Focused research efforts over the last decade have helped establish breast magnetic resonance imaging (MRI) as an important tool for the detection and characterization of breast cancer. The value of breast MR imaging is derived primarily from the high sensitivity of contrast material enhancement in the detection of breast cancer [1]. However, though sensitivity is consistently high across different clinical studies, specificity still remains a challenge [2, 3]. The current consensus is that MR is an excellent tool for determining tumor extent, and should especially be performed in dense breasts. However, due to low specificity, it becomes necessary to perform biopsy of additional lesions detected on MR. The results of multiple studies show that preoperative MR is recommended in patients with breast cancer who are scheduled for breast-conserving surgery. Breast MRI offers a multi-contrast approach which, combined with morphological features, may provide the optimal approach to diagnosis and management [4]. A brief overview of current breast MRI techniques is presented below including dynamic contrast enhanced MRI, diffusion MRI, spectroscopy, perfusion MRI, MR elastography, concluding with a computer aided assessment of breast lesions.
by Prof. Usha Sinha & Prof. Shantanu Sinha

Dynamic Contrast Enhanced MR Imaging (DCE-MRI)
This is the most routinely used MR imaging technique for the assessment of breast lesions. It has shown the most promise for discriminating between malignant and benign tumors; however even with this technique, specificity is in the range of 37%-97% [5]. In DCE-MRI, a dynamic imaging is performed after injection of a contrast agent and the signal intensity on a T1 weighted dynamic sequence has specific contrast uptake patterns for malignant and benign, with malignant lesions showing a higher ‘signal enhancement ratio’, larger values of ‘maximum slope’, and shorter ‘time to peak of enhancement’ [Figure 1]. Multivariate models combining morphology and contrast uptake dynamics have the highest predictive values and exceed that based on lesion architectures or intensity patterns alone [4]. However, the conflicting requirements of high spatial resolution and high temporal resolution have precluded the routine use of both feature types in clinical studies; recent studies explore optimum temporal resolution [6] and pulse sequences for combined high spatial and temporal resolution [7].

The applications of DCE-MRI extend beyond diagnosis to therapeutic monitoring and for breast screening. The non-invasive nature of MRI makes it an ideal candidate for monitoring and increases the potential for detection of early response [8] and for treatment optimization [9]. The American Cancer Society (ACS) has established guidelines for screening with breast MRI as an adjunct to mammography [10]. For breast MRI, the guidelines indicate that screening can potentially identify cancer in patients of specific risk groups, e.g. high-risk patients facing a lifetime risk of ~20–25%.

Diffusion Imaging
Diffusion weighted MR imaging (DW-MRI) provides unique information about the state of the molecular translational motion of water. This allows inference about local tissue architecture which is a sensitive early indicator of abnormality and cellularity. The mean or average diffusivity in tissue is quantified by an index called the Apparent Diffusion Coefficient (ADC). It is anticipated that DW-MRI will detect early changes in morphology and physiology of tissues associated with changes in water content such as changes in the permeability of cell membranes, cell swelling and/or cell lysis.
Early studies using DW-MRI to characterize breast lesions found that the ADC values of malignant breast tumors were significantly reduced compared to benign and normal fibro-glandular tissue, presumably from increased cellularity in malignancy [11]. DW-MRI evaluation of 60 women showed a specificity of 81% and sensitivity of 80% to discriminate malignant from benign lesions [12]. Guo et al [13] showed that tumor cellularity correlated inversely with tumor ADC and that malignant breast tumors had a higher cellularity and a lower ADC than benign breast tumors. A recent study confirmed that DWI shows potential for improving the positive predictive value of breast MRI for lesions of varied types and sizes [14]. Diffusion MRI has also been shown to detect early response to neoadjuvant therapy prior to tumor size changes [8, 15] including voxel by voxel ADC changes [Figure 2] [16] and more recently, to identify patients most likely to respond to treatment [17].

