Breast Diffusion-weighted MR Imaging: Current Applications, Insights from Screening, and Future Directions

Abstract

Breast diffusion weighted MR imaging (DWI) is increasingly used, because it is fast and easy to be added in clinical protocol without contrast agent and provides information of cellularity or tissue microstructure. This review article explores the principles of breast DWI, the standardization of acquisition techniques, and its current clinical applications. We emphasize its role in differentiating benign from malignant lesions, reducing unnecessary biopsies, and discuss the evidence supporting DWI as a potential standalone screening tool. Prognostic indicators derived from DWI parameters and its utility in monitoring treatment responses are discussed. Finally, we look to the future, discussing emerging techniques. This review provides a comprehensive overview of breast DWI’s current status and future potential.

Introduction

Diffusion-weighted imaging (DWI) of the breast has emerged as a powerful tool in breast cancer diagnosis and management. It visualizes and quantifies random movement of water molecules in biological tissue and reflects cell density and microstructure.¹ Shorter image acquisition time, easy availability on commercial scanners, and no need for contrast agent are major advantages of breast DWI. The issues of poor image quality and lack of standardization of breast DWI have been the obstacles to routine use. However, over the past decade, advances in MRI technology and multicenter collaboration are addressing the issues.² Thus, the clinical applications of DWI are expanding from the supplemental role of conventional dynamic contrast enhanced (DCE) MRI or breast cancer characterization to standalone screening modality for breast cancer.

This article aims to summarize the most recent studies on applications and standardization of breast DWI and explores the future.

Principles of DWI

DWI is performed by applying 2 motion-sensitizing gradients. Fixed water proton acquires phase shift during the first gradient, which is cancelled out by second gradient leading to no additional signal loss. If a water molecule diffuses to a different location, refocusing is imperfect, leading to signal loss. So, signal intensity (SI) at DWI is decreased proportional to the water diffusivity as follows:

$$\mathrm{S}\left(\mathrm{b}\right)={\mathrm{S}}_0{e}^{-\mathrm{b}\ast \mathrm{A}\mathrm{D}\mathrm{C}}$$ (1)

where S(b) is the SI with diffusion weighting of b (diffusion sensitizing factor, sec/mm²), S₀ is the SI without diffusion weighting. ADC is the apparent diffusion coefficient (mm²/sec).¹ Because the signal decay shows relatively monoexponential within b-value ranges used in clinical practice, ADC can be calculated from each voxel with 2 b-values, as follows:

$$\mathrm{A}\mathrm{D}\mathrm{C}=\mathrm{l}\mathrm{n}\left({\mathrm{S}}_1/{\mathrm{S}}_2\right)/\left({\mathrm{b}}_2-{\mathrm{b}}_1\right)$$ (2)

where b₂ >b₁, S₁ is SI at b₁, and S₂ is SI at b₂.

Malignant tumors show high SI on DWI and low SI on ADC maps due to reduced water diffusion (Fig. 1).

Fig. 1.

MR images of a 52 year-old woman with a fibroadenoma in the right breast and invasive ductal carcinoma in the left breast. (a) Contrast enhanced T1WI obtained 90 seconds after contrast injection shows oval enhancing masses in both breasts. (b) Axial DWI with b = 0 sec/mm² shows a mass with high signal intensity (arrow) in the right breast and a mass with iso-signal intensity (double arrows) in the left breast. (c) On DWI with b = 800 sec/mm², both masses show high signal intensity. (d) On the ADC map calculated using the b values of 0 sec/mm² and 800 sec/mm², the mass in the right breast shows a high signal intensity (arrow) and the mean ADC value was 2.40 × 10⁻³ mm²/sec. The mass in the left breast shows a dark signal intensity (double arrows) and the mean ADC value was 0.96 × 10⁻³ mm²/sec. ADC map is useful to distinguish true diffusion restriction and T2 shine-through effect. The benign mass in the right breast shows high signal intensity at DWI with b = 0 sec/mm² because of T2-weighting effect. ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging; T1WI, T1-weighted image.

Standardized Techniques of Breast DWI

International breast working group of the European Society of Breast Imaging has addressed a consensus statement for the minimum requirements of acquisition parameters to improve the comparison of ADC values between studies.³ Field strength of 1.5T or higher using a dedicated breast coil with at least 4 channels, echo-planar imaging (EPI)-based axial images, a minimum in-plane resolution of 2 × 2 mm², a section thickness of 4 mm or less, and parallel imaging with an acceleration factor of 2 are recommended.

