Program Information
Volumetric Breast Density: Comparison of Estimates From Tomosynthesis Reconstructions with Mammography
L Shi1*, S Vedantham1 , K Michaelsen2 , V Krishnaswamy2 , A Shenoy2 , B Pogue2 , A Karellas1 , K Paulsen2 , (1) University of Massachusetts Medical School, Worcester, MA, (2) Dartmouth College, Hanover, NH.
Presentations
SU-E-I-54 Sunday 3:00PM - 6:00PM Room: Exhibit HallPurpose: To measure the volumetric breast density (VBD) from digital breast tomosynthesis (DBT) reconstructions and compare with estimates from mammography.
Methods: A kernel-based fuzzy c-means algorithm developed to segment DBT reconstructed slices to provide anatomic priors for near-infrared spectroscopy (NIRS) in a multimodality NIRS/DBT clinical-prototype imaging system was used to quantitate VBD. The VBD was determined from 17 DBT reconstructions of 11 breasts from 9 subjects (6 subjects–unilateral 2-view, 1 subject–unilateral craniocaudal (CC) view, 2 subjects–bilateral CC view) who underwent multimodality NIRS/DBT imaging. VBD from mammography (Quantra, Hologic Inc.) matched for subject, breast laterality and view was used for comparison. To account for differences in compressed breast thickness (T) between the modalities, the metric normalized VBD (nVBD = VBD/T) was considered.
Results: All data satisfied normality assumption (Shapiro-Wilks, p>0.06). Pearson’s correlation (r) and paired t-tests were used for analysis. The normalized VBD (nVBD) was statistically correlated (r=0.66, p<0.01) between the two modalities, but the VBD was not (r=0.47, p=0.06), as the compressed breast thickness (p=0.01, difference=6.6+/-9.8 mm) and the total breast volume (p<0.01, difference=183+/-169 cc) were higher with mammography than with clinical-prototype NIRS-coupled DBT system, while the fibroglandular volume was not different (p=0.21). The intra-modality nVBD assessed between 2-views of the same breast (n=6) were better correlated (r=0.86, p=0.03) in DBT than in mammography (r=0.57, p=0.24).
Conclusion: Normalized VBD (nVBD) could be a better metric when compressed breast thickness differs between the modalities. nVBD estimated from DBT is better correlated between 2-views of the same breast than mammography and could improve the consistency in the estimate. The algorithm, in addition to providing anatomic priors for NIRS reconstruction, may also be suitable for breast density estimation. Ability to use DBT reconstructions could facilitate widespread adaptation. Future studies will investigate VBD from images acquired on a combined mammography/DBT system.
Funding Support, Disclosures, and Conflict of Interest: This work was supported in parts by National Institutes of Health (NIH) grants R01CA139449 and R21CA176470. The contents are solely the responsibility of the authors and do not represent the official views of the NIH or the National Cancer Institute.
Contact Email: