Breast Imaging The potential of - Weighted MRI as an unenhanced breast cancer screening tool

by Dr. N Amornsiripanitch and Dr. SC Partridge

This article provides an overview of the potential publication [8]. Summary of evidence role of diffusion-weighted magnetic resonance to date of DW-MRI in cancer detection, imaging (DW-MRI) as a breast scancer screening optimal approaches, and future consid- tool independent of dynamic contrast-enhance- erations are presented. ment. The article aims to summarize evidence to Current Evidence for DW-MRI in date of DW-MRI in cancer detection and present Breast Cancer The following equation describes optimal approaches and future considerations. DW-MRI signal intensity in rela- tion to water mobility within a : Due to well-documented limita- Considering the constraints of SD=S0e-b*ADC, where SD is defined tions of in the settings contrast-enhanced breast MRI, there as diffusion weighted signal intensity, of women with dense breasts and high- is great clinical value in identifying an S0 the signal intensity without diffusion risk women [1], there has been great unenhanced MRI modality. Diffusion- weighting, b or ‘b-value’ the diffusion interest in identifying imaging tech- weighted (DW) MRI is a technique sensitization factor, which is dependent niques to supplement mammography that does not require external contrast on applied gradient’s strength and tim- in breast cancer screening. Dynamic administration—instead, image con- ing (s/mm2), and the apparent diffusion contrast-enhanced (DCE) MRI is trast is generated from endogenous coefficient (ADC) the rate of diffusion endorsed by multinational organiza- water movement, reflecting multiple or average area occupied by a water tions as a supplemental screening tool tissue factors such as cellular mem- molecule per unit time (mm2/s) [9]. for women at high risk for breast can- brane integrity, density, and organiza- Compared to normal surround- cer [2, 3] due to high sensitivity and tion. DW-MRI has been investigated ing tissue, breast malignancies typi- cancer detection rate [4, 5]. Screening for a variety of breast imaging applica- cally exhibit impeded water diffusion, with DCE-MRI has also been shown tions, most commonly as an adjunct appearing dark on ADC map and to decrease incidental cancer rate in tool for lesion assessment in multipa- bright on DW-MRI sequences [10] women with extremely dense breast [6]. rametric breast MRI examinations and [Figure 1]. A meta-analysis of 73 stud- However, widespread implementation for evaluating response to neoadjuvant ies demonstrated that, using only ADC of DCE-MRI is limited by cost. And chemotherapy. However, a growing measurements, DW-MRI could differ- given the unknown long-term effects number of studies are evaluating the entiate benign versus malignant lesions of gadolinium retention after admin- role of DW-MRI as a stand-alone tool with comparable sensitivity and speci- istration for contrast-enhanced MRI for breast cancer detection. ficity to DCE-MRI (sensitivity=89% [7], caution may be warranted against The goal of this article is to provide vs. 93% and specificity=82% vs. 71%, repeated administrations in a healthy an overview of DW-MRI’s potential as respectively) [11]. In another study, 89% population such as women undergoing a breast cancer screening tool, which of mammographically occult cancers breast cancer screening. is outlined in greater detail in a recent were visually detected on DW-MRI,

The Authors Abbreviations 1 2,3 Nita Amornsiripanitch, MD and Savannah C Partridge, PhD ADC – apparent diffusion coefficient 1. Department of , Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, USA DCE – dynamic contrast-enhanced 2. Department of Radiology, University of Washington, Seattle, Washington, USA DCIS – ductal carcinoma in situ 3. Breast Imaging, Seattle Cancer Care Alliance, Seattle, Washington, USA DW – diffusion-weighted Corresponding Author Dr SC Partridge EPI – echo planar imaging email : [email protected] MIP – maximum intensity projection

