Applied Radiology

Dr. Ingle is chief resident in Diagnostic Radiology at the University of Tennessee at Memphis. He graduated with honors from William Jewell College and Oxford University in 1993, and received his DO from the University of Osteopathic Medicine and Health Sciences in Des Moines, IA. Dr. Ingle plans a fellowship in musculoskeletal imaging at Indiana University beginning July 2002.

Breast cancer accounts for one out of every three cancers diagnosed in women in the United States, with approximately 175,000 new cases and nearly 43,000 deaths annually. Only lung cancer accounts for more cancer deaths in women. ¹ With these sobering numbers in mind, emphasis remains on early detection, diagnosis, and staging of breast cancer, with the goal of improving survival and treatment options.

It is generally accepted that the widespread use of routine screening mammography has played a key role in improving detection and survival. Beginning in 1989, the number of breast cancer deaths decreased significantly, ^1,2 and rates have continued to decline at an average of 1.8% per year. ¹ One study from Malmo, Sweden reported a 43% decrease in the death rate from breast cancer in that city since the introduction of screening mammography. ³

Although screening mammography has had a tremendous impact on overall outcomes, the false-negative rate reportedly ranges from 8% to 10%, ⁴ with some studies suggesting that as many as 25% of breast cancers may be missed on screening mammography. ^4,5 Several recent studies have found MRI to be much more sensitive than mammography for detecting breast cancer. ^6-11 Four of these studies also found MRI to be more sensitive than mammography in defining tumor extent (figures 1 and 2) with reported sensitivities of 100% vs. 33%, ⁶ 85% vs. 32%, ⁷ 98% vs. 55%, ⁸ and 96% vs. 44%, ⁹ respectively.

Contrast-enhanced MRI

Contrast-enhanced MRI has very high sensitivity for detecting invasive breast cancer--approaching 90% to 100% ^10-16 --and can detect mammographically and clinically occult breast cancers (figure 3). ^6-17 One advantage of contrast-enhanced MRI over conventional imaging is the ability to depict malignant changes caused by tumor angiogenesis. ^{9,10,15,18,19} Tumor angiogenesis (the induction of new capillary vessels in response to specific mediators produced by tumor cells) is crucial to the development of various tumors, including breast cancer. Such neovascularity is one of the most important characteristics distinguishing breast cancer from benign processes. ^9,15,18 In fact, the presence of intratumoral, marginal, and peritumoral vascularity on MRI as an indicator of malignancy has been shown to have a positive predictive value of 95%, 87%, and 92% respectively. ¹⁸ Angiogenesis has also been shown to correlate with tumor aggressiveness. ¹⁵

In conjunction with angiogenesis, the contrast-enhancement characteristics and patterns can be used to distinguish benign and malignant lesions. Malignant tissues tend to enhance earlier than benign lesions or normal glandular tissue, with a stronger and faster increase in signal intensity. ^{9,10,14-16,20-23} Breast malignancies demonstrate early enhancement after contrast administration due to a preferential accumulation of gadolinium (figure 4). This shortens the T1 relaxation of water protons, thereby yielding brighter signal intensity on the T1-weighted images. ⁹ An altered or increased vascularity in breast cancers relative to benign or normal breast tissue is the likely reason for this stronger and earlier enhancement. ^9,15,18,23

One method of increasing both the sensitivity and specificity of breast MRI is assessment of enhancement kinetics and the signal intensity time course. This is accomplished by placing a region of interest (ROI) in the most enhancing part of the lesion with a subsequent plot on a time-signal intensity curve. ^{6,10,15,20-23} A relative increase in signal of 80% or more in the first post-
contrast minute is a sign of malignancy, while lesion with values less than 80% are usually benign. ¹⁵ Several studies have shown added specificity by evaluating the intermediate and late post-contrast periods. ^{6,10,15,20-23} Based on these studies, three enhancement patterns have been described: type I demonstrates steady enhancement; type II demonstrates plateau of signal intensity; and type III demonstrates washout of signal intensity. ^6,10,20-23

