There is growing interest in using diffusion-weighted MRI to characterize breast disease and changes in tumors in response to treatment. Our study demonstrates significant contribution to breast ADC measures from unsuppressed intravoxel fat signal, and underlines the importance of achieving robust fat suppression for clinical breast DWI applications.
In this study, 21 women with breast cancer were imaged with DWI both with and without fat suppression. Fat-suppressed mean ADC values in our study (1.59 ± 0.28 × 10−3
/s for tumor, 1.96 ± 0.24 × 10−3
/s for normal fibroglandular tissue) compared well to those reported in other studies after accounting for differences in study design and b-values [4
]. The ADC values for tumor and normal fibroglandular tissue were significantly lower when measured without fat suppression. On the other hand, no difference was observed for ADC measures in a water phantom, validating that the two acquisitions differed only by fat suppression and that the measured ADC differences were not due to other factors. It has been reported previously that fat produces a much lower ADC value than breast fibroglandular tissue due to restricted water mobility [18
]. In our study, breast ADC values were reduced due to subvoxel partial volume averaging of fat and tissue, even when ROIs were carefully placed to avoid adipose tissue regions. This should be taken into consideration if DWI is performed without adequate fat suppression. Furthermore, ADC underestimation was significantly greater in normal tissue than in tumor and in patients with lower mammographic breast density (i.e. fattier breasts), which suggests that there is more partial volume averaging with fat in normal fibroglandular tissue than in tumors and in women with fattier breasts compared to those with dense breasts.
To assess tumor conspicuity, we calculated the tumor ADC contrast-to-noise ratio by comparing ADC values for tumor and normal tissue in each subject. Tumor contrast was higher on the ADC maps created with fat suppression than for those without fat suppression, indicating that tumors could be more easily detected by using fat suppression. This was primarily due to significantly reduced ADC values in the normal tissue without fat suppression. We found no associations between tumor conspicuity and breast density, with similar contrast values observed for each density category. However, there were a limited number of patients with high breast density in our study and our findings need to be validated in a larger cohort. Our results are in contrast to those of Yoshikawa et al. who previously reported better cancer detection rates on DWI in patients with lower breast density [16
]. The difference in our findings may be attributed to study design: in their study, breast tumors were identified by visual inspection of diffusion-weighted images rather than by ADC values and heterogeneously dense and extremely dense were combined into a single breast density category for analysis.
While this study illustrates the importance of fat-suppression, uniform fat suppression is difficult to achieve in the breast. Fat suppression can be implemented through a variety of techniques such as a chemical shift selective (CHESS) pulse, short TI inversion recovery (STIR), and water excitation (the approach used in this study). CHESS selectively nulls signals from fat protons using the chemical shift difference between fat and water protons; however, the shifts of fat and water can be misassigned in breasts with very little fibroglandular tissue in which the fat peak is significantly larger than the water peak. In addition, the shifts may vary due to field inhomogeneities caused by air-fat interfaces, skin folds, and biopsy clips or other hardware [19
]. STIR-based methods may hold more promise in the breast; Kazama et al. found that fat suppression in breast DWI was insufficient in 44% of patients using CHESS and in 0% of patients using STIR [19
]. However Wenkel et al. showed better lesion visibility on breast DWI using CHESS due to reduced signal-to-noise with the STIR technique [20
]. Furthermore, Baron et al. reported that water excitation may provide optimal SNR over other fat suppression methods for breast DWI [9
]. Based on the relationship between fat suppression and ADC, nonuniform fat suppression could impact the ADC differently in different areas of the breast, making it difficult to apply ADC thresholds for cancer detection. In treatment monitoring, nonuniform intra- or inter-scan fat suppression in the tumor region could impact measured ADC changes, potentially resulting in erroneous conclusions about change in tumor ADC and response to treatment. Uniform, reproducible fat-suppression is therefore essential to applications of DW-MRI in the breast.
Our study had limitations. There were a small number of patients in each of the mostly fat (n=1), scattered fibroglandular densities (n=1), and extremely dense (n=5) breast density categories, compared to heterogeneously dense (n=14). We were therefore not able to independently assess the lower density categories, and the finding that ADC error decreases with increasing breast density needs to be validated with larger study groups. In addition, normal tissue ADC values were obtained from the contralateral breast and CNR calculations may not reflect true tumor conspicuity, which would be judged relative to surrounding tissue. The study investigated a single fat suppression method; differences in ADC may vary depending on the approach. However, this should not affect the observation of greater ADC differences (fat-suppressed versus non-fat-suppressed) in normal tissue compared to tumor. Furthermore, ADC was calculated from only two b-values, with minimum b = 0 s/mm2
. Calculation of ADC using a greater number of b-values with a nonzero minimum b would provide more accurate diffusion measures without perfusion effects [21
]. In our study, DWI was performed after injection of a gadolinium-based contrast agent. While several prior studies have not shown a significant effect on ADC following administration of a contrast agent [22
], it may be preferable to acquire DWI sequences prior to contrast injection to avoid any possible confounding effects.
In summary, the results of this study showed that robust fat suppression in DWI is essential for accurate detection and characterization of malignant breast lesions. Intravoxel partial-voluming of unsuppressed fat signal can cause a reduction in the ADC measurement, even in solid appearing breast tumors, and can significantly limit lesion conspicuity. Influences of fat signal on tumor ADC values were most pronounced in breasts with low fibroglandular density. Techniques measuring ADC in the breast without adequately suppressing signal from fat may underestimate the values and be less sensitive to changes with treatment.