Acceleration of Statistical Image Reconstruction in X-ray Phase Tomography

Abstract

Phase imaging can achieve higher contrast resolution in measurement of biological soft tissues than attenuation imaging. Recently, simultaneous statistical image reconstruction methods have been reported. The method can reconstruct attenuation, scatter and phase images directly from a set of measured moiré images. We expect that the method has some potentials, e.g., reduction of noise and measurement data. However, the method takes a large computational time. To resolve this issue, we improve the convergence speed of this iterative method. Our method is based on the FBP embedded-type acceleration method developed in attenuation computed tomography. This acceleration method can reconstruct images with a small number of iterations by embedding the FBP structure in the iterative method. The original method was developed to solve the penalized weighted least square problem. We applied the method to solve the maximum likelihood problem. Using the proposed method, we reconstruct a numerical phantom and a real image of mouse bone sample measured by Talbot interferometer. We show the improvement of a convergence speed by the proposed acceleration method.