Studies by Kuhl et al ^6,20 and Kinkel et al ²² have shown that the contrast washout time-signal intensity curve is extremely useful in differentiating benign and malignant lesions. A washout pattern is characterized by an initial upstroke, after which enhancement is abruptly cut off, followed by a signal intensity decrease (washout) in the intermediate post-contrast period. Kuhl et al ²⁰ showed an 87% likelihood of malignancy with a washout pattern and Kinkel et al ²² reported an 85% likelihood. In comparison, a steady time-signal intensity curve, characterized by continued increase over the entire dynamic period, appeared in 94% of benign lesions in one study ²⁰ and in 95% in the other. ²² Kuhl et al ²⁰ found the signal intensity time course to be substantially more specific in identifying breast cancer than enhancement rate alone, with an improved specificity of 83% vs. 37%, respectively.

Integration of both morphologic features depicted on breast MRI and the dynamic enhancement characteristics are needed to optimize the diagnostic capability of breast MRI. ^{10,11,13,15,16,20-23} This is evidenced by the fact that approximately 10% of malignancies--including lobular, medullary, and some ductal carcinomas, as well as metastases--will enhance slowly this is a pattern typically seen in benign lesions. ^15,21 In these instances, morphology may be critical to making the correct diagnosis. Lesions with irregular, ill-defined or spiculated borders (figure 5); peripheral enhancement; and ductal enhancement are suggestive of malignancy. ^{6,7,10-13,15,16,20-23} Morphologic features suggestive of benign disease include: smooth or lobulated borders, lack of contrast enhancement, non-enhancing internal septations, and patchy parenchymal enhancement. ^13,16,21,22 Nunes et al ¹³ found these characteristics to be highly predictive of benign disease. Lesions with smooth borders demonstrated a 100% negative predictive value (NPV), while lobulated borders had an 87% NPV. Additionally, they demonstrated that the presence of spiculated borders had an 88% positive predictive value (PPV) for malignancy. ¹³

Combining the morphologic features and dynamic enhancement characteristics has led to the following proposed diagnostic criteria: any lesion with spiculated margins, regardless of enhancement pattern, or lesions with lobulated margins that demonstrate washout are considered malignant. Lesions without spiculated margins or washout are considered benign. ^13,15,20,22 Using these criteria, Kinkel et al ²² reported a NPV of 96% and PPV of 97%.

Despite some general agreement among radiologists regarding the importance of both morphologic features and dynamic enhancement characteristics, no standard interpretation guidelines have been formally established. ^21,22 Most published diagnostic criteria have involved site-specific software programs or unique MR sequences. ²² One reason for the lack of a general consensus regarding interpretation criteria is that considerable overlap exists between the enhancement characteristics of benign and malignant lesions, ^{13,15,16,21-23} with specificities ranging from 37% to 97%. ^13,15,16 Benign lesions such as fibroadenomas, fibrocystic changes, radial scars, mastitis, atypical ductal hyperplasia, and lobular neoplasia have been shown to enhance similar to malignant lesions. ^16,21 Normal breast tissue has also demonstrated variable enhancement during different phases of the menstrual cycle, ^21-24 however, this can be minimized by imaging during the second week of the cycle. ²⁴

The recent studies by Kuhl et al ²⁰ and Kinkel et al ²² address many of the issues that have prevented the acceptance of more standard interpretation guidelines. Their work incorporates both lesion morphology and time course kinetics, which optimizes the discriminatory capability of breast MRI. Although the overall study populations have been small, it appears that the principles remain valid. The enhancement kinetics as represented by the time-signal intensity curves differ significantly among benign and malignant lesions. This shows great promise for enabling breast MRI to have both high sensitivity and specificity.

