Ioannis Sechopoulos PhD, Presenter: Institutional research collaboration, General Electric Company

Sankararaman Suryanarayanan PhD, Abstract Co-Author: Research collaboration, General Electric Company

Srinivasan Vedantham PhD, Abstract Co-Author: Institutional research collaboration, General Electric Company

Carl Joseph D'Orsi MD, Abstract Co-Author: Consultant, General Electric Company Consultant, Hologic, Inc (R2 Technology, Inc) Stock options, General Electric Company Stock options, Hologic, Inc (R2 Technology, Inc)

Andrew Karellas PhD, Abstract Co-Author: Institutional research collaboration, General Electric Company

To determine the acquisition geometry that maximizes image reconstruction quality in digital tomosynthesis of the breast.

A 3D portion of breast tissue which included masses, spiculations and microcalcifications was simulated using the known power law noise relationship. Tomosynthesis projections of the breast volume were computed using ray-tracing and then modified to include the effects of x-ray scatter and the performance characteristics of a clinical mammography digital imaging system. X-ray scatter was added using published reports of x-ray scatter in tomosynthesis, while the characteristics of the imaging system were obtained from a serial cascaded model which included the angular dependence of the modulation transfer function. An iterative reconstruction program based on the maximum likelihood expectation maximization (MLEM) algorithm that included correction for x-ray scatter was used. Thirty eight different tomosynthesis projection sets were studied, each with a different combination of angular range and number of projections per set. The exposure per projection was varied depending on the total number of projections and angles in a given set, so that the total glandular dose (2 mGy) to the breast remained constant. The quality of the reconstructions was measured using objective measures for in-plane quality and vertical resolution.

A balance between in-plane image quality and vertical resolution was found in projection sets that included between 17 and 21 projections covering between ±24° and ±30°. If the number of projections is too high, the increase in noise due to the lower exposure available per projection dominates, while a minimum number of projections of 9 was found to be necessary to achieve an acceptable tomographic effect.

The reconstruction quality in breast tomosynthesis is maximized with projection sets of approximately 20 projections covering an angular range between ±24° and ±30°.

The optimization of the acquisition geometry is critical to maximize image quality and minimize radiation dose before the clinical translation of tomosynthesis.

Sechopoulos, I, Suryanarayanan, S, Vedantham, S, D'Orsi, C, Karellas, A, Optimization of Acquisition Geometry in Digital Tomosynthesis Imaging of the Breast.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5009713.html