A new method for cupping and scatter precorrection for flat detector CT

Abstract

Scatter and beam hardening are prominent artifacts in x-ray CT. Currently, there is no precorrection method that inherently accounts for tube voltage modulation and shaped prefiltration. We generalized a method for self-calibration based on binary tomography of homogeneous objects [1] to use this information to preprocess scans of other, non-binary objects, e.g. to reduce artifacts in medical CT applications. Further on we extended the method to handle not only beam hardening but also scatter and to allow for detector pixel-specific precorrections. This implies that our calibration technique handles varying tube voltage and shaped prefiltration. We propose a method that models the beam hardening correction by using a rational function while the scatter component is modeled using the pep-model. A smoothness constraint is applied to the parameter space to regularize the underdetermined system of non-linear equations. The parameters determined are then used to precorrect CT scans. Our algorithm was evaluated using simulated data of a flat panel cone-beam CT scanner with tube voltage variation and bow-tie prefiltration and real data of a flat detector cone-beam CT scanner. In simulation studies our correction model proved to be nearly perfect and the algorithm showed its abilities by correcting the beam hardening and scatter effects. Reconstructions of measured data showed significantly less artifacts than the standard reconstruction.