40 DI EUROPE FEB/MARCH 2020 Figure 1. 52-year-old woman with invasive ductal carcinoma in the right breast detected on DCE-MRI (left). On DW-MRI, the lesion is not visible on the (T2-weighted) b=0 image but is hyperintense to normal breast tissue on the diffusion-sensitized b=800 s/mm2 image (arrow). The lesion shows lower diffusivity (dark) on corresponding ADC map (right, mean ADC=0.89x10-3 mm2/s) compared to normal tissue (mean ADC=2.21x10-3 mm2/s). (Reprinted, with permission, from [7].) suggesting that DW-MRI may be superior sensitivity included only mammographi- imaging modalities was reported in some to mammogram in cancer detection [12]. cally occult cancer and did not exclude of the studies [13-18]. Compared to mam- However, readers in these studies were not studies with suboptimal image quality [15], mography, one study found DW-MRI to blinded to DCE-MRI images when iden- whereas the study with the highest sensi- be more accurate and sensitive in cancer tifying cancers with DW-MRI. Therefore, tivity performed double reading [19]. A detection (area under the receiver oper- these study designs do not approximate a variety of imaging acquisition techniques ating characteristic curve=0.73 vs. 0.64 clinical setting where DW-MRI is poten- were also used. and sensitivity=69% vs. 40%) [18], and tially used as a stand-alone unenhanced Performance of DW-MRI versus other all malignancies detected by DW-MRI in modality to supplement mammogram in screening for breast cancer.

Blinded DW-MRI Reader Studies Several studies to-date have explored the use of DW-MRI in study designs simu- lating a clinical screening setting [13-19]. Readers in these studies retrospectively reviewed only DW-MRI and other unen- hanced MRI sequences without access to contrast-enhanced sequences. Readers assessed exams for level of suspicion for malignancy (assigning a numeric score or positive versus negative assessment). Study designs ranged from inclusion of only asymptomatic intermediate-to-high risk patients (cancer prevalence 1.4-2.5%) [13, 19], enriched asymptomatic cancer population (prevalence 25-67%) [15, 18], to inclusion of symptomatic and/or known cancer patients [14, 16, 17]. However, to simulate a screening experience, readers in the latter study design did not have access to clinical history and other imaging modalities, and, therefore, were not privy to prevalence of cancer in the study popula- tions (prevalence 27-46%) [14, 16, 17]. DW-MRI performances in these seven studies are summarized in Table 1. Briefly, mean sensitivity was 76% (range 45-100%) and mean specificity 90% (range 79–95%). Variation in reported sensitivities is likely due to inclusion criteria and interpreta- tion protocol: the study with the lowest Table 1: Blinded reader studies evaluating DW-MRI performance for breast cancer screening.

FEB/MARCH 2020 DI EUROPE 41 Breast Imaging GE Healthcare

SEE DIFFERENTLY SEE IN CONTRAST

Figure 2. Two examples of DCIS appearance on DW-MRI. (a) 49-year-old woman with left DCIS identified on DCE as 5.2 cm non-mass enhancement and DW-MRI at b = 1000 s/mm2 (arrows, left). (b) 60-year-old woman with DCIS detected on DCE as a 36 mm non-mass enhancement but not detectable on DW-MRI at b = 800 s/mm2. A bright susceptibility-based artifact was also present at the nipple (arrowhead). (Reprinted, with permission, from [7].)

