Statistical Reconstruction for Quantitative CT Applications

Abstract

This paper summarizes considerations in developing statistical reconstruction algorithms for polyenergetic X-ray CT. The algorithms are based on Poisson statistics and polyenergetic X-ray attenuation physics and object models. In single-kVp scans, object models enable estimates of the contributions of bone and soft tissue at every pixel, based on prior assumptions about the tissue properties. In dual-kVp scans, one can estimate water and bone images independently. Preliminary results with fan-beam data from two cone beam systems show better accuracy for iterative methods over FBP.