Calibration of the error from spectrum estimation for a dual energy CT

Abstract

Dual energy computed tomography (DECT) can be used for substance discrimination by reconstructing the effective atomic number and electron density distribution of scanned object. One critical requirement for DECT reconstruction is estimating both high- and low- energy spectra used in DECT imaging. The estimates of spectra are almost-for-sure off from real spectra to some extend depending on the spectrum estimation methods and data available for spectrum estimation. Hence, systematic error in DECT reconstructions is expected due to spectrum mismatch. In this work, we work on a quantitative way to measure the spectrum estimation error and its effect on DECT reconstructions. A new definition of error using a pseudo data set is introduced and the error propagation is studied. With that quantitative estimation of error, we proposed a method to calibrate the general DECT decomposition and hence increase the accuracy of the DECT reconstruction. According to our simulation experiments, the systematic bias in DECT reconstruction due to spectrum estimation error can be largely reduced. Hence, the DECT system can perform better in discriminating different materials.