McDonald et al were missed on mammog- missed [14, 19], which is to be expected to optimize both diagnostic specificity and raphy [15]. Compared to DCE-MRI, three given that typical DW-MRI voxel sizes are image quality (i.e., signal-to-noise ratio). studies reported lower mean sensitivity for larger than that of DCE-MRI, therefore cre- Standardization of optimal diagnostic ADC ™ DW-MRI (78.9%, range 50–94%) compared ating partial volume effect for small lesions. cut-offs is also warranted. A recent multi- Seno HD to that of DCE-MRI (93.4%, range 86-98%) Notable false positives included com- center trial suggested that an ADC cutoff Bright [13, 16, 18], suggesting that DW-MRI may plicated/proteinaceous cysts, fibroadeno- of 1.68 ×10-3 mm2/s using a maximum Contrast Enhanced Spectral Mammography not be as sensitive in detecting cancer com- mas, and artifactual “lesions” [14-17]. In b-value of 800 s/mm2 could avoid 21% of pared to DCE-MRI. None of the above stud- one study, all seven DW-MRI false positives unnecessary breast biopsies prompted by ies directly compared blinded DW-MRI were found to represent complicated cysts DCE-MRI without compromising sensitiv- performance with that of screening whole [16]. Fibroadenomas can present with a ity [24]. Specific to screening applications breast US. However, a non-blinded study wide range of ADC values; a study reported where both sensitivity and specificity are a of 60 mammographically occult cancers that 37% of fibroadenomas exhibited ADCs major consideration, there is also benefit in showed that more cancers were detectable in the same range as those of malignancies acquiring an additional very high b-value on DW-MRI than MRI-guided focused US (all below an ADC cut-off of 1.81 ×10-3 (1200 – 1500 s/mm2) as lesion conspicu- (78% vs 63% respectively) [20]. mm2/s) [22]. Examples of a complicated ity increases with b-value [8]. High diffu- Notable false negatives in these cyst and fibroadenoma on DW-MRI are sion-weighting can also be achieved with DW-MRI studies included ductal carci- shown in Figure 3. Figure 2b also demon- computed DW-MRI, a technique that uses noma in situ (DCIS), malignant lesions with strates an example of artifactual lesion at images acquired at low b-values to synthe- high ADC values, and small cancers. DCIS, the nipple from field inhomogeneity-related size ones at higher b-values. The technique despite being a minority of cancers included distortion, which could be mistaken as a has been shown to generate high image in blinded reader studies, was more likely suspicious lesion. quality and lesion conspicuity without com- missed by DW-MRI than invasive carci- promising scan times [25]. noma (mean false negative rate of 32% vs. Challenges and Future Directions DW-MRI’s performance as a screening SenoBright HD Contrast-Enhanced Spectral Mammography 15%, respectively) [15-17, 19]. Examples More studies directly evaluating tool could be further amplified by advanced technology offers anastonishingly clear way of seeing what of DCIS that were missed versus detected DW-MRI as a screening tool are warranted techniques, which are described in depth in matters in the breast. on DW-MRI are shown in Figure 2. DCIS before clinical implementation. One main a prior review article [10]. Namely, high- has generally been found to have higher consideration in future studies should be resolution acquisition techniques including It subtracts the fibroglandular tissue that can cloud your view ADC measurements compared to invasive standardization of acquisition protocol and multi-shot (e.g., readout-segmented) and and increases the signal of iodine contrast, resulting in carcinomas owing to less cellular density interpretation approach. Recently estab- reduced field-of-view echo planar imaging an unobscured, ultra-informative image that provides and common non-mass morphology [21], lished expert consensus from the European (EPI) could improve lesion conspicuity and 1 which may explain why DCIS is more dif- Society of Breast Radiology provided a produce sharper images, allowing for bet- additional reporting confidence tounite patient and clinical team . ficult to detect on DW-MRI. Mucinous car- number of recommendations towards stan- ter assessment of tumor shape and margin. cinoma, a malignancy with high ADC, had dardizing breast DW-MRI [23]. Because Post-processing techniques can improve a 100% false negative rate in one study that in vivo ADCs vary with applied b-value, image quality by reducing characterized this pathological subtype [16]. a maximum b-value of 800 s/mm2 is rec- inhomogeneity-related EPI distortions Lastly, cancers ,10mm, were more likely ommended for calculating ADC in order and spatial inaccuracies and artifacts due gehealthcare.com/SenoBright

