Towards Quantitative Flat-Detector CT Using Advanced Scatter Correction

Abstract

Quantitative imaging using large area flatdetector CT (FD-CT) scanners is mainly disturbed by the high fraction of scattered radiation, which results in increased scatter artifact content and falsifies the CT values. Pure MCbased methods for scatter correction need several iterations for good results due to insufficient image quality of the initial voxel volume used for the calculations. We applied an empirical cupping correction (ECC) which uses a set of precalculated polynomial correction coefficients for cupping removal. The ECC-corrected raw data were used to improve the initial reconstructed volume used for the MC scatter estimation. Detector elements were binned and the angular increment between the estimated projections was enlarged to speed up the calculation. After spatial and angular interpolation, the scatter estimates were subtracted from original data and were combined with an analytical beam hardening correction to remove remaining beam hardening-induced cupping. The correction method was applied to simulated head and body datasets of patient CT scans for which scatter was generated by MC calculations. For comparison, reference reconstructions of scatter and beam-hardening-free projections were generated. The root-mean-square deviation (RMSD) between both the corrected and uncorrected images relative to the reference was used to calculate the scatter correction quality factor Q (Q = 1 for the ideal image, Q=0 for the uncorrected image). The method was also applied to anthropomorphic phantom data measured on a C-arm FD-CT system (Artis zeego, Siemens Healthcare, Forchheim, Germany) using a large flat detector of 40×30cm. The proposed correction method on the simulated data provided a Q of 0.91 due to significant reduction of scatter artifacts. The number of iterations for scatter correction was reduced from 4 to 1 compared to a pure iterative MC approach data without ECC for equivalent artifact reduction (15 s per iteration on a standard PC). Correction of the measured images of an anthropomorphic phantom led to an increase of contrast and to a general reduction of cupping and streaks.