Parallel multi-material decomposition of Dual-Energy CT data

Abstract

Dual-Energy Computed Tomography (DECT) is a new modality of CT where two images are acquired simultaneously at two energy levels, and then decomposed into two material density images. It is also possible to further decompose these images into volume fraction images that approximate the percentage of a given material at each pixel. Here, we describe a novel parallel version of the multilateral decomposition algorithm proposed by Mendonça et al., which is used to obtain volume fraction images. Our parallel version accelerates decomposition by 200x. We also discuss some of the algorithm limitations.