Along with these technical parameters, achieving effective fat suppression is also crucial. Effective fat suppression is essential in breast DWI, as signals from voxels containing fat can artificially lower ADC values. The spectrally adiabatic inversion recovery method is preferred over short-tau inversion recovery (STIR), as STIR fat suppression may lead to over- or underestimation depending on the tissue’s T1 and ADC characteristics.⁴

Regarding the b-value (diffusion sensitizing factor, sec/mm²) selection, as the b-value increases, difference of signal decay rates between tumor and normal parenchyma increases, which leads to increased contrast-to-noise ratio and improved cancer visibility, but decreased SNR ratio and increased distortion due to susceptibility effects and eddy currents. In addition, at higher b-values (>1000 sec/mm²), the signal attenuation rate is reduced, reflecting not only cell density but also tissue microstructure, which is called non-Gaussian diffusion. Thus, considering these variables, a high b value of 800 sec/mm² and a low b value of 0–50 sec/mm² are recommended for measurement of standardized ADC by the working group.³

However, for much higher quality image is required for standalone screening purpose. The Diffusion Weighted Magnetic Resonance Imaging Screening Trial (DWIST) group has suggested acquisition of a very high b value of 1200 sec/mm² to increase lesion visibility and specificity of lesion detection in spite of lower SNRs.^5–7 Thus, images with 3 b-values including 0 sec/mm², 800 sec/mm² for ADC map generation and 1200 sec/mm² for lesion visualization are used for the DWIST trials. Three b-value images also show advantage in the lesion characterization due to the different signal decay pattern between benign and malignant masses (Figs. 2 and 3). In addition, field strength of 3.0T, dedicated breast coil with at least 16 channels, a minimum in-plane resolution of 1.3 × 1.3 mm² and a section thickness of 3 mm or less are required for standalone DWI screening.^5,6

Fig. 2.

MR images of a 36 year-old woman with a fibroadenoma. (a) Contrast enhanced T1WI obtained 90 seconds after contrast injection shows an oval enhancing mass (arrow) in the right breast. (b) Axial DWI with b = 0 sec/mm² shows a mass with high signal intensity (arrow). (c) DWI with b = 800 sec/mm² shows the mass with slightly high signal intensity (arrow). (d) DWI with b = 1200 sec/mm² shows that the mass is not visualized (arrow). As the b value increases, the signal intensity of background parenchyma and the benign mass decreases. (e) On the ADC map calculated using the b values of 0 sec/mm² and 800 sec/mm², the mass shows high signal intensity (arrow). The ADC value was 1.58 × 10⁻³ mm²/sec. ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging; T1WI, T1-weighted image.

Fig. 3.

MR images of a 45 year-old woman with an invasive ductal carcinoma. (a) Contrast enhanced T1-WI obtained 90 seconds after contrast injection shows an irregular enhancing mass (arrow) in the left breast. (b) Axial DWI with b = 0 sec/mm² shows a mass with iso-signal intensity (arrow) in the left breast. (c) DWI with b = 800 sec/mm² shows a mass with high signal intensity (arrow). (d) DWI with b = 1200 sec/mm² shows a mass with increased high signal intensity (arrow). As the b value increases, the signal intensity of background parenchyma decreases and the cancer mass increases. (e) On the ADC map calculated using the b values of 0 sec/mm² and 800 sec/mm², the mass in the left breast shows a low signal intensity (arrow). The ADC value was 0.92 × 10⁻³ mm²/sec. ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging; T1WI, T1-weighted image.

Differentiation of Benign and Malignant Breast Lesions

Recent meta-analysis including 28 studies with 4406 lesions, published from January 1985 to September 2023, the pooled sensitivity, specificity, and the areas under the receiver operating characteristic curve (AUC) of DWI were 86.5% (95% confidence interval [CI]: 81.4, 90.4), 83.5% (95% CI: 76.9, 88.6), and 0.93 (95% CI: 0.91, 0.95).⁸ In a direct head-to-head comparison with contrast-enhanced (CE) MRI including 18 studies, CE MRI showed higher sensitivity than DWI (95.1% [95% CI: 92.9, 96.7] vs. 88.9% [95% CI: 82.4, 93.1], P = 0.004) and similar specificity to DWI (82.2% [95% CI: 75.0, 87.7] vs. 82.0% [95% CI: 74.8, 87.6], P = 0.97).⁸ Authors noted that there was significant between-study heterogeneity in study design, acquisition techniques, and patient cohorts. Sensitivity was highest in studies using DWI for further characterization of suspicious lesions at other imaging modalities and was lowest at a screening setting.⁸

Earlier interpretation of DWI is mainly based on quantification of ADC due to the lower spatial resolution and artifacts.³ Lesion detection is primarily at the CE MRI. When an enhancing lesion is identified at CE MRI, cross-correlation of high b-value image and ADC maps should be performed to exclude T2 shine-through effect (Fig. 1). T2 shine-through effect means that high SI on DWI not due to diffusion restriction but due to T2-weighting effect. If a lesion shows bright SI on high b-value image and dark SI on ADC map generated from b0 sec/mm² to 800 sec/mm² images, the ADC value is further measured for characterization.^1,3 ROI using at least 3 pixels is placed on the darkest portion of the lesion on ADC map and completely within the bright SI on high b image to avoid necrosis, fat, or artifact. Then, the mean ADC value within the ROI is reported.

