal resolution or slower high-spatial resolution, is often known as MRM. Pathophysiologic basis of breast cancer The basis of the high sensitivity of MRM is the tumor angiogenesis that accompanies a majority of breast cancers, even early ones. Contributing to this process is the secretion of factors such as vascular endothelial growth factor (VEGF), which in turn is strongly correlated with contrast enhancement (Frouge, 1994; Buadu, 1996). Conversely, the suppression of VEGF has been shown to reduce such tumoral enhancement (Pham, 1998). This angiogenesis has been shown to occur microscopically, even with DCIS, with a consequent increase in capillary density (Gilles, 1995). The abnormal permeability that accompanies the neovasculature leads to a rapid and marked leakage of the injected contrast medium at the site of neoplasia within the first few minutes after the injection; this characteristic permits tumor detection as areas of contrast enhancement during MRM. This factor is thought to be even more important than capillary density in determining tumor enhancement (Knopp, 1999). The clinical significance of neoangiogenesis in breast cancer remains somewhat unclear. Although the fact that metastatic disease cannot occur without angiogenesis is established, mixed evidence suggests that tumor angiogenesis is a good independent prognostic indicator for the survival outcome of women with breast cancer. Some researchers (Weidner, 1992; Toi, 1993) claim that it is, and others claim that it is more closely related to the age of the patient or the type of tumor (Miliaras, 1995). INDICATIONS AND CONTRAINDICATIONSSection 3 of 11
Advantages of MRM The many advantages of MRM over conventional breast imaging for the detection of malignancy have become apparent with increasing clinical experience. These advantages include the following:
Disadvantages of MRM However, the widespread use of MRM for the detection breast malignancy also has many disadvantages. These are as follows:
Prerequisites for MRM MRM is a highly specialized diagnostic technique that complements clinical assessment and conventional imaging with mammography and US. It does not replace these techniques except in certain unusual situations. In general, MRM should not be performed without conventional imaging first. MRM is best performed in a multidisciplinary setting with access to additional breast imaging, as well as close collaboration between the surgeon, radiologist, and pathologist. Radiologists experienced in MRI but without a strong knowledge of breast disease and diagnosis often have major difficulties with the interpretation of breast MRI studies. The radiologist embarking on MRM must (1) have a thorough understanding of breast pathology and the management of breast diseases, (2) work closely with a breast surgeon and pathologist to ensure the diagnostic accuracy of MRM, (3) be experienced in the interpretation of mammograms and breast sonograms, and (4) be experienced with image-guided breast needle-biopsy techniques. Indications for MRM Over the last decade, a variety of roles for MRM have been proposed. The high cost and limited availability of MRM, as well as the difficulties inherent in performing and interpreting the studies, require careful recommendations for its use. The following are common agreed-upon and useful indications for MRM:
MRM is used as an adjunct to conventional mammographic assessment because it is inconsistent in the diagnosis of DCIS, unless significant neoangiogenesis is present as well. This finding is typically seen in high-nuclear-grade DCIS. In some cases, DCIS has only weak enhancement that is indistinguishable from that of benign breast tissue. In addition, new applications for MRM are emerging. In particular, it is finding use in patients who are proven carriers of the BRCA gene or those who have undergone prior radiation therapy in the chest wall (eg, those with Hodgkin disease). However, while MRM has been shown to be effective for detecting small cancers before they are apparent with other imaging modalities, the cost of detection is high, and the low specificity means that women with suspect findings usually have to undergo multiple scans and MRI-guided breast biopsies (Kuhl, 2000). It remains to be seen whether this has any impact on survival and mortality. After radiation therapy and after the reactive inflammatory changes have settled, the normal breast tissue enhances only weakly. On this background of weak enhancement, any new focal strong enhancement is highly suggestive of malignancy, with high specificity. MRM in not always used to screen for breast cancer because the incidence of breast cancer is only 0.05 case per 1,000 women undergoing screening and because MRM is unreliable in detecting DCIS. However, whether MRM is useful for high-risk carriers is unclear. Ongoing studies are addressing the benefits of using MRM in conjunction with conventional imaging for surveillance screening in women who are known to be at high risk, either because they have the gene for breast cancer or because they have an extremely strong family history of breast and/or ovarian cancer in a first-degree relative. To date, these studies have shown that MRM can depict small invasive breast cancers before they become apparent on sonograms or mammograms. However, the cost of MRM is high, the images often have to be repeated to exclude cyclical enhancement, and no current evidence suggests a mortality benefit. Contraindications to MRM Conversely, in a number of situations, MRM is essentially contraindicated, usually because of physical constraints that prevent adequate patient positioning. These constraints include the following:
Relative contraindications also exist. These are essentially based on the high sensitivity but limited specificity of the technique. MRM may not be useful for the following:
Gadolinium-based contrast agents (gadopentetate dimeglumine [Magnevist], gadobenate dimeglumine [MultiHance], gadodiamide [Omniscan], gadoversetamide [OptiMARK], gadoteridol [ProHance]) have recently been linked to the development of nephrogenic systemic fibrosis (NSF) or nephrogenic fibrosing dermopathy (NFD). For more information, see the eMedicine topic Nephrogenic Fibrosing Dermopathy. The disease has occurred in patients with moderate to end-stage renal disease after being given a gadolinium-based contrast agent to enhance MRI or MRA scans. TECHNIQUESSection 4 of 11
EquipmentField strength Although MRM was initially successfully developed on the basis of magnets with field strengths of 0.35-0.5 T, it is now generally accepted that MRM should be performed at field strengths of 1 T or greater.
