Michael Phillip Kempston, Presenter: Nothing to Disclose

James Gordon Mainprize PhD, Abstract Co-Author: Nothing to Disclose

Martin Joel Yaffe PhD, Abstract Co-Author: Nothing to Disclose

Tomosynthesis is a limited-view cone-beam CT technique that can potentially improve lesion visibility and localization in breast imaging. To maintain the x-ray dose at an acceptable level while achieving high image quality, careful optimization of the acquisition technique is necessary. The purpose of this study is to examine the impact of the number of x-ray projection images and the angular range over which they are acquired on the quality of the reconstructed image.

A full-field digital x-ray detector was used to acquire two data sets: a simple phantom consisting of two low-contrast objects in a homogeneous breast tissue-equivalent background, and a lumpectomy sample embedded in agar. A complete set of N=101 projections was acquired at 2° intervals. Reconstructed volumes were created by selecting different subsets of the projections from the original data. Reconstruction quality was evaluated by determining the contrast, relative noise, and signal difference to noise ratio (SDNR) for background and lesion regions of interest in several slices.

Background noise in the reconstructed slices consists of quantum fluctuations, reconstruction artifacts and tissue structure noise. When a fixed dose per projection was used, the relative noise for the reconstructed lumpectomy sample decreased from 100% (N=9) to 37% (N=23), with a lesser reduction for high N (18% at N=93). Extrapolating to a fixed total dose (i.e. normalizing to the dose used for N=23 projections) suggests that the relative noise falls from 100% (N=9) to 60% (N=23), with a negligible decline at high N (58% at N=93). This agrees with qualitative observations, where both the background suppression and image uniformity improved dramatically as N increased, with less noticeable improvement for N>23.

A pronounced reduction in background noise, with an acceptable level of image uniformity and minimal presence of artifacts can be achieved using only a relatively small number of projections (N≈20). Further analysis is required to evaluate the quality of reconstructed images for different tissue samples and phantoms under various dose conditions.

M.P.K.,J.G.M.,M.J.Y.: Martin Yaffe's lab has a research collaboration with GE Healthcare on certain topics in breast cancer imaging.

Kempston, M, Mainprize, J, Yaffe, M, The Effect of Projection Number and Angular Spacing on Image Quality in Breast Tomosynthesis.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4408602.html