Based on the ADC value, the lesion is classified as very low (range of ADC, ≤0.9 × 10⁻³ mm²/sec); low (range of ADC, 0.9–1.3 × 10⁻³ mm²/sec); intermediate (range of ADC, 1.3–1.7 × 10⁻³ mm²/sec); high (range of ADC, 1.7–2.1 × 10⁻³ mm²/sec) and very high (range of ADC, >2.1 × 10⁻³ mm²/sec).³ Invasive cancers tend to show low to very low ADC values and benign lesions or normal tissue tend to show high to very high ADC values. However, specific cutoff values to differentiate benign and malignant lesions were not suggested by the working group, because there are wide variations from 1.1 × 10⁻³ to 1.6 × 10⁻³ mm²/sec and it depends on protocol variations including b-values.³ Higher cutoff value tends to increase sensitivity and lower cutoff value tends to increase specificity.⁷ Thus, they have suggested that lesion differentiation should be based on the diffusion level combined with anatomical information from other images.³

For standalone screening, the DWIST group has suggested standardized interpretation guideline based on both morphology assessment and the ADC measurement.^6,7 When a unique finding showing high-signal on high b-images is identified, it is classified as a focus, mass, and nonmass based on morphology.^6,7 Similar to Breast Imaging Reporting and Data System (BI-RADS) lexicon, the shape and internal pattern for mass or focus, or the distribution and internal pattern for nonmass are further classified. Then, the ADC value on ADC map generated from b0 sec/mm² and 800 sec/mm² images is measured.⁷ Briefly summarizing the interpretation algorithm, if a lesion shows a suspicious morphology, a biopsy (ADC ≤1.3 × 10⁻³ mm²/sec) or a 6 month follow-up (ADC >1.3 × 10⁻³ mm²/sec) is recommended depending on the ADC value (Fig. 4). If a lesion shows a non-suspicious morphology and an ADC ≤1.3 × 10⁻³ mm²/sec, a biopsy (low to iso SI on b0) or a 6 month follow-up (high SI on b0) is recommended depending on the SI on b0. If a lesion shows a non-suspicious morphology and an ADC >1.3 × 10⁻³ mm²/sec, a 6 month (low SI) or 1 year follow-up (high SI) is recommended depending on the SI on b0.⁷ However, since the lesions detected in the screening setting are relatively small and subtle, the morphology evaluation and ADC measurement at standalone screening DWI are sometimes challenging due to the lower spatial resolution of DWI compared with other images (Figs. 5 and 6).

Fig. 4.

Interpretation flow chart modified from the DWIST for unique findings on standalone DWI. DWI, diffusion weighted imaging; DWIST, Diffusion Weighted Magnetic Resonance Imaging Screening Trial.

Fig. 5.

Images of a 59 year-old woman with an invasive ductal carcinoma detected by standalone screening DWI. She underwent contralateral mastectomy due to invasive ductal carcinoma 26 months ago. (a) Axial DWI with b = 0 sec/mm² shows a lesion with slightly high-signal intensity (arrow) in the left breast. (b) DWI with b = 800 sec/mm² shows an oval mass with high signal intensity (arrow). (c) DWI with b = 1200 sec/mm² shows an oval mass with high signal intensity (arrow). (d) On the ADC map calculated using the b values of 0 sec/mm² and 800 sec/mm², the mass shows a low signal intensity (arrow). The ADC value was 0.89 × 10⁻³ mm²/sec. (e) Second-look ultrasonography with Doppler study shows an irregular mass with internal increased vascularity. (f) Spot compression magnification view shows a mass with an indistinct margin and internal calcifications. Surgical histopathology revealed a 1.1 cm invasive ductal carcinoma. ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging.

Fig. 6.

Images of a 43 year-old woman with false-negative finding at standalone screening DWI. She underwent contralateral skin sparing total mastectomy due to invasive ductal carcinoma and reconstruction 5 years ago. (a) Axial DWI with b = 0 sec/mm² shows a linear nonmass with high signal intensity (arrow) in the right breast. (b) DWI with b = 800 sec/mm² shows a linear nonmass with high signal intensity (arrow). (c) DWI with b = 1200 sec/mm² shows a linear nonmass with high signal intensity (arrow). (d) On the ADC map calculated using the b values of 0 sec/mm² and 800 sec/mm², ADC measurement is not available due to small lesion size. It was initially interpreted as negative finding. (e) Eight months later, screening mammography shows grouped pleomorphic calcifications in the corresponding area of right breast. Surgical histopathology revealed a 1.6 cm ductal carcinoma in situ with 0.2 cm invasive ductal carcinoma component. ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging.