Breast coil design A dedicated double breast surface coil is essential because it permits simultaneous high-resolution, high-quality imaging of 1 or both breasts. An example of such a coil is shown (see Image 1). The patient lies prone with both breasts freely suspended in these coils. Such coils should have excellent homogeneity with minimal image shading and hot spots; otherwise some regions show low signal intensity and poor enhancement, and fat suppression may be unpredictable. A phased-array design is considered ideal today, but simpler coupled or switched coil designs are still commonplace. Newer designs are open on the sides, permitting access during imaging for interventional procedures, such as hookwire localization and needle biopsy. A symmetric coil platform design gives the patient the option of entering the magnet bore feet first or head first. Because of the need for IV access and because of claustrophobia concerns, the authors perform almost all studies with the patient entering feet first. Contrast agentsIn general, the group of extracellular fluid (ECF) paramagnetic gadolinium-based intravascular contrast agents have virtually identical pharmacokinetics and contrast-enhancement characteristics, and they are all equally suitable for breast MRI. All of these agents shorten T1 and increase tissue relaxation rates, increasing the signal intensity on T1W or spoiled GRE images. At the time of this writing, no agent has been shown to be superior to any other. Suitable agents include gadolinium dimeglumine (Magnevist, Schering), gadoteric acid (Gd-DOTA, Dotarem; Guerbet), gadoteridol (Gd-HPDO3A, ProHance; Squibb), gadodiamide (Gd-DTPA-BMA, Omniscan; Nycomed), and gadobenate dimeglumine (Gd-BOPTA, MultiHance; Bracco). The administration of contrast material should be by means of a rapid bolus injection. The authors routinely use an MRI-compatible power injector at a rate of 2 mL/s, followed by a 20-mL isotonic sodium chloride flush. Although they are less convenient, hand injections are also successful. For 2D GRE sequences, a dose of 0.16 mmol/kg appears to provide better sensitivity for lesion detection compared with the standard dose of 0.1 mmol/kg at field strengths less than 1.0 T (Heywang-Köbrunner, 1994). However, good evidence suggests that, with 3D gradient echo sequences at higher field strengths, the standard dose of 0.1 mmol/kg does not reduce the sensitivity of the test (Nunes, 1997), and the authors routinely use this dose successfully. Injection timing is important for routine MRM, but it is not as critical as it is for magnetic resonance (MR) angiography because of the inherent variation and unpredictability of the rate and intensity of lesion enhancement. Most investigators aim to acquire a complete post-contrast dataset within 1-2 minutes, preferably within 30-60 seconds. The authors time the injection to commence about 15 seconds before starting the first dynamic imaging sequence so that the contrast agent bolus arrives in the arterial phase at about the time that the central lines of k-space are being acquired in the 50-second acquisition used (see Ideal sequence, below). However, this time may vary with other factors such as the patient's age, cardiovascular status, and others. In general, these issues are not particularly critical if at least 1 dataset is acquired within 1-2 minutes after the injection. The exception to this is the use of high-speed dynamic first-pass contrast susceptibility imaging. First-pass T2*-susceptibility imaging is further discussed under Improved image acquisition, below. Patient compliancePatients undergoing breast MRI are often anxious because they either may have a known diagnosis of breast cancer or because they are expecting one. Other forms of breast imaging are rapid and usually do not require either an IV injection or prolonged enclosure in a long tunnel in which the patient experiences loud tapping noises. Anxious or claustrophobic patients can usually undergo imaging satisfactorily. In these patients, the physician and technologist can use a variety of strategies. For example, they do the following:
Experienced MRI technologists are crucial for successful completion of many studies. Failing these methods, extremely anxious patients may require IV sedation. The authors have found that 3-5 mg of midazolam via slow IV injection is effective. Imaging sequencesIdeal sequence Numerous MRM protocols have been published, and the inexperienced reader can easily be confused as to which to follow. Interestingly, while some observers have criticized this wide variety, all the published techniques are similar in the detection of breast malignancy. Therefore, it is more important for each site to be comfortable and familiar with a specific technique and protocol to ensure a high level of diagnostic accuracy than for them to try to always update to the latest sequence or technique. The ideal contrast-enhanced MRM sequence has the following parameters:
Because of current hardware limitations, no current sequence has all of these characteristics. However, novel methods for k-space filling, such as spiral 3D imaging, may allow this ideal sequence to be achieved in the near future. In general, all postcontrast imaging should be completed within 5-7 minutes because the diffusion of contrast material into normal tissues limits diagnostic characterization after this time. To detect suspicious contrast enhancement, imaging times shorter than 10-60 seconds are generally unnecessary; as many as 5-10% of carcinomas enhance relatively slowly, reaching peak enhancement at 3 minutes or even longer (Heywang-Köbrunner, 1996). Currently 4 clinically proven approaches are widely used to obtain satisfactory diagnostic accuracy and sensitivity with MRM:
In summary, crucial factors include the following:
Acquisition plane The acquisition plane has a significant impact on the pulse sequence. Phase-direction motion artifacts due to breathing and heart motion are minimized by ensuring that the frequency-encoding direction is in the anteroposterior direction for axial and sagittal imaging. In the coronal plane, switching the phase direction craniocaudally allows the use of a rectangular field of view with reduced phase encoding, reducing the acquisition time. However, intramammary mapping of lesion location can be difficult in the coronal plane, respiratory motion may produce unpredictable variations in signal across the images, and some coils and sequences generate unwanted moiré-like image artifacts at the edges of the breast due to variations in bulk tissue susceptibility. For these reasons, the authors prefer the axial and sagittal planes. These are also generally easier to correlate with the mammographic projections. Despite these problems with coronal imaging, the International Multicentre Breast MRI Study (Heywang-Kobrunner, 2001) used the coronal plane for dynamic acquisitions, with excellent results. Again, this testifies to the overall robustness of MRM as a technique, even with major differences in acquisition methods. Removal of fat signal In general, T1W sequences of any type that are sensitive to contrast enhancement are also highly sensitive to other intrinsic short-T1 substances, the most common of which in the breast is fat. While almost all large malignancies produce enhancement sufficiently strong to be detectable on routine T1W images, the signal intensity of fat is generally adequate to at least partly reduce the sensitivity of detection of small lesions or more subtle enhancement. Early publications on dynamic MRM quoted enhancement thresholds for malignancy in raw units of signal-intensity change. However, these figures were too field-strength and machine-specific to be used widely. A more robust and less instrumentation-dependent method is to use the percentage enhancement above baseline, with all values corrected to that of nonenhanced fat. The equation for this calculation is shown in Image 2. Various groups have reported that threshold enhancement values 50-90% above baseline have been reported as being highly suggestive of malignancy. However, such values vary with the field strength, pulse