42 DI EUROPE FEB/MARCH 2020 1. 510(k) K103485, 2. C.R.L.P.N. Jeukens “Chapter 2: Physics of Contrast-Enhanced Mammography” in: Lobbes M, Jochelson MS (ed). Contrast-Enhanced Mammography. Springer 2019. (page 30). SenoBright is a trademark of General Electric Company. Breast Imaging

2 PERIPHERAL A RTERIAL

DISEASE DAY Figure 3. Two common DW-MRI false positive lesions. (a) 53-year-old female with a proteinaceous cyst in the right breast detected on DCE-MRI (arrow) as a 10-mm round mass with circumscribed margins and thin rim enhancement and exhibiting T2 hyperintensity. On DW-MRI at b = 800 s/mm2, the cyst demonstrates restricted diffusion with corresponding low ADC value (0.48 x 10-3 mm2/s) and could be mischaracterized as a suspicious finding without access to contrast enhanced images. (b) 48-year-old woman with a fibroadenoma detected on DCE-MRI (arrow) as a 9-mm enhancing oval circumscribed mass with dark internal septations and isointense T2 signal (arrow). On DW-MRI at b = 600 s/mm2, the fibroadenoma demonstrates restricted diffusion and appears bright with corresponding low ADC value (1.38 ×10−3 mm2/s). (Reprinted, with permission, from Partridge et al. Magn Reson Imaging Clin N Am 2013;21(3):601-24 and Parsian et al. Clin Imaging, 2016;40(5):1047-54.)