Reduction of Unnecessary Biopsies

CE MRI is the most sensitive modality for breast cancer detection, but it shows various specificities. As high false-positive findings of CE MRI have limited the applicability of breast MRI, supplemental role of DWI to improve specificity has been explored. According to previous meta-analyses, the pooled specificity of DWI is 75.6% to 85%, better than of CE MRI’s pooled specificity of 71%,^9–11 which differs from the most recent meta-analysis report that DWI shows similar specificity (82.0% and 82.2%, P = 0.97) to regular CE MRI.⁸ Thus, there have been multiple studies to determine whether adding DWI to CE MRI can improve the performance by reducing false positives of CE MRI and what is the appropriate ADC value for this.

The Eastern Cooperative Oncology Group (ECOG) –American College of Radiology Imaging Network (ACRIN) Cancer Research Group A6702 trial was designed to confirm ADC differences between malignant and benign lesions and to identify a generalizable ADC threshold. They found that the optimal cutoff of ADC value was 1.53 × 10⁻³ mm²/sec.¹² And this cutoff could have avoided 21% of benign biopsies without missing cancers. Subsequent retrospective study also validated that ADC cutoff of 1.50 × 10⁻³ mm²/sec could have allowed downgrading of lesions classified as BI-RADS 4 and 33% of unnecessary biopsies could have been avoided in a large heterogeneous multicenter data set.¹³

Recent study further validated the study results in a clinical setting.¹⁴ Of the 240 BI-RADS category 4 or 5 lesions, biopsy could have been avoided by 15.8% (38 of 240) when A6702 ADC cutoff value (1.53 × 10⁻³ mm²/sec) was applied. Also, biopsy could have been avoided by 10.4% (25 of 240) when a conservative cutoff value (1.68 × 10⁻³ mm²/sec) was applied.¹⁴ Presumed sensitivity was 92.1% (58 of 63) and 95.2% (60 of 63), respectively. Notably, 5 false negative findings were all nonmass enhancements and were 4 ductal carcinomas in situ (DCIS) and 1 lobular carcinoma.¹⁴ Thus, the use of ADC cutoffs in breast MRI to avoid unnecessary biopsy while maintaining diagnostic accuracy has proven feasible in clinical practice, particularly for suspicious masses identified on CE MRI. However, as nonmass enhancements tend to have less clear diffusion characteristics, when applying ADC thresholds to nonmass enhancements, caution is required and additional imaging should be considered (Figs. 6 and 7).

Fig. 7.

MR images of a 57 year-old woman with a pure ductal carcinoma in situ in the left breast. (a) Screening mammography shows a subtle architectural distortion (arrows) in the left outer breast. (b) Axial DWI with b = 0 sec/mm² shows a subtle nonmass with high signal intensity at the glandular-fat junction of the left breast (arrows). (c) DWI with b = 800 sec/mm² shows a subtle nonmass with high signal intensity (arrows). (d) DWI with b = 1200 sec/mm² shows a subtle nonmass with high signal intensity (arrows). (e) On the ADC map calculated using the b values of 0 sec/mm² and 800 sec/mm², mean ADC value was 1.31 × 10⁻³ mm²/sec (arrows). (f) Contrast enhanced T1WI obtained 90 seconds after contrast injection shows the lesion as a segmental nonmass enhancement (arrows). Surgical histopathology revealed a 5.5 cm low grade ductal carcinoma in situ (ER-positive, PR-positive, and HER2-negative). ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging; T1WI, T1-weighted image.

Standalone Screening

Although CE MRI offers the highest sensitivity in breast cancer screening, but has limitations such as high cost, long scan time, and the need for gadolinium-based contrast agents, which are unsuitable for pregnant women or renal failure patients. Concerns have also emerged about gadolinium accumulation in the brain with unclear long-term effects. Thus, clinical trials are currently underway on the potential of DWI as a standalone screening method.⁵ Multiple studies have investigated performance of unenhanced DWI for cancer detection in simulating screening settings. In the studies, readers retrospectively assessed DWI in cohorts comprising patients with positive or negative imaging findings, as well as healthy controls. The evaluations were performed with or without inclusion of T1-weighted or T2-weighted images, while readers remained blinded to CE MRI findings to minimize bias.^15–18 The results showed that the mean sensitivity was 77% (range 45%–94%) and the mean specificity was 89% (range 79%–94%) in the studies.

