Pre- and post-reconstruction digital image processing solutions for computed tomography with spectral photon counting detectors

Abstract

Due to several issues still affecting pixelated photon-counting detector technology, such as e.g. inhomogeneities in the energy threshold calibration, polarization effects of high-Z semiconductor sensors and limitations of the hardware solutions aimed at compensating for the charge sharing issue, several artifacts corrupt the reconstructed images when adopting these detectors for computed tomography (CT). This article presents and discusses the most recurrent artifacts occurring in the raw images acquired with the innovative CdTe Pixirad-1/Pixie-III detector. By considering an experimental CT dataset of a suitable test object, digital image processing solutions for spectral CT imaging are here suggested and optimized. The proposed solutions consist of a pre-reconstruction despeckle filtering of each acquired projection and a post-reconstruction ring artifacts removal. The effects of these filters are quantitatively analyzed in terms of spatial resolution and signal-to-noise ratio. Finally, the suitability of the optimized image processing procedure is validated with two practical CT applications in which a plastinated small animal specimen is considered.