Ingrid Reiser PhD, Presenter: Nothing to Disclose

Robert Mark Nishikawa PhD, Abstract Co-Author: Nothing to Disclose

Maryellen Lissak Giger PhD, Abstract Co-Author: Nothing to Disclose

Elizabeth Ann Rafferty MD, Abstract Co-Author: Nothing to Disclose

Daniel Benjamin Kopans MD, Abstract Co-Author: Nothing to Disclose

Richard H. Moore MD, PhD, Abstract Co-Author: Nothing to Disclose

Tao Wu PhD, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

Digital Breast Tomosynthesis has recently emerged as a promising improvement to mammography. Reconstruction algorithms for this modality are not yet fully optimized. Because computerized lesion detection in the breast volume will be affected by the reconstruction technique, we developed a mass detection algorithm that operates on the set of raw projection images.

An image set for one breast consisted of 11 projection images, at source angles equally spaced over an arc of 50 degrees. The detection algorithm followed a two-step approach. First, lesion candidates were obtained for each projection image separately, using a mass detection algorithm which was adapted from screen-film mammography. The second step geometrically projected lesion candidate locations into the breast volume. In this feature volume, voxel intensities were a combined measure of detection frequency (e.g. the number of projections in which a given lesion candidate was detected), as well as a measure of the angular range over which a given lesion was detected. The database used to test the algorithm consisted of 21 mass cases (13 malignant, 8 benign).

Based on this database, the algorithm yielded a sensitivity of 86% at 6 false positives per breast volume. Algorithm performance can be further improved by including feature analysis of the lesion candidates.

Our results indicate that computerized mass detection in projection images is successful despite the higher noise level in those images. Taking advantage of the correlation within the image set allows to efficiently reduce false positive detections.

D.B.K.: Patent holder for the technique of Digital Breast Tomosynthesis.E.A.R.,D.B.K.,R.H.M.: Receive some research support from the General Electric Company through the Breast Imaging Division at the Massachusetts General Hospital.M.L.G.: Shareholder & received funds for research from: R2 Technology, Inc., Sunnyvale, CA.R.M.N.: Shareholders in R2 Technology, Inc. (Sunnyvale, CA).

Reiser, I, Nishikawa, R, Giger, M, Rafferty, E, Kopans, D, Moore, R, Wu, T, et al, , Computerized Mass Detection for Digital Breast Tomosynthesis Directly from the Projection Images.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4411369.html