Sensitivity tended to decrease when study design was similar to screening settings, such as low cancer prevalence or mammographically occult cancer alone cohort. In addition, advanced DWI techniques such as readout-segmented (RS) EPI DWI with background suppression used in studies by Kang et al. and Telegrafa et al. were associated with higher sensitivity than other studies in which single-shot EPI DWI was used.⁵

Previous studies also reported that DWI detected more cancers than mammography alone,^17,19,20 or MRI-guided focused ultrasonography (US).²⁰ Compared with abbreviated breast MRI, DWI showed comparable cancer detection rate.^21,22 Compared with DCE MRI, sensitivity of DWI was consistently lower.^23–25

Notably, in a blind reader study using a cohort more similar to screening setting, which included 1130 patients with approximately 2% prevalence of clinically occult contralateral breast cancer, DWI using RS EPI showed lower sensitivity (77.8% vs. 96.8%, P < 0.001) and higher specificity (87.3% vs. 84.6%, P < 0.001) compared with DCE MRI.²⁵ Compared with combined mammography with US, however, DWI showed higher sensitivity (76.7% vs. 40.0%) and higher PPV2 (positive predictive value2, 42% vs. 18.5%) in other study using a similar cohort.²⁶

Based on these studies, several prospective multicenter trials are underway to investigate the performance and to identify adequate indications of standalone DWI screening. Multi-institutional DWIST group proposed standardized lexicon and interpretation criteria for the trials (Fig. 4). They are comparing the diagnostic performance of standalone DWI with that of mammography, US, and DCE MRI for women at a high risk of developing breast cancer (ClinicalTrials.gov identifier NCT03835897). In addition, they are comparing the performance of DWI + DCE MRI versus DCE MRI for women with a personal history of breast cancer (NCT04619186), contralateral breast cancer (NCT05307757), or multifocal, multicentric breast cancer (NCT04656639) in women with newly diagnosed breast cancer.^27,28 Other group is also evaluating the usage of standalone DWI screening for women with dense breasts (NCT03607552).²⁹

Of the multi-institutional trials, the DWIST group reported the first-year outcome for 1040 high-risk women.³⁰ In the analysis, the sensitivity of DWI was 73.2% (95% CI 58.1–84.3) and the specificity was 80.0% (95% CI 77.1–82.1). The sensitivity of DWI was higher than that of mammography (39.0%, 95% CI 25.7–54.3, P = 0.002), similar to US (58.5%, 95% CI 43.4–72.2, P = 0.149), but lower than that of DCE MRI (90.2%, 95% CI 77.5–96.1, P = 0.023). The specificity of DWI was lower than that of mammography (88.6%, 95% CI 86.5–90.4) and similar to that of US (80.2%, 95% CI 77.6–82.5) and DCE MRI (76.7%, 95% CI 74.0–79.2). Majority (85.4%, 35 of 41) of detected cancers were DCIS (n = 12, 29.3%) or T1 tumors (n = 23, 56.1%). These results suggest that standalone screening DWI could be used as a supplemental screening modality to mammography. In addition, it could be an appropriate alternative to DCE MRI for high-risk women who cannot undergo gadolinium-based contrast agent use, as well as those with an intermediate risk of breast cancer.

Prognostic Indicators

As DWI reflects cell density and tissue microstructure, there has been a lot of studies investigating association between DWI findings and tumor characterization such as tumor grade, molecular subtype, or invasiveness of DCIS lesions diagnosed at core needle biopsy.

Regarding the differentiation of pure DCIS and invasive cancers, earlier studies compared the ADC values between pure DCIS and DCIS with invasive components and consistently reported that pure DCIS shows higher ADC values than DCIS with invasive components due to the lower cellularity (Figs. 8 and 9).^31–34 Recent study found that a predictive model including preoperative DWI, clinical and pathological factors showed good performance for predicting upstage to invasive cancer of DCIS at percutaneous biopsy.³⁵ In the study, in addition to the ADC value, integrating clinicopathologic features such as the biopsy method (14-gauge vs. 10-gauge needle), nuclear grade (high nuclear grade vs. low to intermediate grade), lesion size (>5 cm vs. ≤5 cm) and lesion type (mass vs. nonmass) improved characterization performances.³⁵ The predictive model combining DWI and clinical-pathologic factors showed AUC of 0.87.³⁵

Fig. 8.

