Predrag R. Bakic PhD, Presenter: Nothing to Disclose

Ann Katherine Carton PhD, Abstract Co-Author: Nothing to Disclose

Cuiping Zhang PhD, Abstract Co-Author: Nothing to Disclose

Martin J. Yaffe PhD, Abstract Co-Author: Research collaboration, General Electric Company

Andrew D.A. Maidment PhD, Abstract Co-Author: Nothing to Disclose

To evaluate observers’ agreement in estimation of breast percent density (PD) from clinical digital mammography (DM) and digital breast tomosynthesis (DBT) central projection images.

We analyzed mediolateral (MLO) DM and DBT projection images from 39 women, acquired in the period of September 2004 - April 2005, using a GE Senograph 2000D DM machine. The images were acquired as a part of a multimodality clinical study of high breast cancer risk population. Three experienced observers estimated PD from DM and DBT central projection images, using the Cumulus 3 software package developed at the University of Toronto. The PD estimation study was repeated for one of the observers after two months. The DBT central projection images were acquired in the same geometry as mammographic MLO views, with smaller amount of compression than DM, and using on average 22% of the radiation dose of a DM projection. The agreement between observer performance was tested using Pearson correlation coefficients. For each observer, we computed the correlation between PD estimates from DM and DBT images. In addition, we computed the correlation between PD estimates by different observers, separately for DM and DBT images.

Correlation between PD estimates from DM and DBT central projection images, for three observers was 0.852 +/- 0.047 (range 0.802-0.900). Correlation between PD estimates by three different observers was 0.823 +/- 0.030 (range 0.774-0.850) for DM images, and 0.890 +/- 0.028 (range 0.860-0.925) for DBT central projections. Correlation between repeated PD estimation studies by one of the observers was 0.904 for DM images and 0.937 for DBT central projections.

We observed high values of inter- and intra-observer correlation between breast PD estimates from clinical DM and DBT images of 39 women. Inter-observer correlation for DBT central projections was slightly higher than for the DM images.

Mammographic PD is used as a biomarker for breast cancer risk. DBT central projection images, although acquired with lower radiation dose than in DM, can be used for accurate estimation of PD.

Bakic, P, Carton, A, Zhang, C, Yaffe, M, Maidment, A, Inter and Intra-Observer Correlation of Breast Percent Density Estimation from Mammograms and Central Tomosynthesis Projections.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5011676.html