Much of the reason for the lack of consensus regarding diagnostic criteria for breast lesion evaluation has stemmed from different image acquisition protocols emphasizing either high-spatial-resolution images or dynamic breast MRI. ^{13,15,16,20-22} These two concepts involve the competing demands of adequate spatial and temporal resolution. High-spatial-resolution techniques place emphasis on morphologic features, such as margins and internal characteristics, while sacrificing some temporal resolution. ^{13,15,16,20-22} In comparison, contrast-enhanced dynamic MRI emphasizes the enhancement characteristics and involves high-temporal resolution of approximately one minute or less. ²² Since both types of information have been shown to be useful, integrating both the kinetic and morphologic information through an acceptable imaging technique seems logical. The recent work of Kinkel et al ²² provides an example of one such technique combining the morphologic and semi-dynamic data by acquiring three-high-spatial-resolution MR sequences.

Technical factors

Just as no formal interpretation guidelines exist, standardized imaging protocols have not yet been developed either. Part of the reason for this is the fact that nearly all of the early and current work has occurred primarily in a research setting. The work of Kuhl et al ²⁰ and Kinkel et al ²² combining both high-temporal and spatial resolution, is such an example. A wide range of protocols exist based on the different hardware and software available, clinical indications, desired results, clinical experience, ¹⁵ type of breast coil, magnetic field strength, and imaging parameters. ¹⁶

Despite this variability, some agreement does exist regarding general imaging techniques. It is agreed that the signal from fat should be removed because enhancing lesions may become isointense to fat after gadolinium contrast administration and therefore may be missed. ^14,16 This can be accomplished by either post-processing image subtraction or by different fat suppression techniques. ¹⁶ One pitfall to post-processing image subtraction is that there can be no patient motion between the pre-contrast and post-contrast images. ¹⁶ The rotating delivery of excitation off resonance (RODEO) sequence is an example of a fat suppression technique combined with magnetization transfer and T1 weighting in an efficient non-selective three-dimensional acquisition. ^7,12,14,25 Radiofrequency spoiling is used to depict lesions with a high fluid content. This helps to distinguish cysts from enhancing masses, which produce opposite signals. Masses are hypointense on pre-contrast and hyperintense on post-contrast RODEO MR images, while cysts are hyperintense on pre-contrast and hypointense on post-contrast images. ^12,14 With this technique, Harms et al ²⁵ demonstrated a nearly 100% negative predictive value for malignancy, with only 3 false negatives in more than 3,000 cases.

Currently, most MRI protocols use two- and three-dimensional T1-weighted gradient echo techniques for dynamic breast imaging over 6 to10 minutes after bolus injection of gadolinium-based contrast agents. A T2-weighted sequence is also performed to depict cysts or cystic lesions. One of the more recent techniques being investigated involves fast T2*-weighted perfusion imaging, which uses susceptibility-mediated perfusion imaging of the breast to distinguish between benign and malignant tumors. The individual sequence patterns and imaging protocols vary according to the manufacturer. High magnetic fields and strong gradients are recommended. Also, breast coils with compression devices help reduce motion artifacts, thus improving subtraction images. ¹⁵

Conclusion

While breast cancer remains one of the most devastating diseases in women's health, advances and experience in MRI of the breast hold great promise for allowing earlier and more accurate detection of breast cancer. In turn, it is hoped, this will provide more favorable and effective treatment op-tions. MRI of the breast has the ability to detect both clinically and mammographically occult malignancies. It also has a very high sensitivity and specificity for breast cancer detection. Perhaps just as important, it has a negative predictive value of malignancy approaching 100%.

Tremendous advances have been made in the past 10 years in breast MRI, including improved MRI units, breast coils, imaging sequences, clinical experience, and the introduction of contrast-enhanced imaging. Future directions of breast MRI are focused on developing additional imaging sequences, new contrast agents, and MRI needle localization-biopsy systems.

Acknowledgements

I wish to thank Dr. Waleed Qaisi for his valuable support in preparing this manuscript. I also would like to thank Dr. Steven Harms for allowing the generous use of images.