MR images of a 56 year-old woman with a pure ductal carcinoma in situ. It was a ductal carcinoma in situ at initial needle biopsy. Surgical histopathology revealed a 2.0 cm high grade ductal carcinoma in situ (ER-negative, PR-negative, and HER2-negative). (a) Mammography shows grouped pleomorphic calcifications in the left lower breast. (b) Axial DWI with b = 0 sec/mm² shows a focal nonmass with high signal intensity (arrow) in the medial portion of left breast. (c) DWI with b = 800 sec/mm² shows a focal nonmass with high signal intensity (arrow). (d) DWI with b = 1200 sec/mm² shows a focal nonmass with high signal intensity (arrow). (e) On the ADC map calculated using the b values of 0 sec/mm² and 800 sec/mm², the mean ADC value of the lesion was 1.29 × 10⁻³ mm²/sec (arrow). (f) Contrast enhanced T1WI obtained 90 seconds after contrast injection shows the lesion as a focal nonmass enhancement (arrow). ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging; ER, estrogen receptor; PR, progesterone receptor; T1WI, T1-weighted image.

Fig. 9.

MR images of a 57 year-old woman with histologic upgrading at surgery. It was a ductal carcinoma in situ at initial needle biopsy. Surgical histopathology revealed a 2.5 cm high grade invasive ductal carcinoma (ER-negative, PR-negative, and HER2-negative). (a) Mammography shows a round mass in the outer breast. (b) Axial DWI with b = 0 sec/mm² shows a round mass with high signal intensity (arrow) in the left breast. (c) DWI with b = 800 sec/mm² shows a round mass with high signal intensity (arrow). (d) DWI with b = 1200 sec/mm² shows a round mass with high signal intensity (arrow). (e) On the ADC map calculated using the b values of 0 sec/mm² and 800 sec/mm², the mean ADC value of the mass was 0.95 × 10⁻³ mm²/sec (arrow). (f) Contrast enhanced T1WI obtained 90 seconds after contrast injection shows the lesion as a round enhancing mass (arrow). ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging; ER, estrogen receptor; PR, progesterone receptor; T1WI, T1-weighted image.

In addition to predicting presence of invasive components of DCIS lesions diagnosed by percutaneous biopsies, many studies have explored whether ADC values can distinguish between tumor groups with better prognosis and tumor groups with worse prognosis. However, there have been various results in the literature. In early studies, there was an inverse correlation between ADC values and histologic grade. Grade 1 tumors showed higher ADC value than Grade 3 tumors.^36,37 But, another studies did not find any significant differences in ADC values according to the histologic grade.^31,38,39

With respect to estrogen receptor (ER) or progesterone receptor (PR) expression or molecular subtypes, several studies reported that lower ADC values were associated with ER-positive tumors.^31,40–43 As for HER2-positive tumors, significant correlation was found between ADC value and HER2 expression and higher ADC values were associated with HER2-positive tumors.^40,43 In other studies, however, no significant correlation was found between mean ADC values and HER2-positive tumors.^31,44 Park et al. reported that ER and PR status were not correlated with mean ADC values, but HER2-positive cancers showed higher mean ADC values.⁴⁵ Several studies found no significant correlation between ADC and ER or HER status.^46–48

Regarding Ki-67, in tumors with higher Ki-67 had significantly lower mean ADC values than in tumors with lower Ki-67.^49,50 Histogram analysis parameters such as minimum, mean, 25th, 50th, and 75th percentiles of tumor ADC also negatively correlated with the Ki-67 labeling index.⁵¹ However, a subsequent multicenter study including 845 patients reported a weak statistical significance between ADC values and Ki-67 expression.³⁹ As in previous studies, tumors with low Ki-67 tended to show higher ADC values than those with high Ki-67, but there was a large overlap between the 2 groups. Therefore, receiver operating characteristic curve analysis to distinguish tumors with high versus low Ki-67 tumors using ADC values revealed that the AUC was only 0.574.³⁹ The authors suggested that ADC could not be used as a marker for proliferation activity or tumor grade.

Recent systematic review including publications between January 2008 and July 2023 reported that ER-positive or PR-positive cancer showed significantly lower ADCs than ER-negative or PR-negative cancer, respectively.⁵² HER2-negative cancers had significantly lower ADCs than HER2-positive cancers.⁵² Ki-67-positive cancers had significantly lower ADCs than Ki-67-negative cancers, although there was a large heterogeneity.⁵² However, other meta-analysis reported that ADC cannot discriminate between molecular subtypes.⁵³

Inconsistent results regarding the association between ADC values and prognostic factors in the literature might be due to different histological thresholds such as Ki-67 positivity or different techniques for image analysis. Ki-67 positivity ranges from 10% to 50% in the literature.³⁹ In terms of image analysis, Arponent et al. found that lower ADC values were correlated with axillary lymph node metastasis and higher tumor grade in the group using a small ROI, but not in the group using whole-tumor ROI.⁴⁸ The authors explained that these results were probably due to the fact that the small ROI method could represent the most aggressive tissue components, similar to histological diagnoses, and was able to minimize the partial volume average of fat or fibroglandular tissue.

