Maximum Likelihood CT reconstruction from material-decomposed sinograms using fisher information

Recently, a K-edge imaging technique for energy-resolving photon-counting detectors based on Maximum-Likelihood estimation for material separation in the projection domain with subsequent image formation using filtered back projection (FBP) was presented. Despite its computational load, use of statistical reconstruction techniques for image reconstruction from material-specific sinograms is favourable as they feature superior signal to noise ratio (SNR). This work presents an estimation of the noise in material-decomposed sinograms using Fisher information in order to enable statistical image reconstruction. Simulations demonstrate that a gain of two in SNR compared to FBP reconstruction can be obtained. This improvement may be used for x-ray dose reduction but is in particular desirable to limit count rates and, thus, which is a significant challenge with current photon-counting detectors.