to eddy-currents and motion, as well as 18(7): p. 1307-18. Mammographically Occult Breast Cancer in Elevated- 3. Saslow, D., et al., American Cancer Society guidelines Risk Women With Dense Breasts. AJR Am J Roentgenol, correcting b-value inaccuracies due to for breast screening with MRI as an adjunct to mammog- 2016. 207(1): p. 205-16. gradient nonlinearities. Display enhancing raphy. CA Cancer J Clin, 2007. 57(2): p. 75-89. 16. Telegrafo, M., et al., Unenhanced breast MRI (STIR, techniques such as maximum intensity pro- 4. Lehman, C.D., et al., Screening women at high risk for T2-weighted TSE, DWIBS): An accurate and alternative breast cancer with mammography and magnetic reso- strategy for detecting and differentiating breast lesions. jections (MIPs), which renders a 3-dimen- nance imaging. Cancer, 2005. 103(9): p. 1898-905. Magn Reson Imaging, 2015. 33(8): p. 951-5. sional display of DW-MRI, and fusion of 5. Kriege, M., et al., Efficacy of MRI and mammography 17. Trimboli, R.M., et al., Breast cancer detection DW-MR images to unenhanced T1- or for breast-cancer screening in women with a familial or using double reading of unenhanced MRI including genetic predisposition. N Engl J Med, 2004. 351(5): T1-weighted, T2-weighted STIR, and diffusion-weighted T2-weighted images may also improve can- p. 427-37. imaging: a proof of concept study. AJR Am J Roentgenol, cer detection accuracy and reduce reading 6. Bakker, M.F., et al., Supplemental MRI Screening for 2014. 203(3): p. 674-81. time [13, 16]. Women with Extremely Dense Breast Tissue. N Engl J 18. Yabuuchi, H., et al., Detection of non-palpable Med, 2019. 381(22): p. 2091-2102. breast cancer in asymptomatic women by using unen- 7. Gulani, V., et al., Gadolinium deposition in the brain: hanced diffusion-weighted and T2-weighted MR imaging: Conclusion summary of evidence and recommendations. Lancet comparison with mammography and dynamic contrast- Neurol, 2017. 16(7): p. 564-570. enhanced MR imaging. Eur Radiol, 2011. 21(1): p. Several studies to date suggest that 11-7. 8. Amornsiripanitch, N., et al., Diffusion-weighted MRI for DW-MRI may be more sensitive than Unenhanced Breast Cancer Screening. Radiology, 2019. 19. Rotili, A., et al., Double reading of diffusion-weighted mammography in breast cancer detec- 293(3): p. 504-520. magnetic resonance imaging for breast cancer detec- 35 years of excellence in IR tion. Breast Cancer Res Treat, 2020. tion and support further investigation 9. Le Bihan, D., et al., MR imaging of intravoxel incoher- ent motions: application to diffusion and perfusion in 20. Amornsiripanitch, N., et al., Visibility of mammo- into DW-MRI as an unenhanced option neurologic disorders. Radiology, 1986. 161(2): p. 401-7. graphically occult breast cancer on diffusion-weighted MRI versus ultrasound. Clin Imaging, 2017. 49: p. CATCH 10. Partridge, S.C., et al., Diffusion-weighted breast MRI: for supplemental breast cancer screening. 37-43. Eight clinical tracks on all IR topics, including one Clinical applications and emerging techniques. J Magn Moreover, DW-MRI’s potential to detect Reson Imaging, 2017. 45(2): p. 337-355. 21. Choi, S.Y., et al., Correlation of the apparent diffu- THE EARLY sion coefficiency values on diffusion-weighted imaging especially dedicated to oncologic interventions clinically and mammographically occult 11. Baxter, G.C., et al., A Meta-analysis of the Diagnostic with prognostic factors for breast cancer. Br J Radiol, Performance of Diffusion MRI for Breast Lesion BIRD FEE! cancer may be further enhanced by proto- 2012. 85(1016): p. e474-9. Characterization. Radiology, 2019. 291(3): p. 632-641. col optimization and advanced techniques. 22. Parsian, S., et al., Nonmalignant breast lesions: Browse the preliminary programme to find out more! 12. Partridge, S.C., et al., Differential diagnosis of mam- ADCs of benign and high-risk subtypes assessed Larger standardized multicenter trials are mographically and clinically occult breast lesions on as false-positive at dynamic enhanced MR imaging. diffusion-weighted MRI. J Magn Reson Imaging, 2010. warranted to confirm DW-MRI’s potential Radiology, 2012. 265(3): p. 696-706. 31(3): p. 562-70. as a safe and convenient alternative to con- 23. Baltzer, P., et al., Diffusion-weighted imaging of the 13. Kang, J.W., et al., Unenhanced magnetic resonance www.cirse.org breast-a consensus and mission statement from the ventional contrast-enhanced MRI for breast screening using fused diffusion-weighted imaging and EUSOBI International Breast Diffusion-Weighted Imaging maximum-intensity projection in patients with a personal screening. working group. Eur Radiol, 2019. history of breast cancer: role of fused DWI for postopera- tive screening. Breast Cancer Res Treat, 2017. 165(1): 24. Rahbar, H., et al., Utility of Diffusion-weighted Imaging References p. 119-128. to Decrease Unnecessary Biopsies Prompted by Breast MRI: A Trial of the ECOG-ACRIN Cancer Research Group 1. Carney PA, et al., Individual and combined effects of 14. Kazama, T., et al., Diffusion-weighted MRI as an (A6702). Clin Cancer Res, 2019. 25(6): p. 1756-1765. age, breast density, and hormone replacement therapy adjunct to mammography in women under 50 years use on the accuracy of screening mammography. Ann of age: an initial study. J Magn Reson Imaging, 2012. 25. O’Flynn, E.A., et al., Evaluating the diagnostic sen- Intern Med, 2003. 138(3): p. 168-75. 36(1): p. 139-44. sitivity of computed diffusion-weighted MR imaging in the detection of breast cancer. J Magn Reson Imaging, 2. Mann, R.M., et al., Breast MRI: guidelines from the 15. McDonald, E.S., et al., Performance of DWI as 2016. 44(1): p. 130-7. European Society of Breast Imaging. Eur Radiol, 2008. a Rapid Unenhanced Technique for Detecting Cardiovascular and Interventional Radiological Society of Europe FEB/MARCH 2020 DI EUROPE 45