sequence, sequence timing, and dose of contrast agent. It is now accepted that no fixed enhancement threshold that invariably excludes invasive malignancy, though the absence of enhancement is rare in carcinomas. Four major approaches have been used to reduce or remove fat signal to show enhancement more clearly: GRE technique, digital image subtraction, spectral fat saturation, and magnetization transfer suppression (MTS). GRE imaging can be performed with a carefully chosen repetition time (TR), echo time (TE), and flip angle to minimize the (in-phase) fat signal while remaining sensitive to the presence of gadolinium enhancement. This approach was initially developed with field strengths of less than 1 T, and it tends to be less suitable with high field strengths, where the T1 of fat is longer. Because this method is usually a 2D technique with relatively thick sections, the risk of missing a small lesion is significant. Examples of images obtained with this technique are shown in Images 6-7. Pixel-by-pixel digital image subtraction with precontrast and postcontrast images is the only reliable method of fat suppression with low field strengths. This method permits the short imaging times required for MRM. Inversion recovery and Dixon techniques tend to be slow at low field strengths, largely because the poor SNR. Subtraction is the best means of canceling signal inhomogeneity across the breast at any field strength. However, it is time-consuming and susceptible to motion artifacts. Worse, misregistration due to patient motion may cause a lesion to become less visible. Therefore, if subtraction is used, the source images must be carefully reviewed. Despite a clear need, no commercial, automated nonrigid volumetric subtraction technique is available today. Spectral fat saturation using frequency-selective pulses is another technique. At less than 1 T, the spectral separation of fat and water resonances is too narrow for this technique to be reliable. Successful application requires careful shimming; a good coil design; and, frequently, manual preimaging tuning. Even so, homogeneous fat suppression may not be possible with large breasts. Nevertheless, this technique remains the best method of obtaining high-spatial-resolution 3D scans without resorting to subtraction. MTS method reduces the glandular tissue signal by using detuned saturation pulses prior to the imaging sequence; this shortens the water relaxation by coupling it to macromolecular motion. This technique as either an adjunct to or a substitution for spectral fat suppression achieves high sensitivity in terms of contrast enhancement sensitivity (Flamig, 1992; Santyr, 1996; Schreiber, 1996). However, this technique has not been formally studied to determine whether it is consistently superior to other methods. Of all these approaches, image subtraction and spectral fat suppression are the 2 most commonly used strategies for improving enhancement detection. Quantitation of enhancementThe use of region-of-interest (ROI) temporal-enhancement curves has been strongly highlighted by proponents of dynamic MRM. Three basic curve shapes have been described by Kuhl and co-investigators (Kuhl, 1999) (see Image 3): Type I is slowly enhancing in which gradual steady enhancement occurs over about 5 minutes. This has been further classified into type Ia, which is slow linearly progressive enhancement, and Ib, which is slow enhancement with a late plateau that produces a bowlike curve. Type II is a plateau with early strong enhancement (1-2 min) and a subsequent plateau phase. This is suspicious and often indicates malignancy. Type III is washout with early strong enhancement (1-2 min) and a subsequent decline in enhancement. This produces a characteristic peak that some investigators have dubbed the cancer corner. This enhancement curve is strongly associated with malignancy. The use of these curves is discussed in the next section, Hybrid MRM protocol. Current MR graphics workstations can have analysis software that produces pixel-by-pixel parametric color maps of the enhancement rate and intensity. These maps are arbitrary, but typically, lesions that are strongly and/or rapidly enhancing appear red, whereas slowly or weakly enhancing lesions appear blue or green. Intermediate lesions usually are orange or yellow. These have been developed (Knopp, 1994) and commercialized to simplify the analysis of the large amount of image data. These maps are helpful for the rapid evaluation of multiple foci of enhancement (see Image 4). Even if such a workstation is lacking, all MR consoles have ROI measurement capabilities. The detection of suspicious foci depends then on visual analysis, with ROIs placed over each suspect area and the mean signal intensity being read off the screen. This data can be entered into standard software programs, such as spreadsheet programs, for further analysis and charting. Hybrid MRM protocolA hybrid protocol using dynamic subtraction images and high-resolution fat-suppressed MRM is the authors' current protocol. This technique is performed at 1.5 T on a Signa MRI unit (GE Medical Systems). We use a GE Advantage Windows workstation with FuncTool software for image postprocessing, parametric map generation, and ROI enhancement measurement. A survey of a wide range of protocols has been published (Heywang-Köbrunner, 1996). This approach combines dynamic 2D or 3D images without fat suppression after contrast enhancement with slightly delayed high-resolution fat-suppressed 3D images. The authors have chosen the axial plane for the dynamic series because bilateral breast imaging is possible, and it permits comparison with craniocaudal mammographic projections. The authors try to avoid oversampling (no phase wrap [NPW]) to suppress spatial aliasing in the phase-encoding direction, because this increases the imaging time. High-resolution images are acquired in the sagittal plane because this permits faster acquisition than the axial plane if phase encoding is set to the head-foot direction and if NPW is turned off. These images can also be readily compared with mediolateral mammographic projections, and these are useful for section-by-section correlation with mammograms of mastectomy specimens. The authors find that the high-resolution axial images are often not useful because the leakage of contrast material into normal tissues may obscure lesions by the time these images are obtained. Nevertheless, on occasion, this sequence provides excellent depiction of lesions. The dynamic images are postprocessed by using batch subtraction, and then the location of each section is analyzed with parametric color maps to determine the lesion enhancement rate, peak, and time (see Image 4). Suspicious lesions are reviewed with a small ROI cursor with instantaneous automated curve plotting, and the results are compared with the enhancement curves from fat, with the normal glandular parenchyma, and with the enhancement in the aorta or internal mammary arteries. Table 1. Protocol parameters
Note.—All TEs are fractional echoes. FSE indicates fast spin-echo; FSPGR, fast spoiled GRE; eff, effective; ETL, echo train length; L-R, left to right; NEX, number of excitations; S-I, superior-inferior; SPF, swap phase and frequency; ZIP, zero-filled interpolation. *With a 0.1-mmol/kg IV injection of gadopentetate dimeglumine at a rate of 2 mL/s and a 20-mL sodium chloride push, the imaging time is 30 seconds. FINDINGSSection 5 of 11
Diagnostic criteria for malignant and benign processesThe following summary Tables show the diagnostic criteria that permit characterization of an enhancing lesion as either malignant or benign. The data have been collated from a variety of publications and represent pooled approximations of published results. Table 2. MRM criteria for malignant processes
Note–Values are from various authors and the architectural interpretation model developed by Nunes et al in 1997. PPV indicates positive predictive value. *Sherif, 1997. †Nunes, 1997. Table 3. MRM criteria for benign processes
