Image Quality Improvement in Sparse-View X-Ray Phase-Contrast Trimodal CBCT With Multifrequency Fringe Modulation and Iterative Methods

X-ray grating interferometer can simultaneously retrieve attenuation, refraction, and scattering signals of the imaged sample, which shows its great potential in biomedical imaging. However, fast scan of gratings and high centrifugal forces during rotating in X-ray grating-based cone-beam computed tomography (CBCT) system may lead to phase-stepping (PS) errors and grating position fluctuations, which cause reconstruction failure with PS method. What is more, the reconstructed slices usually suffer from severe artifacts under low dose conditions. Herein, we propose a multifrequency model to simulate the mechanical instability of the system and put forward novel iterative image reconstruction methods named multifrequency fringe modulated Talbot-Lau interferometry (MFM-TLI) and maximum likelihood expectation maximization (MLEM)-fast iterative shrinkage-thresholding algorithm (FISTA)-total variation (TV) to improve the image quality in sparse-view trimodal CBCT. Results reveal that the model and MFM-TLI is effective to retrieve high-quality absorption, differential phase contrast (DPC), and dark-field (DF) signals under mechanical instability, significantly outperforming the PS method. Image quality reconstructed by MLEM-FISTA-TV is improved compared to other iterative and analytical algorithms with sparse-view data. To the best of our knowledge, this is the first time multifrequency method is applied in X-ray grating interferometer and iterative MLEM-TV algorithm is evaluated on experimental sparse-view DF CBCT data. This work might provide support for future trimodal CBCT reconstruction.

Automated summary

X-ray grating interferometer can simultaneously retrieve various signals of imaged sample. However, the mechanical instability of the system can cause reconstruction failure and severe artifacts. To improve the image quality, a multifrequency model and novel iterative image reconstruction methods are proposed. The results show that the proposed methods are effective in retrieving high-quality signals and improving image quality, making them potential techniques for trimodal CBCT reconstruction.