Recently, radiomics and super-resolution ADC images have been used for the breast cancer characterization. In a study, the use of deep learning-based interpolation method and radiomics analysis improved the prediction of Ki-67 expression and tumor grade as well as the spatial resolution of ADC images.⁵⁴ Judging from these results, deep learning-based image interpolation and radiomics analysis could be promising techniques to improve the performance of DWI diagnosis.

Treatment Monitoring

Neoadjuvant chemotherapy (NAC) is used to increase the chances of breast-conserving surgery in breast cancer patients by downstaging primary breast tumors. Also, achieving pathological complete response (pCR) following NAC is a favorable prognostic factor.⁵⁵ However, as approximately 10% of tumors do not respond to NAC, early prediction of response might provide an opportunity to modify treatment and to avoid adverse effects of ineffective treatment.^56,57

DWI reflects decreased cell membrane integrity, cell density, and tissue microstructural changes due to cytotoxic chemotherapy. Thus, it has been expected that DWI would be useful for the prediction or monitoring of response to NAC. There have been various study results depending on the treatment regimens, timing of MRI examinations (pretreatment vs. following 1st cycle of NAC vs. completion of NAC), techniques of DWI image acquisition, image analysis methods (mean or median value of single ROI or whole tumor vs. texture analysis), and definition of response criteria (Response Evaluation Criteria in Solid Tumors vs. pathological response grading vs. pCR).

In terms of pre-treatment timing of MRI examinations, several studies reported that lower pretreatment tumor ADC was associated with clinical or pathological responses;^58–60 however, other studies did not.^61–64

Many studies have reported the potential of DWI as an early predictor of response to NAC. Tumor ADC value tended to increase following NAC before tumor size changes.^63,65 Changes of tumor ADC value following treatment was significantly greater in patients with pCR than those with non-pCR in multiple studies (Figs. 10 and 11).^60,63 Notably, in a prospective study including 62 patients, at post-1st cycle of NAC, the % change of tumor ADC was significantly higher in the pathological response group than in nonresponse group, while the % change of tumor size at CE MRI was not different according to the pathological responsiveness.⁶³ However, in a prospective multicenter study of the ACRIN 6698 trial including 272 patients, the change of tumor ADC was not predictive of pathological response at early-treatment (3 weeks), but predictive of pCR at mid treatment time point (12 weeks).⁶⁶ In the study, patients with pCR showed greater increases in ADC than patients without pCR. Also, higher predictive performance of ADC change was seen in hormone receptor-positive, HER2-negative tumors.⁶⁶ Meta-analyses reported that the pooled sensitivity, specificity, and AUC of DWI in evaluating pathological responses were 88%–89%, 72%–79%, and 0.91%.^67,68

Fig. 10.

MR images of a 42 year-old woman with invasive ductal carcinoma with partial response following chemotherapy. Axial DWIs with b = 1000 sec/mm² at baseline (a), at 3 weeks after chemotherapy (b), and at 5 weeks after chemotherapy (c) show an irregular mass with high signal intensity without definite interval changes following chemotherapy. ADC maps calculated using the b values of 0 sec/mm² and 1000 sec/mm² at baseline (d), at 3 weeks after chemotherapy (e), and at 5 weeks after chemotherapy (f) show an irregular mass with diffusion restriction. The mean ADC value of the tumor was 0.795 × 10⁻³ mm²/sec, 0.844 × 10⁻³ mm²/sec, and 0.881 × 10⁻³ mm²/sec, respectively. After completion of chemotherapy, contrast enhanced T1WI obtained 450 seconds after contrast injection (g) shows an irregular enhancing mass (thick arrows) with internal clip (thin arrow). Surgical histopathology revealed residual 2.5 cm invasive ductal cancer (ER-negative, PR-negative, and HER2-negative). ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging; ER, estrogen receptor; PR, progesterone receptor; T1WI, T1-weighted image.

Fig. 11.

MR images of a 45 year-old woman with pathological complete response following chemotherapy in the right breast (ER-negative, PR-negative, and HER2-negative). Axial DWIs with b = 1000 sec/mm² at baseline (a), at 3 weeks after chemotherapy (b), and at 5 weeks after chemotherapy (c) show an irregular mass with decreased sizes following chemotherapy. ADC maps calculated using the b values of 0 sec/mm² and 1000 sec/mm² at baseline (d), at 3 weeks after chemotherapy (e), and at 5 weeks after chemotherapy (f) show an irregular mass with diffusion restriction (arrow). The mean ADC value of the tumor was 0.778 × 10⁻³ mm²/sec, 0.809 × 10⁻³ mm²/sec, and 1.100 × 10⁻³ mm²/sec, respectively. After completion of chemotherapy, contrast enhanced T1WI obtained 450 seconds after contrast injection (g) shows no residual enhancement around the clip (arrow). Surgical histopathology revealed no residual tumor. ADC, apparent diffusion coefficient; DWI, diffusion weighted imaging; ER, estrogen receptor; PR, progesterone receptor; T1WI, T1-weighted image.