Note–Values are from various authors and the architectural interpretation model developed by Nunes et al in 1997. NPV indicates negative predictive value. Normal tissuesFat Normal fat appears moderately bright on most images obtained with gadolinium-sensitive non–fat-saturated sequences used for MRM. For all practical purposes, normal fat does not enhance after the administration of gadolinium-based contrast material; thus, the signal intensity from the nearby fat may be used as an internal control for the standardization and correction of ROI enhancement calculations. With fat suppression, normal fat typically appears dark grey, except in areas where either paradoxical water suppression or poor fat suppression occurs; both are usually due to poor local field homogeneity or incorrect preimaging tuning. Glandular parenchyma The enhancement of normal glandular parenchyma may have several different patterns:
These patterns and the intensity of enhancement depend on the patient's age, the stage of her menstrual cycle, and the patient's prior medical treatment. Normal parenchyma enhances more strongly in the 35- to 50-year age range and least in the second and third weeks of the menstrual cycle (Mueller-Schimpfe, 1995). Nonspecific focal areas of enhancement may resolve or fluctuate in size from month to month (Kuhl, 1995; Kuhl, 1994). In practice, most dynamic MRMs show no enhancement, with variable amounts of parenchymal enhancement on delayed images. However, nonspecific regional enhancement may sometimes occur early; in these cases, it may be difficult to differentiate from malignancy (Nunes, 1997). The use of MRI-guided US and/or repeat MRM at a different part of the menstrual cycle may be necessary to confirm benign enhancement (see the discussions below). Nipple and areola Not infrequently, the normal nipple enhances intensely and rapidly. Retroareolar ducts may enhance normally, though usually not as intensely. In cases of extensive high-grade DCIS, these changes can make the interpretation of nipple enhancement problematic. In contrast, the areola does not normally enhance, and it appears only slightly thicker than adjacent skin; this characteristic permits the detection of areolar infiltration by a malignancy. Comparison with the contralateral side may be helpful if nipple disease is suspected; asymmetric hyperenhancement in the presence of DCIS may indicate the presence of nipple involvement, even in the absence of clinically obvious Paget disease (see Image 5). Intramammary and axially lymph nodes Normal intramammary and axillary nodes may enhance moderately intensely, either slowly or rapidly. They may even show a washout time-enhancement curve in the absence of malignancy. Therefore, the enhancement pattern is generally not useful for distinguishing benign nodes from involved nodes. A node may be diagnosed with confidence on high-resolution images by the presence of a fatty hilus and a sharply defined smooth contour. Careful comparison with the matching mammograms sometimes increases the diagnostic confidence if the typical mammographic appearance of the node can be demonstrated in the corresponding position. Postprocedural changesChanges after needle biopsy Normal tissues do not significantly alter their enhancement after routine needle biopsy. Foci of punctate hemorrhage or a focal hypointense ferromagnetic metal artifact are occasionally seen at the biopsy site, particularly on GRE images. Hematoma and seroma Hematomas and seromas are usually seen shortly after surgery, and they appear as fluid-filled nonenhancing, smoothly bounded cavities. Seromas are hyperintense on T2W images, and hematomas may be hyperintense or hypointense on T1W images, depending on the age and oxygenation state of the blood products. Both usually show a rim of irregular and hypointense but enhancing granulation tissue. This enhancement may be intense, potentially masking residual malignancy at the boundary of the collection. This reactive postsurgical enhancement decreases over time. A delay of at least 28 days after surgery is needed to achieve a reasonable specificity (75%) and an NPV (86%) for the detection of residual tumor (Frei, 2000). Fat necrosis Fat necrosis at the site of surgery usually manifests as a small nonenhancing hyperintense on T1W images. It appears as an ovoid focus with irregular stellate granulation tissue, which may be intensely enhancing after the administration of contrast material. This necrosis resembles a contracted hematoma or seroma (see Image 6). While this occasionally mimics a recurrent tumor, the typical lesion is usually readily diagnosable by its shape and enhancement pattern. Postoperative scarring Scar tissue generally appears as a low-signal-intensity linear irregularity with variable enhancement, which largely depends on the interval since treatment. In the first few months after surgery, the borders of the surgical cavity may have strong enhancement, particularly if hemorrhage or fat necrosis has occurred (see Fat necrosis above and Image 6). This reactive enhancement gradually subsides. At 6 months after surgery without radiation therapy, most images show slow, minimal, or no enhancement (see Image 7). In such cases, the appearance of abnormal enhancement at the scar after 6 months should raise the suspicion of a recurrent malignancy (Heywang, 1990). Again, tiny ferromagnetic hypointense sucker tip and instrument artifacts may be evident at the site of surgery, particularly if the native unsubtracted GRE images are reviewed (see Image 6). Appearances after radiation therapy In the first 9-12 months after radiation therapy, a diffuse increase in capillary permeability occurs. This change initially causes marked parenchymal enhancement that later becomes patchy. In most women, this enhancement gradually declines after 18 months because of fibrosis (Heywang-Köbrunner, 1993). As a result, normal tissues have minimal enhancement, and any enhancing lesion on this background is suggestive of a recurrent tumor, which may appear nodular with carcinoma recurrence or linear with DCIS (Gilles, 1993). Some patients may nevertheless benefit from MRM soon after surgery and radiation therapy if the presence of residual disease is strongly suspected. Although diffuse parenchymal enhancement is of little diagnostic value, the demonstration of typically malignant enhancement in a focal lesion should prompt repeat excision. Appearances after neoadjuvant chemotherapy MRM has been used successfully before, during, and after neoadjuvant chemotherapy to assess the preoperative tumor response in advanced local malignancy. Typically, such monitoring is performed by means of clinical palpation of the size of the tumor. However, this method can be highly inaccurate. For example, a good tumor response with necrosis and fibrosis may occur with only a slight reduction in the clinical size of a tumor. A clinical complete response may result in no enhancement or residual neovascularity visible on MRMs. In the latter case, this indicates residual viable tumor with a high degree of reliability. The correlation between MRM and pathologic findings of residual tumor is usually good (Abraham, 1996). Chemotherapy does not produce the initial edema response seen with radiation therapy, and parenchymal enhancement becomes bland soon after the administration of contrast material; this characteristic also helps in the differentiation of normal tissues from malignant tissues. Benign abnormalitiesFibrocystic disease Perimenopausally, fibrocystic change is common and best detected with T2W fast spin-echo images, on which cysts appear uniformly hyperintense. Multiple hyperintense, sharply circumscribed, nonenhancing, fluid-filled cysts of variable size are the hallmarks of this condition (see Image 8). As a result of their thick, viscous, inspissated contents, older cysts are occasionally hyperintense on T1W images or hypointense on T2W images. Mammary duct ectasia Mammary duct ectasia may be visible on precontrast T1W images because of high protein inspissated secretions and hyperintense dilated retroareolar ducts. Such ducts are readily overlooked, particularly with subtraction techniques. With fat-suppressed 3D images, they can mimic prominent enhancing ducts, and they are potentially