A novel iterative reconstruction method for dual-energy computed tomography based on polychromatic forward-projection calibration

Dual-energy computed tomography (DECT) scans the measured object using two different detected X-ray spectra. The acquired polychromatic projections can be used to calculate the density and composition distribution of the scanned object. Due to the non-linear relationship between the polychromatic projections and the images to be reconstructed, precise DECT image reconstruction is generally a difficult task. This paper proposes an iterative reconstruction method for DECT based on polychromatic forward-projection calibration. The method has a fast convergence rate and fits for any scanning configurations commonly used, including both geometrically consistent and inconsistent measurements for different spectra. In contrast to other iterative reconstruction algorithms, the proposed method requires neither knowledge of the X-ray spectrum nor of the attenuation coefficients. The proposed method also has a high degree of parallelism and is suitable for acceleration on graphic processing units (GPUs). The method is validated with numerical experiments from the simulated noise-free and noisy, consistent and inconsistent polychromatic projection data of a FORBILD thorax phantom. The reconstructed linear attenuation coefficients of the test phantom are in excellent agreement with the theoretical ones, and beam hardening artefacts can be completely eliminated.