Hormonal effect on time-dependent diffusion of the breast fibroglandular tissue
Sungheon Gene Kim1,2, Eric Sigmund1,2, Melanie Moccaldi1,2, Thorsten Feiweier3, and Linda Moy1,2

1Center for Advanced Imaging Innovation and Research, Radiology, New York University School of Medicine, New York, NY, United States, 2Bernard and Irene Schwartz Center for Biomedical Imaging, Radiology, New York University School of Medicine, New York, NY, United States, 3Siemens Healthcare GmbH, Erlangen, Germany

### Synopsis

This study is to investigate the potential of the surface-to-volume ratio obtained from multiple diffusion times to measure mammary duct microstructural changes induced by hormonal variation. Seven premenopausal women were scanned twice using a stimulated-echo diffusion sequence; one in the follicular phase of the menstrual cycle and again in the luteal phase. In 6 out of 7 subjects, the surface-to-volume ratio measured with 69 and 173 ms was significantly reduced in the luteal phase compared to the follicular phase. The length scales obtained in our study are consistent with the duct diameters reported in previous ex-vivo studies.

### PURPOSE

Recently, it has been reported that the fractional anisotropy of fibroglandular tissue (FGT) increases significantly when a longer diffusion time is used.1 Furthermore, it was reported that the ADC measurement with same b-value and multiple diffusion times can be used to estimate the surface-to-volume ratio (SVR) of the microstructural component that restricts water diffusion, which was proposed to be dominated by mammary duct inner lining.2 This study is to investigate the potential of the SVR measured from multiple diffusion times as a novel marker for the microstructural changes of the mammary duct induced by the hormonal variation during the menstrual cycle.

### METHODS

Seven asymptomatic, premenopausal women (32±7 years) were scanned twice; once in Week2 (follicular phase) of the menstrual cycle and again in Week4 (luteal phase). All scans were performed using a whole-body MAGNETOM Trio 3T scanner (Siemens Healthcare, Erlangen, Germany) with a 7-element breast coil array. We measured diffusion using a prototype stimulated-echo acquisition mode (STEAM)-DTI sequence with an echo-planar imaging (EPI) readout and SPAIR fat suppression (TR/TE = 11500/45 ms, matrix = 192x132x10, resolution = 2.1x2.1x5 mm3, 6 directions, 3 averages, parallel imaging factor = 2) with two different b-values (0 and 500 s/mm2). The acquisition was repeated using 4 different diffusion times (Δ=69, 173, 466 and 903 ms) by varying the mixing time.

Diffusion-weighted images were corrected for eddy-current-induced distortion using a non-affine registration. The contribution of the imaging gradients was included in the b-matrix for tensor estimation. Parametric maps of mean diffusivity (MD), fractional anisotropy (FA), axial diffusivity (AD), and radial diffusivity (RD) were derived from the diffusion eigenvalues (λ1, λ2 and λ3) with AD = λ1 and RD = (λ23)/2. Regions of interest (ROI) were manually drawn to select the FGT in the bilateral breast of all imaging slices. For the voxels in the ROI, selection criteria of AD > 1.0 μm2/ms and AD < 2.7 μm2/ms were used to exclude any voxels with a substantial partial-volume effect, with unsuppressed fat or noise, respectively. The selected voxels were then used to quantify the diffusion characteristics of the FGT using a bootstrapping analysis: the eigenvalues were calculated from the average data of randomly selected 30% of the selected voxels for each diffusion time; this process was repeated 500 times to calculate the mean and standard deviation of the FGT diffusion metrics. With the RD values of diffusion times 69 and 173 ms, surface-to-volume ratio (SVR) was calculated using the Mitra short time limit model3: $$D(\Delta)=D_0[1-SVR\frac{4\sqrt{D_0\Delta}}{6\sqrt{\pi}}]$$

### RESULTS

Figure 1 shows representative images of one subject at two diffusion times, 69 and 173 ms, that show reduced diffusivity at Δ=173ms, compared to Δ=69ms. This trend was consistently observed in all subjects. Figure 2 shows the eigenvalues combined from all subjects in Week2 and Week4. At both menstrual periods, the decrease of the RD was mostly between 69 and 173 ms. Figure 3 shows that the mean AD and RD measured with 69 and 173 ms did not change noticeably between Week2 and Week4. However, the SVR was significantly reduced in Week4 for 6 out of 7 subjects as shown in Figure 4, highlighting the potential benefit of individualized model fitting.

### DISCUSSION and CONCLUSION

Assuming that mammary ducts are the main underlying component in the fibroglandular voxels, the SVR values can be converted to the diameters of a cylindrical model (d=4/SVR). Few studies have explored the size of the milk ducts in the non-lactating breast. Taneri et al4 reported a mean diameter of 0.57 mm in their series of 226 nipples, and Rusby et al5 reported a diameter of 0.7 mm at 3 mm beneath the nipple. These diameters are about 2-3 times larger than what could be estimated from our SVR values. Note that our measurement includes the entire ductal tree rather than only the sub-areolar region. The discrepancy might also be due to the difference between in-vivo and ex-vivo measurements. Nevertheless, the length scales obtained in our study are in the same order of magnitude as in these previous reports. Hence, the findings of this study suggest that the SVR measurement may be used as a means to measure the average duct size of the FGT in vivo. The ability to non-invasively measure hormonal dependent changes in the size of lactiferous duct may allow for improved characterization of these structural changes. A larger cohort of women is necessary to determine if these normal fluctuations in the size of the milk ducts are also seen in premenopausal women with breast cancer.

### Acknowledgements

This work was supported by the NIH (CA160620).

### References

1. Cho G, Leite APK, Baete S, et al. Diffusion Tensor Imaging (DTI) at Multiple Diffusion Times in Mammary Fibroglandular Tissue and Cancerous Lesions. Proceedings of the 21st Annual Meeting of ISMRM 2013; p.3373

2. Teruel J, Cho G, Ostenson J, et al. Stimulated echo diffusion tensor imaging with varying diffusion times as a probe of breast tissue. Proceedings of the 23rd Annual Meeting of ISMRM 2015;p.0883.

3. Mitra PP, Sen PN, Schwartz LM. Short-time behavior of the diffusion coefficient as a geometrical probe of porous media. Physical review B, Condensed matter. 1993;47(14):8565-74.

4. Taneri F, Kurukahvecioglu O, Akyurek N, et al. Microanatomy of milk ducts in the nipple. European surgical research Europaische chirurgische Forschung Recherches chirurgicales europeennes. 2006;38(6):545-9.

5. Rusby JE, Brachtel EF, Michaelson JS, et al. Breast duct anatomy in the human nipple: three-dimensional patterns and clinical implications. Breast cancer research and treatment. 2007;106(2):171-9

### Figures

Figure 1: Representative images of raw images with b=0 (a,b) and MD (c, d) measured at diffusion time Δ=69 ms (a,c) and 173 ms (b, d).

Figure 2: Average eigenvalues (black; O for λ1, flipped-Δ for λ2, and Δ for λ3) from the ROIs of all subjects at the follicular phase (a) and the luteal phase (b). Red markers and lines represent RD ((λ23)/2). Error bars are for standard deviations.

Figure 3: Comparison of average AD and RD between Week2 and Week4 at two diffusion times, 69 and 173 ms. The error bars represents standard deviations.

Figure 4: Comparison of average SVR between Week2 and Week4. The error bars represents standard deviations.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2450