misleading; the use of precontrast images prevents this pitfall. Unlike DCIS, this condition always has sharply defined lesions on high-resolution images, and they remain unchanged in intensity throughout the time course of the study. Sclerosing adenosis Sclerosing adenosis is typically indistinguishable from glandular parenchyma. However, it may appear as an irregular area of slow but strong focal enhancement. If dynamic imaging has not been performed, this finding may appear suspicious. Fibroadenomas may enhance rapidly and strongly when they are myxoid (usually in younger women). Such enhancement may be similar to invasive malignancy in terms of its rate and intensity, though a type III (washout) curve is not seen. They usually have sharply defined smooth or slightly lobulated borders with an ovoid shape (see Image 9). These features usually permit the distinction of a fibroadenoma from a malignancy, particularly mucinous or myxoid cancers. Although they are sharply defined, they are generally round. These features are analogous to the US criterion for the height-to-width ratio of a lesion: Solid nodules with a ratio of 0.5 are almost always benign, whereas those with a ratio of about 1 are generally malignant. Fibroadenomas may also have internal septa that do not enhance. This is a highly reliable sign of a benign lesion (see Image 10). When fibromas are hyalinized (usually in older women), they generally enhance more slowly and weakly than do carcinomas. Larger fibroadenomas may appear moderately bright on T2W precontrast imaging, whereas almost 90% of larger carcinomas have a signal intensity lower than that of normal parenchyma (Kuhl, 1999). Densely hyalinized fibroadenomas show minimal or no enhancement. Other benign tumors Papillomas may appear indistinguishable from fibroadenomas in many cases. Before enhancement, they are typically hypointense ovoid masses, and they may enhance strongly, moderately, or not at all (see Image 11). They usually have well-defined borders and can then be often classified as benign, though they must be distinguished from medullary and mucinous carcinomas. High-resolution US may be useful to determine whether the lesion is intraductal. Because of their cellularity and papillary growth pattern, pathologists usually recommend their excision to confirm the diagnosis. Lipomas and fibroadenolipomas (hamartomas) are benign mesenchymal lesions, and they are usually readily diagnosed with conventional imaging. Usually, they are imaged with MRM only incidentally. Characteristically, the lesions have internal fat, and they are easily distinguishable from malignancies. Breast infection In acute infective mastitis, conventional assessment is usually sufficient, and MRM has a limited role. The main differential diagnosis is inflammatory carcinoma, for which, again, MRM has few indications. If MRM is performed for the assessment of infections, strong rapid enhancement may be present, usually with a poorly defined regional or diffuse pattern. Granulomatous mastitis is a rare and problematic inflammatory condition that may mimic acute infective mastitis or even invasive breast cancer. It is usually idiopathic, but it may be caused by various mycobacteria or Actinomyces species, and it has also been described in association with sarcoidosis and Wegener granulomatosis. The patient experiences recurrent bouts of breast sepsis with sinus tracks and the discharge of purulent material. Cultures of the material are usually negative, and biopsy shows granulomatous inflammation. MRM shows areas of strong irregular enhancement around pockets of fluid-filled infective material (see Image 12). MRM can be useful in mapping the full extent of intramammary disease in this condition. This mapping permits a more accurate evaluation of the extent of involvement, which helps in planning and monitoring therapy. This condition is notoriously difficult to define and treat otherwise. Surgery or corticosteroid therapy is the treatments of choice. Borderline abnormalitiesProliferative dysplasias Proliferative dysplasias, including florid epithelial hyperplasia, usually appear as foci or regions of slow-to-moderate enhancement after the administration of contrast material. Generally, this enhancement is indistinguishable from that of normal parenchyma, and the diagnosis is made at pathologic examination. Atypical ductal hyperplasia Atypical ductal hyperplasia (ADH) is a recognized high-risk condition that may progress to DCIS and, eventually, invasive carcinoma. Not infrequently, its presence indicates an adjacent malignancy, which usually justifies surgical excision biopsy when it is diagnosed. Not surprisingly, ADH may have associated neoangiogenesis, which may result in focal or regional suspicious enhancement indistinguishable from that of a malignancy (Heywang-Köbrunner, 1996). Radial scar or complex sclerosis lesion These lesions are difficult to diagnose with all imaging modalities. Typically, they are detected as poorly defined, focal areas of stellate distortion on mammograms. They may have proliferative dysplasia, ADH, DCIS or even a focus of invasive carcinoma centrally. With MRM, they may enhance fairly intensely and show an irregular spiculated border (see Image 13). Sometimes, these may have an apparent central mass; in this case, they are indistinguishable from carcinomas. Surgical removal is routinely recommended for such lesions because pathologists cannot confidently diagnose these lesions by using needle biopsy samples. Juvenile papillomatosis Juvenile papillomatosis is an uncommon condition seen in young women, and it may produce an appearance of multiple masses with marked distortion on mammograms. On MRMs, this may appear as a network of enhancing beadlike nodules that are connected together by enhancing ducts (see Image 14). Alternatively, they may appear as a lobulated, enhancing mass with small internal cysts. Lobular carcinoma in situ Lobular carcinoma in situ (LCIS) is a nonmalignant proliferative condition that is a marker for increased risk of breast malignancy. It is usually indistinguishable from benign parenchyma on MRMs. However, it has been described as occasionally showing intense suspicious enhancement. MalignanciesDuctal carcinoma in situ DCIS has a variety of enhancement patterns on MRMs, including minimal or no enhancement. Any combination of these patterns may be present, and commonly, 1 or more is seen in association with invasive carcinoma. The enhancement patterns include the following: ductal, focal nodular, regional, and benign. The ductal pattern is a treelike, linear, branching or reticular pattern. It is usually seen with high-nuclear grade DCIS (Orel, 1997), and it is due to periductal angiogenesis. The enhancement may extend to the nipple; this suggests nipple involvement before Paget disease is clinically evident (see Image 5). Focal masses may be present at junctions of the branching pattern; these may be due to clumped DCIS, focal microinvasion, or small invasive carcinomas. In the focal nodular pattern, DCIS may appear masslike or stellate, and it may mimic an invasive carcinoma on conventional and MRM images. The diagnosis is usually made only by means of excision biopsy. This well-recognized but atypical type of DCIS is sometimes noncalcified. With the regional pattern, ductal enhancement may appear as a region of strong enhancement on early or delayed images. This pattern may be bandlike or irregular, or it may appear as a geographic area of enhancement (see Image 15). DCIS may also appear as a poorly defined region of strong enhancement, usually around an invasive malignancy. It may even enlarge the apparent size of a small invasive tumor; this is a cause of the overestimation of tumor size with MRM. In the benign pattern, 15-40% of DCIS shows minimal-to-moderate enhancement indistinguishable from that of normal glandular tissues. This pattern tends to occur in low-grade DCIS (Orel, 1997), but it has also been described in comedocarcinoma (Gilles, 1995). The variability of DCIS enhancement is due to variations in angiogenesis, which in turn is somewhat related to the histologic grade (Orel, 1997). DCIS