Diffusion MRI as a Stand-Alone Unenhanced Approach for Breast Imaging and Screening

List of Abbreviations

Although dynamic contrast-enhanced magnetic resonance imaging (DCE MRI) is highly sensitive and endorsed by multinational organizations as a supplemental screening tool for high-risk women, widespread implementation of DCE MRI is limited by high cost and uncertain long-term effects of gadolinium retention from contrast administration. In addition, the cost-effectiveness of DCE MRI in intermediate risk patients, such as those with history of breast cancer and dense breasts, remains unclear. Therefore there is great interest in identifying an affordable, unenhanced imaging modality suitable for breast cancer screening. Diffusion-weighted (DW) MRI has emerged as one of the leading options, owing to its short scan time, relative availability and promising sensitivity for identifying breast cancer. DW MRI enables detection of breast malignancies without the need for administering a contrast agent, based instead on microstructural characteristics (e.g., cellular density), as reflected by endogenous diffusional water movement ( Fig. 6.1 ). To date, most experimental and clinical uses of DW MRI have been as an adjunct to DCE MRI in lesion assessment, for preoperative staging of ipsilateral and contralateral breasts, and for evaluating the response to neoadjuvant chemotherapy. However, there is increasing interest in exploring the use of DW MRI as a stand-alone tool for breast cancer detection.

This chapter summarizes the evidence for DW MRI in cancer detection and describes the optimal unenhanced breast cancer screening methods. In addition, the chapter discusses ongoing multicenter DW MRI screening trials and issues associated with clinical implementation.

Current Evidence for DW MRI as a Stand-Alone Modality

Real-world performance of DW MRI for noncontrast cancer detection in the clinical screening setting has been investigated in a variety of reader studies, most of which were performed retrospectively. The readers in these studies assessed only unenhanced MRI sequences (i.e., DW MRI with or without anatomical nonenhanced T1- or T2-weighted sequences) for suspicious findings and were blinded to DCE MRI. They assigned either a binary category (suspicious vs. benign/negative) or a number on a scale corresponding to the level of suspicion, similar to the Breast Imaging Reporting and Data System (BI-RADS) categories. Study designs ranged from inclusion of only asymptomatic intermediate- to high-risk patients, patients with suspicious imaging or clinical symptoms, to those with known malignancy; some studies included a combination of more than one of the above.

DW MRI cancer detection performance across these various studies is summarized in Table 6.1 . The mean sensitivity was 81% (range 44%–97%), and the mean specificity was 88% (range 73%–96%). However, among studies that simulated clinical screening experience by including negative/benign cases, mean sensitivity was 76% (range 45%–100%), and the mean specificity was 89% (range 79%–95%). Variation in the reported sensitivities is likely due to the inclusion criteria, imaging, and interpretation protocol. The study with the lowest sensitivity included only mammographically occult cancer and used relatively low maximum b values (600–800 s/mm ² ), whereas some studies with higher sensitivities evaluated only (previously biopsied) known malignancy, used advanced imaging acquisition techniques, or performed double reading.

Some previous studies have reported on the performance of DW MRI versus other imaging modalities, including mammography, DCE MRI, abbreviated breast MRI, and ultrasound. Compared with mammography, DW MRI was found to be more sensitive (mean sensitivity across studies 78% vs. 59% for DW MRI vs. mammogram, respectively). Compared with DCE MRI, DW MRI was found to be less sensitive (mean sensitivity across studies 81% vs. 95% for DW MRI vs. DCE MRI, respectively). No studies directly compared blinded DW MRI performance with that of screening whole-breast ultrasonography; however, a nonblinded study of 60 mammographically occult cancers showed that more cancers were detectable on DW MRI (78%) compared with MRI-guided focused ultrasound (63%). In another study of 1146 women with newly diagnosed breast cancer, DW MRI of the contralateral breast showed higher sensitivity than mammography (77% and 30%, respectively) or combined mammography and ultrasound (40%) in detecting clinically occult cancer. The cancer detection rate (20 per 1000 examinations) and positive predictive value (42%) for biopsy recommendation of DW MRI was also higher compared with combined mammography and ultrasound (10 per 1000 examinations and 19%, respectively; Fig. 6.2 ) .

Common false-negative lesions in DW MRI include ductal carcinoma in situ (DCIS), mucinous carcinomas, and cancers presenting as nonmass enhancement and small masses. DCIS was more likely to be missed by DW MRI than invasive ductal carcinoma. DCIS commonly presents as a nonmass enhancement on DCE MRI with higher apparent diffusion coefficient (ADC) than invasive carcinomas, making it difficult to detect ( Fig. 6.3 ), with false-negative rates reported as high as 86% using this technique. Tumors with high liquid content, such as mucinous cancers and necrotic triple-negative cancers, can also exhibit a high ADC. Mucinous carcinoma was frequently missed on DW MRI ( Fig. 6.4 ), with a false-negative rate as high as 100%. Finally, small cancers (<10–12 mm) and invasive lobular cancer were also frequently missed ( Fig. 6.5 ). This was due to the low spatial resolution of conventional DW MRI techniques, which may lead to partial volume averaging, producing results that are not significantly better than the standard in-plane resolution and slice thickness (commonly 2 × 2 mm ² and 3–5 mm, respectively). Expected false negative lesions on DW MRI also could include tumors with a low water content (e.g., low cellularity cancers with extensive desmoplastic stromal fibrosis).

The notable false positives using DW MRI included complicated/proteinaceous cysts, fibroadenomas, and artifactual “lesions.” Complicated/proteinaceous cysts, which are known to exhibit restricted diffusion ( Fig. 6.6 ), represent the majority of DW MRI false positives in some settings. Similarly, fibroadenomas may be mistaken as a suspicious finding due to the wide range of possible ADC values. More than a third of fibroadenomas have ADCs in the same range as those of malignancies. Finally, false positives can be produced by artifactual signal, such as near the nipple-areolar complex, an area known to be prone to susceptibility-based distortion in DW MRI.

Technical Requirements as an Unenhanced Screening Modality

Expert consensus from the European Society of Breast Imaging (EUSOBI) recommended standardized parameters for high-quality breast DW MRI, which were extended specifically for the application of unenhanced breast cancer screening in a DW trial protocol ( Table 6.2 ).

Given that b values are known to directly affect the image’s signal-to-noise ratio, lesion contrast-to-noise ratio, and ADC values, selecting an optimal b value may improve cancer detection. As the maximum b value increases, the contrast-to-noise ratio and the differences in signal decay between cancer and normal/benign tissues increase, improving cancer visibility and specificity. On the other hand, as b value increases, the signal-to-noise ratio decreases. Furthermore, image acquisition at high b values is lengthy and prone to distortions due to increased susceptibility and eddy currents. Based on the data and expert consensus, an optimal maximum b value of 800 s/mm ² is recommended for generalizable quantitative ADC mapping. In a screening setting, evidence suggests using a higher b value of 1200 to 1500 s/mm ² to maximize lesion contrast and visibility, although this may result in longer acquisition times. Computed DW MRI, a technique that synthesizes higher b value images from images acquired at lower b values, could provide higher image quality and lesion conspicuity compared with those acquired directly ( Fig. 6.7 ). However, one should remain cautious that extrapolation from lower b value images by assuming diffusion is monoexponential (Gaussian) may result in interpretation errors of some findings (see Chapter 1 ).

DW MRI’s performance as a screening tool could be further enhanced by advanced techniques. High-resolution acquisition techniques including multishot (e.g., readout-segmented and simultaneous multislice read-out segmented) and reduced field-of-view echo-planar imaging (EPI) could improve lesion conspicuity and produce sharper images, which would enable better assessment of tumor shape and margin. Image registration algorithms can reduce magnetic field inhomogeneity-related EPI distortions as well as spatial inaccuracies and artifacts caused by eddy currents and motion. Postprocessing techniques can also correct b value inaccuracies due to gradient nonlinearities. Display-enhancing techniques can also be used to improve cancer detection accuracy and reduce reading time. These methods include maximum intensity projections (MIPs), which render a 3D display of DW MRI, and the fusion of DW images to nonenhanced anatomical T1- or T2-weighted images.

Image Interpretation Strategies

Standardized terminology is needed to describe unenhanced DW MRI findings, interpretation, and management recommendations, similar to the standardized classification for other breast imaging modalities. A variety of interpretation strategies for noncontrast breast screening with DW MRI have been implemented in prior reader studies, with a common approach to first identify unique areas of signal hyperintensity on diffusion-weighted images but varying reliance on quantitative ADC assessment, morphology, and appearance on other unenhanced sequences (T1 and T2 weighted). Based on the prior work, a standardized lexicon and interpretation criteria for DW MRI and precontrast T1- and T2-weighted sequences ( Tables 6.3 and 6.4 ) was proposed by the Diffusion-Weighted MRI Screening Trial (DWIST) group in 2019, which has been used in over 1200 patients in the stand-alone DW MRI program at Seoul National University Hospital since January 2020. The approach is summarized in Fig. 6.8 , and the interpretation algorithms for unique findings on DW MRI are outlined in Fig. 6.9 . In this guideline, an ADC cutoff value of 1.3 × 10 ⁻³ mm ² /s was determined based on a diffusion level lexicon from EUSOBI guidelines in order to decrease false positives and increase cancer detection.