A flexible deconvolution-based reconstruction pipeline for edge illumination phase-contrast computed tomography

Differential phase contrast (DPC) computed tomography (CT), such as Edge Illumination (EI) CT, traditionally incorporates phase integration into the tomographic reconstruction process using Hilbert filtering followed by backprojection. Although this approach is fast, effective, and parameter-free, it offers limited flexibility for noise handling and precludes the use of algebraic reconstruction methods. In this work, we reformulate the phase integration step as a deconvolution problem and propose a practical pipeline that preserves the quantitativeness of phase contrast data as well as enables the use of algebraic reconstruction techniques. We demonstrate and compare several deconvolution strategies on both simulated and experimental data. Qualitative and quantitative evaluations consistently highlight the advantages of the proposed approach.