of a high nuclear grade tends to have stronger enhancement than that of a lower-grade DCIS. Note that 40% of DCIS is not calcified, even when it is high grade (Evans, 1994) and present with concomitant angiogenesis. Paradoxically, although MRM is unsuitable for an evaluation of microcalcifications, it sometimes shows the extent of DCIS (calcified or noncalcified) better than mammography. Because this information has the potential to change the type and extent of surgical excision, MRM occasionally proves useful for the intramammary staging of DCIS, particularly when the noncalcified component is significant. For the evaluation of resection margins, however, no evidence suggests that MRM is useful, either in the early postoperative period or in the surveillance period after resection. The ultimate role of MRM in the management of DCIS requires further investigation. Invasive ductal carcinoma On MRMs, most invasive ductal carcinoma (IDC) appears as irregular, spiculated, or multilobulated nodular masses. They have strong rapid contrast enhancement that is at least 60% above baseline (see Image 16). Rim or inhomogeneous centripetal enhancement on dynamic scans may be present. Typically, either a type II or a type III enhancement curve is observed (see Image 3). Surrounding architectural distortion may be noted. About 5% of IDCs enhance slowly and/or less strongly, particularly if they are highly scirrhous. (Heywang-Köbrunner, 1996). Surrounding enhancement of variable intensity may represent DCIS, florid dysplasia, or benign parenchymal enhancement. In larger or multifocal lesions, MRM may show nipple or chest-wall involvement, which may not be otherwise evident (see Image 17). Multifocal IDC may show moderate segmental ductal enhancement connecting the masses; such masses are thus seen to be part of the same breast segment, even if they are not close to one another (see Images 18-19). Sometimes, internal enhancing septa are seen in invasive carcinomas; these must be distinguished from nonenhancing septations, which are typical in fibroadenomas. Invasive lobular carcinoma ILC accounts for 10-15% of breast carcinomas, and they can be mammographically occult or subtle in 20–40% of cases (Hilleren, 1991; Le Gal, 1992). As many as 85% are isointense relative to glandular parenchyma, and a minority have malignant microcalcifications (Newstead, 1992). The incidence of multifocal, multicentric, and bilateral synchronous or metachronous involvement is much higher with ILCs than IDCs; this involvement is found in as many as one half of all cases. Mammography and US tend to cause marked underestimations of the extent of ILCs. MRM has been shown to be more accurate, correctly demonstrating the extent of disease in about 85% of cases (Rodenko, 1996). In the authors' center, MRM is routinely used in all women with a preoperative diagnosis of ILC. In most cases, ILC shows focal irregular, strong, rapid enhancement typical of a malignancy. Single or multiple masses in 1 or more quadrants are sometimes demonstrated. However, ILC occasionally have weak or moderate enhancement, and they may be difficult to distinguish from glandular parenchyma with this criterion alone (Heywang-Köbrunner, 1996). Recognizing this problem is easier if the diagnosis is already known from needle biopsy results. In these cases, a masslike contour or associated architectural distortion is usually present. Despite this limitation, MRM is better than conventional imaging for preoperative staging of breast ILCs. Other breast malignancies Mucinous or colloid carcinoma may be well defined, with a lobulated border and homogeneous enhancement. Superficially, these tumors may resemble a large fibroadenoma. However, these lesions are typically round rather than oval. Enhancing internal septa may be visible; if present, malignancy can be correctly diagnosed (Nunes, 1997). If a great excess of mucin is present with relatively little malignant tissue, enhancement may be unremarkable or even absent in rare cases. Papillary and tubular carcinomas may enhance strongly and rapidly. However, some of these tumors have weak angiogenesis, which reflects their relatively low biologic aggression. These lesions may then enhance relatively slowly and/or weakly (Heywang-Köbrunner, 1996). Non-Hodgkin lymphoma of the breast is rare and usually secondary to extensive involvement elsewhere in the body. It appears as a focal, well-defined mass with suggestive enhancement (Heywang-Köbrunner, 1996). Primary lymphoma may be synchronously or metachronously bilateral; in some cases, it grows rapidly to become large. An unwary pathologist may occasionally confuse this type of malignancy with an invasive carcinoma, particularly if needle-biopsy specimens are examined. MRM results are accurate in staging the intramammary extent of disease in such cases, and MRM may be useful for monitoring the response to subsequent chemotherapy, as with neoadjuvant therapy for large carcinomas. In the breast, sarcomas are rare malignant mesenchymal neoplasms. Metaplastic carcinomas are also rare, but they may undergo sarcomatous transformation, most typically to osteosarcoma. These generally appear as an otherwise nonspecific focal mass with suggestive enhancement. Sometimes, they are remarkably rounded with well-defined margins, but they may have markedly heterogeneous enhancement secondary to tumor necrosis. Phyllodes tumors are usually diagnosed with mammography, US, and needle biopsy, and MRM adds little other than a true size measurement of large lesions. These lesions appear as large, well-circumscribed masses with rapid, strong enhancement; they often have internal lobulation and cystic spaces (Heywang-Köbrunner, 1996). SPECIFICITY OF MRM AND DIFFERENTIAL DIAGNOSISSection 6 of 11
If early rapid enhancement due to neovascularity were unique to malignant tissues, MRM would be the standard in clinical practice today. Unfortunately, such enhancement is not specific, and several benign conditions may enhance in a fashion similar to cancer. Conversely, a small percentage of malignancies either enhance identically to benign breast parenchyma, or rarely, they do not enhance at all. Nonmalignant pathology that may mimic malignant enhancement:
Malignancy that may show benign-type enhancement:
With any single criterion, such as the presence of early enhancement, a fixed enhancement threshold at a specific time point or a simple analysis of the features of the lesions leads to relatively low specificity rates of as low as 37%, despite the high sensitivity of 91% (Kuhl, 1999). Clearly, such simplistic analysis is insufficient for high diagnostic accuracy, and 2 major and somewhat opposing approaches have been used to try to improve diagnostic specificity. These are time-enhancement analysis using dynamic MRM and detailed lesion architectural analysis using a single high-spatial-resolution 3D acquisition. Temporal enhancement analysis The use of ROI temporal-enhancement curves has been strongly highlighted by proponents of dynamic MRM. As noted previously, 3 basic curves have been described: (1) type I, or slowly enhancing; (2) type II, or plateau; and (3) type III, or washout. In a recent study to determine the value of using such signal-time measurements with dynamic 2D MRM, Kuhl et al studied 266 lesions. These were subsequently excised, and 101 were proven to be malignant. The researchers showed that 9% of breast cancers had a type I enhancement curve; 33.6%, type II; and 57.4%, type III. Conversely, 83% of benign lesions had a type I curve; 11.5%, type II; and 5.5%, type III. In short, lesions with type I enhancement were more likely to be benign than malignant, whereas lesions with a type II or III enhancement curve were more likely to be malignant. In this analysis, the sensitivity, specificity, and diagnostic accuracy were 91%, 83%, and 86%, respectively. These results were significantly better than those obtained with the use of simple enhancement threshold measurements alone. The development of neoangiogenesis in cancers results in early rim enhancement with centripetal slower internal enhancement; conversely, benign lesions tend to enhance