Abstract Archives of the RSNA, 2008
Despina Kontos PhD, Presenter: Nothing to Disclose
Predrag R. Bakic PhD, Abstract Co-Author: Nothing to Disclose
Ann Katherine Carton PhD, Abstract Co-Author: Nothing to Disclose
Andrea B. Troxel DSc, Abstract Co-Author: Nothing to Disclose
Emily F. Conant MD, Abstract Co-Author: Nothing to Disclose
Andrew D.A. Maidment PhD, Abstract Co-Author: Consultant, XCounter AB
Scientific Advisor, Real-Time Tomography, LLC
Spouse, Owner, Real-Time Tomography, LLC
Spouse, CEO, Real-Time Tomography, LLC
To explore the potential advantages of digital breast tomosynthesis (DBT) parenchymal texture analysis for breast cancer risk estimation. We compare the performance of DBT and digital mammography (DM) texture features in correlating to breast percent density (PD), an established surrogate of breast cancer risk.
We analyzed DBT and DM images from 39 women with recently detected abnormalities and/or previously diagnosed breast cancer (age range 31-80 yrs, mean 51.4 yrs). DBT and DM acquisition was performed on the same day with a GE Senographe 2000D FFDM system modified to allow DBT. Filtered-backprojection was used to reconstruct DBT tomographic planes in 1 mm increments. Retroareolar (2.5 cm)³ ROIs were manually segmented from the DBT images of the unaffected breasts; corresponding (2.5 cm)² ROIs were segmented from the DM images. Texture features of skewness, coarseness, contrast, and energy were computed. Mammographic breast percent density (PD) was estimated using the thresholding technique of Cumulus (Ver. 4.0, Univ Toronto). The Pearson correlation (r) was estimated between the computed texture features and the corresponding PD estimates. Linear regression was performed to examine differences in texture patterns between groups of women with increasing breast PD.
DBT texture features have a stronger correlation with breast PD, in comparison to DM. The correlation between DBT texture features and PD was equal to r=0.18 for skewness (p=0.26), r=0.46 for coarseness (p=0.003), r=-0.31 for contrast (p=0.05), and r=-0.36 for energy (p=0.03); for DM the corresponding correlations were equal to r=-0.18 for skewness (p=0.26), r=0.15 for coarseness (p=0.34), r=-0.25 for contrast (p=0.13), and r=-0.29 for energy (p=0.07). The association between increasing PD and texture features was also stronger for DBT than for DM, as evidenced by regression lines with steeper slopes and statistically significant p-values (p≤0.05).
By alleviating the effect of tissue superimposition, DBT parenchymal texture analysis could offer more accurate features to characterize breast density patterns and could ultimately provide more reliable measures for breast cancer risk estimation.
DBT breast cancer risk biomarkers could be used to offer customized screening options, tailor individual treatment and form preventive strategies, especially for women at high risk of breast cancer.
Kontos, D,
Bakic, P,
Carton, A,
Troxel, A,
Conant, E,
Maidment, A,
Parenchymal Texture Analysis in Digital Breast Tomosynthesis for Computer-assisted Risk Estimation (CARe) of Breast Cancer: A Preliminary Evaluation. Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL.
http://archive.rsna.org/2008/6010863.html