Spectroscopy
MR spectroscopy (MRS) is a biochemical marker that allows the non-invasive detection of proton metabolites. Cancerous lesions demonstrate elevated composite choline levels arising from increased cellular proliferation. Several groups have demonstrated that MRS either alone or in conjunction with MRI improves specificity of breast MR [18]. A multi-institutional study by Kratz-Bull et al. confirmed the robustness of total choline-containing (tCho) compounds from MRS and the improvements in specificity in characterizing breast lesions [19]. The consensus is that MRS is useful in larger tumors (>1 cm); however it may be limited in its ability to discriminate benign breast lesions from phyllodes tumors of benign and borderline malignancy [20]. MR spectroscopic imaging has recently produced the highest spectroscopic resolution (0.25 cm3) and offers the potential for mapping choline
distributions across the lesion [21].

Perfusion MR imaging of the breast
Perfusion imaging provides information on microvascular distribution and density and is performed by tracking the changes in the apparent relaxation rate T2* during the first passage of a contrast bolus. T2* refers to the measured loss of transverse magnetization in a gradient-echo sequence that is due to the combined effects of the inherent spin-spin relaxation time T2 and macroscopic magnetic field homogeneities (static field and paramagnetic contrast agent). The main limitation of including perfusion imaging in a MR breast protocol is the need for two contrast boluses: one for the first pass bolus tracking to monitor susceptibility induced losses and the other to monitor the contrast uptake. An early study evaluated perfusion imaging versus DCE-MR [22] and found a strong intensity loss in malignant breast tumors. This is in marked contrast to fibroadenomas that showed little or no perfusion effects, whereas in DCE-MRI the enhancement patterns of malignant lesions and fibroadenomas overlapped. Recent studies on larger population cohorts concluded that perfusion imaging increases the level of specificity, even up to 100% in one study [23, 24].

Magnetic Resonance Elastography (MRE):
MRE is an elegant, relatively new approach and measures the elastic properties of tissue. It is a phase-contrast-based MRI imaging technique that can directly visualize and quantitatively measure propagating acoustic strain waves in tissue-like materials subjected to harmonic mechanical excitation. Quantitative values of elasticity, e.g. shear modulus can be calculated from the acquired data [Figure 1]. There have only been a few preliminary studies using this technique; initial studies have indeed confirmed that MRE could be a non-invasive ‘palpation’ in that breast tumors revealed higher shear elasticity than normal breast tissue [25]. Xydeas et al. have applied MRE to discriminate between malignant and benign tumors [26]. The increased diagnostic accuracy of a combined MRE and DCE-MR in a recent study confirms the potential of MRE [27]. MRE technology is not, as yet, mature for clinical use but the initial results are quite promising.

Computer aided analysis for Breast MRI
Automated breast lesion segmentation methods have been explored by several groups and are the first step in computer aided analysis. Proposed segmentation methods include fuzzy c-means clustering [28], neural network classifiers [29], and multi-parametric techniques [30]. Fuzzy c-means clustering refined by level sets has been proposed recently for segmenting lesions from dynamic scans and shows potential for accurate tracking of lesion volume changes in treatment [31]. Lesion features have been extracted from both 2D as well as from the entire 3D
segmentation. Several measures of lesion shape and lesion border characteristics have been evaluated as potential indices of the lesion structure [32)]. A recent study indicates that texture analysis of breast MRI can differentiate cancer from normal tissue with the potential to distinguish between lobular and ductal cancer, thus providing accurate computer-assisted characterization of breast lesions [33].

Several studies have shown that computer-assisted breast MR can be useful either alone or in conjuction with a radiologist [34]. A few groups have also worked on automated segmentation, rating and classification of breast tumors [35], combining automatically extracted morphological and dynamic features, and visualization of the data [36]. A promising new report is on extending computer-aided diagnosis of breast DCE MR lesions from discriminating to prognostic tasks [37].
 
Conclusions
Breast MRI is an area of active research and offers significant potential for lesion characterization: contrast uptake with DCE-MRI, morphology with high resolution structural imaging, cellularity with diffusion imaging, vascular networks through perfusion MR imaging, and biochemical composition with MR spectroscopy. This battery of techniques allows a multi-pronged approach that can be directed toward diagnosis and therapy monitoring. Challenges still remain and further improvement in acquisition methods will be required to enable a comprehensive multi-parametric examination of the bilateral breast in clinically relevant times [38]. Inclusion of the indices extracted from the multi-parametric imaging in a robust classifier framework will ultimately help achieve a higher specificity for breast MR imaging; establishing it as a one-stop examination for the screening, diagnosis, prognosis, longitudinal monitoring and management of patients with breast lesions.

References
Readers are referred to Sinha et al for a more detailed review of the information presented in the above article[3].

1. Rotaru N & Luciani A. J BUON 2004; 9: 77-82.
2. Morris EA. Magn Reson Imaging Clin N Am 2010; 18: 57-74.
3. Sinha S, Sinha U. Recent advances in breast MRI and MRS. NMR Biomed 2009; 22: 3-16.
4. Schnall MD et al. Radiology 2006; 238: 42-53.
5. Buadu LD et al. Radiology 1996; 200: 639–649.
6. El Khouli RH et al. J Magn Reson Imaging 2009; 30: 999-1004.
7. Medved M et al. Magn Reson Imaging 2010; 28:16-21.
8. Pickles MD et al. Breast Cancer Res Treat 2005; 91: 1-10.
9. Martincich L et al. Breast Cancer Res Treat 2004; 83: 67-76.
10. Saslow D et al. Cancer J Clin 2007; 57: 75–89.
11. Sinha S et al. J Magn Reson Imaging 2002; 15: 693–704.
12. Marini C et al. Eur Radiol. 2007; (Epub ahead of print).
13. Guo Y et al. J Magn Reson Imaging 2002; 16: 172-178.
14. Partridge SC et al. Am J Roentgenol. 2009; 193:1716-1722.
15. Theilmann RJ et al. Neoplasia 2004; 6: 831-837.
16. Ma B et al. Inf Process Med Imaging 2009; 21:276-287.
17. Iacconi C et al. Eur Radiol 2010; 20:303-308.
18. Meisamy S et al. Radiology 2005; 236: 465-475.
19. Katz-Brull R et al. J Natl Cancer Inst 2002; 94: 1197-1203.
20. Tse GM et al. Am J Roentgenol 2003; 181: 1267-1272.
21. Hu J et al.  Med Phys 2009; 36: 4870-4877.
22. Kuhl CK et al. Radiology 1997; 202: 87–95.
23. Kvistad KA et al. Radiology 2000; 216: 545–553.
24. Huang W, et al. Radiology 2004; 232: 585–591.
25. Lorenzen J et al. Rofo 2002; 174: 830-834.
26. Xydeas T et al. Invest Radiol 2005; 40: 412-420.
27. Siegmann KC et al. Eur Radiol 2010 ; 20: 318-325.
28. Chen Wet al. Acad Radiol 2006; 13: 63–72.
29. Woods BJ et al. J Magn Reson Imaging. 2007; 25: 495-501.
30. Jacobs MA et al. Radiology 2003; 229: 225–232.
31. Shi J et al. Med Phys 2009; 36: 5052-5063.
32. Liney GP et al. J Magn Reson Imaging 2006; 23: 493–498.
33. Holli K et al. Acad Radiol 2010; 17:135-141.
34. Chen W et al. Med Phys 2004; 31: 1076–1082.
35. Gilhuijs KG et al. Radiology 2002; 225: 907–916.
36. Subramanian KR et al. Comput Med Imaging Graph 2004; 28: 435–444.
37. Bhooshan N et al. Radiology 2010 Feb 1. [Epub ahead of print]
38. Han M et al. Magn Reson Med 2009; 62:1221-1231.


The authors:
Usha Sinha, Ph.D.
Professor and Chairperson,
 Department of Physics,
San Diego State University,
5500 Campanile Dr.
San Diego, CA-92182-1233.
Ph.No. (619)-594-1791
Cell: (310)-460-8897
E-mail: usinha@sciences.sdsu.edu

Author for Correspondence:
Shantanu Sinha, Ph.D.
Professor, Department of Radiology,
UCSD School of Medicine,
3510 Dunhill Street
San Diego, CA-92121-0852.
Ph.No. (858)-534-2004
Cell: (310)-435-3994
E-mail: shsinha@ucsd.edu