Three-Dimensional Tomosynthesis Reconstruction from 1D and 2D X-ray Source Arrays

Abstract

We study the effects of geometric design on the reconstruction of 3D images from an X-ray tomosynthesis system using microfabricated discrete X-ray sources. Carbon-nanotube-based field-emission X-ray sources can be fabricated in arrays; however, little is known about the effects of the geometry of such a system on reconstruction of tomosynthesis data. We produced simulated X-ray projection data for several source array geometries including a 1 times 11 array, and a 3 times 11 array. The phantom simulates a mammography task, with seven 400 mum spheres embedded in a uniformly-absorbing background. Data from both geometries was reconstructed using the ordered-subset convex (OSC) algorithm specially implemented for these array geometries. Reconstruction was performed on an 800 times 800 (lateral) times 20 (depth) array of noncubic voxels of size 200 mum times 200 mum (lateral) times 2.5 mm (depth). Contrast, resolution, and noise measurements on the reconstructed spheres were used to compare results from the different geometries. The reconstruction from the 2D array had higher contrast than that from the 1D array, but the 2D array image also suffered from higher levels of noise for the same total exposure. The contrast-to-noise ratio for the 2D array, was 5-17% higher than the 1D array. Other studies showed that staggering a 2D array is beneficial, and that increasing the density of sources is detrimental after a certain optimal density is reached We conclude that, in a tomosynthesis system such as that described here, a 2D source array will outperform a 1D array in terms of contrast-to-noise and resolution, while also increasing the usable field-of-view, and that further studies encompassing the effects of scatter will be needed to optimize such systems.