Spatial Frequency Characterization of the X-Ray Scatter Signal in Breast Imaging
I Sechopoulos1*, B Fei1, K Bliznakova2, (1) Emory University, Atlanta, Georgia, (2) Technical University of Varna, Varna, BulgariaWE-G-103-5 Wednesday 4:30PM - 6:00PM Room: 103
Purpose: To determine the frequency characteristics of the x-ray scatter signal in mammography and breast tomosynthesis projections.
Methods: Dedicated breast computed tomography images of 19 patients were classified into skin, adipose, and glandular tissue using a previously validated, automatic classification method. The segmented breasts underwent simulated mechanical compression to mimic breast compression during mammographic and tomosynthesis acquisition. Using Monte Carlo methods, simulated projection images of the primary and scatter x-ray signals were generated for each patient breast with the x-ray source located at both zero and thirty degrees. The noise power spectra (NPS) of the scatter and total (primary + scatter) signals were obtained for each patient image and projection angle. The resulting NPS were radially averaged and the results for all patients combined. As validation, the total NPS were fit to the expected power-law relationship NPS(f) = k/f^β to compare the value of β with previous published reports. The scatter NPS were inspected visually and a power-law fit was also performed.
Results: The values of β for the total NPS fit were 3.39 and 3.35 for the zero and thirty degree projection angles, respectively. As expected, the power of the scatter signal decreases rapidly with spatial frequency, with a reduction of four orders of magnitude at 0.1 lp/mm, beyond which the random white noise due to the statistical nature of the Monte Carlo simulations dominates. The power-law fit for the x-ray scatter signal were 6.10 and 6.42 for the zero and thirty degree projection angles, respectively.
Conclusion: Although it is known that the x-ray scatter signal in mammography and breast tomosynthesis is dominated by low frequencies, quantitative evaluation of its frequency characteristics had not been undertaken before. This characterization will allow for more accurate software-based scatter reduction algorithms for mammography and breast tomosynthesis.
Funding Support, Disclosures, and Conflict of Interest: Funding Support: This research was supported in part by National Institutes of Health (NIH) Grants Nos. R01CA156775 (PI: B. Fei), R01CA163746 (PI: Sechopoulos), Georgia Cancer Coalition Distinguished Clinicians and Scientists Award (PI: Fei), Emory Molecular and Translational Imaging Center (NIH P50CA128301), SPORE in Head and Neck Cancer (NIH P50CA128613), and Atlanta Clinical and Translational Science Institute (ACTSI) that is supported by the PHS Grant No. UL1 RR025008 from the Clinical and Translational Science Award program. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health. Disclosure: Institutional Research Grants, Fuji (Sechopoulos)