DWI has shown promise as an early predictor of chemotherapy response in many studies, offering the potential for personalized treatment strategies. However, findings are various depending on the timing of MRI examinations, image analysis techniques, and the molecular subtypes of breast tumors, which limits its clinical applicability. With further evidence from large, multicenter prospective studies, DWI may evolve into a valuable tool for early response prediction in breast cancer treatment.

Emerging Techniques

The image quality issues have limited clinical usage of breast DWI. Prospective multicenter trials such as ACRIN 6698 and 6702 reported that 13%–30% of the initially enrolled lesions were non-evaluable due to misregistration, inadequate fat suppression or excessive artifacts.^69,70 The most commonly used sequence for DWI is single-shot EPI in which k-space encoding data are obtained for a single slice after each slice excitation. High susceptibility artifacts, spatial distortions, and low spatial resolution are related with the single-shot EPI.

To improve image quality, multi-shot techniques, such as Siemens’ RS EPI and GE’s multiplexed sensitivity encoding, have been implemented in breast imaging.^71,72 These methods divide k-space data acquisition across multiple segments, reducing the echo train length to achieve higher spatial resolution and minimize susceptibility artifacts and image distortions.^71,72 Because multi-shot techniques require longer acquisition times, the simultaneous multislice (SMS) acceleration technique—based on the blipped ‘Controlled Aliasing In Parallel Imaging Results In Higher Acceleration’ approach—was developed to expedite scanning by simultaneously exciting multiple k-space slices.⁷³ The SMS technique has inherent limitation of lower SNR—due to cross-talk, signal division, and increased noise from g-factor penalty.⁷⁴ This issue might affect the detectability of small breast lesions. However, in comparisons, SMS RS-EPI showed better image quality and comparable ADC measurements to RS-EPI, while reducing scan time by 44%.⁷³ Another study demonstrated that multiband (a type of SMS) sensitivity encoding-accelerated single-shot EPI achieved similar image quality and ADC measurements as conventional single-shot EPI.²⁹ Based on these findings, the lower SNR of the SMS technique does not appear to significantly impact qualities of breast DWI, supporting its effectiveness in clinical applications.

Reduced FOV EPI was also proposed to improve spatial resolution, SNR, and artifacts.⁷⁵ However, it cannot be used for whole breast coverage, which limits clinical usage for cancer detection, although it might help tumor characterization. In addition, post-processing techniques such as fusion images, maximal intensity projection, background suppression, and computed DWI could also improve image quality.^18,76 Using deep learning to generate simulated CE MRI from non-contrast images including T1WI, T2WI, DWI, and ADC may be an alternative to breast cancer imaging modality in women unable to use gadolinium.⁷⁷ These acquisition time reduction and post-processing techniques could lead to high spatial resolution DWI approaching conventional DCE MRI, which will further expand the clinical applications of breast DWI.

Conclusion

In the field of breast DWI, recent technical advancements and collaborative efforts have addressed the artifact issues and the standardization issues. The applications as a prognostic predictor and a predictor of chemotherapy response have not yet been translated into clinical practice. In clinical practice, the complementary role of DWI is evident in reducing unnecessary biopsies. Particularly for masses, DWI has been shown to reduce biopsies for additional suspicious lesions detected by DCE MRI by 15%–30%. However, for the nonmass enhancements, caution is required and other images should be considered. With regard to a standalone screening tool, the DWIST trial results have shown that the diagnostic performance of DWI surpasses that of mammography and is comparable to US. However, DWI remains inferior to DCE MRI in terms of sensitivity. Thus, defining appropriate indications for DWI as a standalone screening tool is essential to optimize its clinical utility, such as populations unable to use gadolinium contrast agents or those at intermediate risk for breast cancer. As MRI technology continues to advance, efforts to reduce acquisition time, minimize artifacts, and improve spatial resolution—approaching that of conventional DCE MRI—are ongoing. Additionally, post-processing techniques, such as deep learning-based image interpolation and radiomics, are being integrated into breast DWI. These advancements might expand the role of DWI from its current use in tumor characterization and detection to real-time monitoring of treatment responses and physiological changes, offering a broader scope of applications in the near future.