Preliminary Investigation of Phase-Contrast CT Image Reconstruction From Limited-Angular-Range Data
Presentations
SU-H330-IePD-F8-6 (Sunday, 7/10/2022) 3:30 PM - 4:00 PM [Eastern Time (GMT-4)]
Exhibit Hall | Forum 8
Purpose: We investigate phase-contrast computed tomography (PCCT) image reconstruction from data collected over limited-angular ranges (LARs) by using directional-total-variation (DTV) algorithm, which yields reconstruction with reduced LAR artifacts.
Methods: We simulate a propagation-based (PB) PCCT imaging application with a numerical mouse phantom discretized on a 256x256 image array. We design two LAR imaging configurations: a single-LAR configuration containing a single LAR of ατ, and a two-orthogonal-LAR configuration containing two identical LARs α1 and α2 separated by 90°. We generate both noiseless and noisy data based on the PB-PCCT imaging model. The total angular ranges are selected as ατ=α1+α2=30°, 60°, 90°, 120°, 150°, and 180° for both single-LAR and two-orthogonal-LAR configurations. We then conduct image reconstruction by using DTV algorithm, which is a primal-dual algorithm solving an optimization problem with a data-L2-minimization term and image DTV constraints. For comparison, we also reconstruct images by using the FBP algorithm. We evaluate reconstruction performance by using quantitative metrics such as normalized root mean square error (nRMSE), Pearson correlation coefficient (PCC), and normalized mutual information (nMI).
Results: LAR artifacts in the FBP reconstructions are significantly reduced in the DTV reconstructions. In addition, the two-orthogonal-LAR configuration generally yields improved reconstruction over its single-LAR counterpart, especially for the noisy-data study. Quantitative analysis reveals that the reconstruction accuracy enhances as the total angular range ατ increases. Moreover, with fixed ατ, the two-orthogonal-LAR configuration leads to more accurate image reconstruction than the single-LAR configuration.
Conclusion: We have investigated DTV image reconstruction from LAR data in PB-PCCT. The DTV algorithm yields images from LAR data with reduced LAR artifacts observed in FBP images. Observations can also be made that the two-orthogonal-LAR configurations further improve reconstruction accuracy over the single-LAR configurations. This work may provide insights into the design LAR scanning configurations for potentially reducing imaging time in PCCT.
Funding Support, Disclosures, and Conflict of Interest: XP serves as the Editor-in-Chief of IEEE Transaction on Biomedical Engineering, and is a shareholder and co-founder of Clarix Imaging Co. and XPIM LLC. There is no conflict of interest.
Keywords
Not Applicable / None Entered.
Taxonomy
Contact Email