Bias-Corrected Spectrum Decomposition for X-Ray Computed Tomography (ct): Theory and Method

Hewei Gao,Geng Fu,Yannan Jin,Peter M. Edic

DOI: https://doi.org/10.1109/nssmic.2016.8069482

2016-01-01

Abstract:Modern diagnostic X-ray computed tomography (CT) usually provides a limited number of standard X-ray tube operating voltages (kVps) that are well calibrated and optimized. In certain scenarios, however, a spectrum using a non-standard kVp may have potential of achieving improved image quality and dose efficiency. For a CT scan using a non-standard kVp, estimating an effective spectrum for calibration is a problem to be solved. An efficient but not exact method of spectrum estimation in practice is to decompose the non-standard spectrum into a linear combination of spectra generated using standard kVps, e.g. spectrum decomposition. This approach works well when the spectrum generated at a non-standard kVp is reasonably close to that generated by a standard kVp. If it is too far away, the residual error (systematic bias) could be significant enough to cause artifact/inaccuracy in CT images. To improve spectrum decomposition, we present a bias-corrected spectrum decomposition approach. A lookup table (LUT) is established in advance to store the spectrum estimation errors using a set of known spectra generated using known kVps from simulation that do not have to be well calibrated. A reasonable and key assumption here is that the spectral estimation errors in the LUT generation would be close to that of actual spectrum estimation errors. Simulations of a dual-energy CT scan at non-standard kVps demonstrate the effectiveness of this approach, leading to more quantitative CT images.