Quantitative volumetric breast density estimation using phase contrast mammography. (1st May 2015)

• Record Type: Journal Article
• Authors: Wang, Zhentian; Hauser, Nik; Kubik-Huch, Rahel A; D'Isidoro, Fabio; Stampanoni, Marco
• Is Part Of: Physics in medicine & biology. Volume 60:Number 10(2015:May)
• Page Start: 4123
• Page End: 4135
• Publication Date: 2015-05-01
• Subjects: phase contrast imaging -- mammography -- breast imaging -- grating interferometry -- volumetric breast density; Biophysics -- Periodicals; Medical physics -- Periodicals; 610.153
• DOI: 10.1088/0031-9155/60/10/4123
• Languages: English
• ISSNs: 0031-9155

Abstract

Abstract: Phase contrast mammography using a grating interferometer is an emerging technology for breast imaging. It provides complementary information to the conventional absorption-based methods. Additional diagnostic values could be further obtained by retrieving quantitative information from the three physical signals (absorption, differential phase and small-angle scattering) yielded simultaneously. We report a non-parametric quantitative volumetric breast density estimation method by exploiting the ratio (dubbed the R value) of the absorption signal to the small-angle scattering signal. The R value is used to determine breast composition and the volumetric breast density (VBD) of the whole breast is obtained analytically by deducing the relationship between the R value and the pixel-wise breast density. The proposed method is tested by a phantom study and a group of 27 mastectomy samples. In the clinical evaluation, the estimated VBD values from both cranio–caudal (CC) and anterior–posterior (AP) views are compared with the ACR scores given by radiologists to the pre-surgical mammograms. The results show that the estimated VBD results using the proposed method are consistent with the pre-surgical ACR scores, indicating the effectiveness of this method in breast density estimation. A positive correlation is found between the estimated VBD and the diagnostic ACR score for both the CC view ( ) and AP view ( ). A linear regression between the results of the CC view and AP view showed a correlation coefficient γ = 0.77, which indicates the robustness of the proposed method and the quantitative character of the additional information obtained with our approach.