centrifugally the temporal resolution is sufficient (Boetes, 1994; Mussurakis, 1998). Such enhancement heterogeneity can be assessed by the use of small ROIs, with the user roaming around the image to detect the area of most suspicious enhancement. However, this approach is tedious and prone to operator error. Marked interoperator variation can occur in the ROI measurements in the same image datasets, particularly between experienced and inexperienced readers (Mussurakis, 1996). The use of automated pixel-by-pixel color parametric map analysis of the enhancement rate and intensity is useful for improving reliability of assessment of the large number of images typically obtained with dynamic MRM (Mussurakis, 1995). This technique can also be used to instantly graph the time-enhancement curve as the cursor is moved across the dataset; this information adds to the diagnostic confidence in assessing the nature of an enhancing area. High-resolution spatial analysis of lesion architecture In 1997, Nunes et al published a somewhat complex but comprehensive, and most importantly validated, image-analysis decision model to improve the classification of enhancing lesions depicted on low-temporal-resolution high-spatial-resolution 3D fat-suppressed (3D FSPGR) images. This detailed method yields high diagnostic accuracy and has formed the basis for subsequent studies of architectural-feature assessment. Recently, this model was further validated, updated, and slightly modified. The resultant values for sensitivity, specificity, NPV, PPV, and accuracy were 96%, 80%, 96%, 78%, and 87%, respectively (Sherif, 1997). These diagnostic performance indicators are slightly more sensitive but otherwise similar to those obtained by enhancement–time course analysis by using dynamic 2D MRM. Combined qualitative and quantitative analysis Recently, Liu et al studied the diagnostic performance of combining enhancement quantitation with qualitative feature analysis and obtained sensitivity, specificity, and accuracy statistics of 93%, 74%, and 85%, respectively (Liu, 1998). These statistics are essentially identical to those obtained with either time-enhancement–curve analysis or high-resolution spatial analysis. DIAGNOSIS OF LESIONS WITH MR TECHNIQUESSection 7 of 11
The increasing use of MRM is inevitably accompanied by incidental enhancing abnormalities, which typically had not been detected on earlier conventional images. These apparent lesions may represent normal or dysplastic tissues, cyclic hormonal changes, benign tumors, or even unexpected malignant foci. If the enhancement rate, intensity, or pattern are suspicious, the nature of such foci must be clarified to not miss a cancer. Three strategies are commonly used to diagnose these lesions: performing MRI-guided repeat US, repeating the MRM examination at another suitable time, or performing MRI-guided needle biopsy. MRI-guided US High-resolution US of the suspect region of the breast should be the first method used because it is rapid and can be performed immediately after the MR examination. MRM is used to guide the examination to determine the size, shape, and position of the suspect lesion. Using this technique, the present authors and others have found malignancies previously missed with routine US, particularly in cases with multifocal or occult malignancies (see Image 19). Sometimes, such lesions are subtle. They may even appear benign on sonograms, being confidently detected only because the operator was aware of the existence of the lesion. If MRI-guided US depicts the same lesion as does MRM, US-guided needle biopsy and/or localization of the lesion for excision can then be routinely performed. However, even when the examination is performed with care and experienced operators, MRI-guided US can still fail to depict a tumor, possibly because of its size, location, or US characteristics. Repeat MRM In premenopausal patients, cyclic hormonal enhancement is a common cause of false-positive focal or multifocal enhancement. Although efforts to image such women in midcycle usually reduce the incidence of such foci, repeat midcycle MRM can show whether an apparent lesion changes from one cycle to the next. This is characteristic of hormonally influenced benign tissue. If a repeat examination shows a persistent abnormality with suspicious features that cannot be localized with US, the choice is to either continue to observe the lesion or to perform breast biopsy to achieve a definitive diagnosis. The decision should be based on the level of suspicion of the MRM findings; the options should be presented to and discussed with the patient. MRI-guided biopsy or localization MRI-guided hookwire localization or some form of needle biopsy is an inevitable consequence of performing MRM, because MRM can depict lesions that are occult with all other forms of breast imaging. A number of research centers have been independently developed and reported various techniques, confirming that MRI-guided biopsy is at least practical and that it can be used to verify a malignant diagnosis of otherwise occult neoplasms (Heywang-Köbrunner, 1994; Heywang-Kobrunner, 1999; Spielmann, 2000; Orel, 1994). Early breast coils had a completely closed design that precluded biopsy or needle-based interventions in the breast. Recently, breast coils that have an open design have also been developed for routine imaging. These permit free access to the breast from most directions. Unilateral dedicated biopsy coils that have been developed recently allow complete access to the breast and much of the chest wall and axillae. Several investigators have developed various techniques for MRI-guided fine-needle biopsy or localization (Heywang-Köbrunner, 1994; Heywang-Kobrunner, 2000; Fischer, 1994; Fischer, 1995; Fischer, 1995; deSouza, 1995; Wald, 1996; Doler, 1996; Fischer, 1998; Daniel, 1998). These techniques range from simple freehand techniques, which remain useful for hookwire localization, to complex, robotic, automated systems, which are intended primarily for vacuum-assisted biopsy with large-bore needles. Stereotactic needle guides and non-ferromagnetic biopsy needles and hookwires made specifically for breast MR interventions have become commercially available only in the last few years. Typically, these devices are non-ferromagnetic, they compress the breast mediolaterally, they have a perforated needle-guide compression plate, they have MRI-visible markers, and they use mechanical needle positioning. Currently, none of the available devices show any clear superiority over other designs. User preference and familiarity with conventional image-guided breast biopsy are critical factors for success, regardless of the device and choice of needle-biopsy technology. Typically, imaging is performed with gentle mediolateral compression by using compression plates to stabilize the breast. It is important not to apply too much pressure because this may reduce lesion enhancement, sometimes markedly (Heywang-Köbrunner, 1996). After contrast enhancement and appropriate targeting, the suspicious lesion is punctured with an MRI-compatible (non-ferromagnetic) needle by using repeated imaging to confirm the position of the needle. Needle biopsy can then be performed once the lesion is accurately localized, or a hookwire can be deployed for subsequent surgical excision. Although time-consuming, this procedure can confirm malignancy when the results of all other tests are negative. FUTURE DEVELOPMENTSSection 8 of 11
MRM has already shown itself to be helpful in certain clinical settings in which conventional imaging is less sensitive. With either careful time-enhancement or lesion architectural analysis, high levels of diagnostic performance are possible. However, the complexity, the long learning curve, and the high cost of MRM and the still-significant nonspecific abnormalities that are detected has limited the widespread adoption of the technique. Ongoing research into methods of improving the diagnostic performance and accuracy of MRM must continue if the technique is to finally (and properly) become part of the standard breast imaging armamentarium. To this end, 3 major areas of research and development are being pursued to improve the diagnostic performance of MRM, or evaluate new applications. These areas include improved image acquisition, computerized image analysis, and improved dynamic subtraction. Improved image acquisition A variety of approaches have been developed to increase temporal resolution, spatial resolution, or both, with the aim of improving diagnostic specificity and improving the visualization of early intense enhancement. Broadly, these are divided into rapid GRE imaging, first-pass susceptibility imaging, limited k-space acquisitions (keyhole imaging), and non-Cartesian k-space filling methods. Regarding rapid GRE enhancement imaging, the use of a fast sequential spoiled GRE technique with a temporal resolution of about 1 second has been described for analysis of a single lesion after the rapid injection of a contrast agent bolus (Boetes, 1994). With this method, malignant lesions typically have centripetal enhancement, whereas benign lesions with otherwise similar time-enhancement curves show centrifugal enhancement. However, although this technique exploits the distribution of lesion vasculature and is perhaps more specific than standard dynamic MRM, it remains impractical for a survey of the breast and for an analysis of lesions smaller than 5 mm in diameter. The technique of first-pass T2*-susceptibility imaging was initially developed for brain imaging for the purpose of detecting and quantifying tissue perfusion and blood flow. More recently, this technology was successfully applied to the analysis of single breast lesions (Kuhl, 1997). With this technique, normal breast parenchyma shows no signal-intensity change, cancers show a marked initial loss of signal due to first-pass perfusion, and benign lesions (eg, fibroadenomas) show no such signal-intensity loss, even when the time course of their T1 enhancement and intensities are similar to those of a malignancy. This method is highly specific and can even be applied after standard MRM is used. Unfortunately, the technique has limitations that prevent its widespread adoption. These include the following: the need to use an additional bolus of contrast material, with the associated additional cost and time; the current inability to image the entire breast with this technique; limitations in evaluating lesions smaller than 5 mm; and susceptibility artifacts resulting from air-tissue interfaces. Ongoing technical developments to improve multisection volumetric perfusion imaging may permit imaging of the entire breast with this technique; these may lead to wider use of this technique. The keyhole technique for dynamic contrast-enhanced MRI was introduced in the mid-1990s. It combines a high-spatial-resolution non-enhanced dataset with dynamic high-temporal-resolution acquisitions of only the central lines of k space after the injection of contrast material. This procedure is used to produce synthesized contrast-enhanced datasets with a large image matrix, with features of both a high-temporal-resolution and a high-spatial-resolution acquisition. Keyhole imaging has since been largely abandoned because it produces significant artifacts (Plewes, 1995), it has major limitations for the detection and characterization of small lesions (Bishop, 1997), and newer technical developments in hardware and software have made it unnecessary. Novel k-space-filling strategies are experimental approaches to increase the speed of high-resolution volumetric acquisitions. These techniques all involve reducing the time required to fill k space. Of these methods, the recently described, and as yet non-commercial, spiral 3D acquisition appears to be the most promising at present (Yen, 2000; Daniel, 1998). This technique perhaps comes the closest to the Ideal sequence described above. Computerized image analysis Except for the single high-resolution 3D acquisition strategy, all MRM techniques produce a large number of images, often with differing time points. Direct visual inspection, with or without automated parametric analysis, remains the standard method of diagnostic appraisal of these studies. However, accurate interpretation requires much experience and careful time-consuming analysis. Recently, computerized image analysis methods using the automated extraction of architectural features and processing with artificial neural networks have been reported (el-Kwae, 1998; Abdolmaleki, 1997). Computerized multifactorial feature analysis using a semiautomated system has also been developed (Partridge, 1999). Both of these approaches allow the analysis of a large number of factors and features in a short time, and they appear to markedly reduce the time for dataset analysis without a loss of accuracy. Neither approach replaces the radiologist, but both approaches result in a marked reduction of the time spent in analyzing most of the dataset containing only normal or benign findings. Improved dynamic subtraction Subtraction postprocessing for dynamic MRM has been shown to be highly desirable for detection of some suspicious lesions. However, a major limitation of subtraction postprocessing is misregistration of the precontrast and postcontrast datasets. Patient respiration and minor movements between acquisitions can make subtraction useless, creating apparent artifactual lesions. Or even worse, artifacts can mask true malignant foci. Automated motion tracking during MRM is reportedly useful (Zuo, 1996), but it has a limited application. Standard software on MR workstations permits only section-by-section subtraction with no adjustment or interactive shifting of the datasets. In reality, the motion problem is complex because any part of the readily deformable breast tissue has 6 degrees of motional freedom, and the adjacent sections to be simultaneously adjusted are numerous. In-plane rotation and through-plane translation make simple pixel shifting impractical and inaccurate. Tissue distortion may also occur between datasets as a result of gross patient motion because the breast tissue is so easily deformable. Therefore, a general solution to this problem requires both automated computerized volumetric feature matching and elastic image warping to optimize the quality and accuracy of the subtractions. Recently, this approach has been successfully applied; it shows promise in markedly improving the registration of such subtracted datasets (Lucht, 2000; Rueckert, 1999). Improved contrast agents A problem with the current generation of contrast agents used for MRM is that they are universal agents with a rate of high leakage through normal capillaries into the extracellular fluid compartment. This leakage results in the nonspecific enhancement of nonmalignant tissues, which is so troublesome in MRM today. The design of a contrast agent with more specific affinity to abnormal neovasculature is highly desirable, but this has yet to be achieved. Currently, the most popular approach is to use macromolecular agents, which have a high relaxivity and large molecular weight. Such agents may be polymeric, dendritic, or encapsulated. They may even transiently attach themselves to serum proteins to achieve the desired level of prolonged intravascular retention. Typically, these agents do not pass through normal capillary membranes, but they are designed to leak through the abnormal hyperpermeable membranes of tumor vessels. This approach has shown promising results in animal experiments using the prototypical agent albumin–gadopentetate dimeglumine. Although selective tumor neovascular mapping and even strong correlation with tumor histologic grade may be obtained (Turetschek, 2001; Daldrup, AJR Am J Roentgenol 1998; Daldrup, Pediatr Radiol 1998; van Dijke, 1996), whether a suitable safe contrast agent of this type can be developed for human use remains to be seen. Ultrasmall particulate iron oxides (USPIO) have also been successfully used to highlight the increased microvascular density. USPIO are also useful in calculating capillary permeability in implanted breast cancer tumors (Turetschek, 2001), despite their being larger than macromolecules. CONCLUSIONSSection 9 of 11
|