Investigation of correction and decomposition algorithms in bedside x-ray imaging simulating a multi-layer flat panel detector

Abstract

Recently introduced multi-layer flat panel detectors (FPDs) enable single acquisition spectral radiography. We perform an in-depth simulation study to investigate different decomposition algorithms under the influence of adipose tissue and scattered radiation using physics-based material decomposition algorithms for the task of bone removal. We examine a matrix-based material decomposition (MBMD) under assumption of monoenergetic Xray spectra (equivalent to weighted logarithmic subtraction (WLS)), a matrix-based material decomposition with polynomial beam hardening pre-correction (MBMD-PBC) and a projection domain decomposition (PDD). The simulated setup corresponds to an intensive care unit (ICU) anterior posterior (AP) bedside chest examination (contact scan). The limitations of the three algorithms are evaluated using a high-fidelity X-ray simulator with five phantom realizations that differ in terms of added adipose tissue. For each simulated phantom realization, different amounts of scatter correction are considered, ranging from no correction at all to an ideal scatter correction. Unless quantitative imaging is required, the three algorithms are capable of removing bone structures when adipose tissue is present. Bone removal using a multi-layer FPDs in an ICU setup is feasible. However, uncorrected scatter can lead to bone structures becoming visible in the soft tissue image. This indicates the need for accurate scatter estimation and correction algorithms, especially when using quantitative algorithms such as PDD.