AIR: Fused Analytical and Iterative Reconstruction Method for Computed Tomography
H Gao1*, L Yang1, S Qi2, (1) Emory University, Atlanta, GA,(2) UCLA School of Medicine, Los Angeles, CATH-C-103-10 Thursday 10:30AM - 12:30PM Room: 103
Purpose: CT image reconstruction techniques can be classified roughly into two categories: analytical reconstruction method and iterative reconstruction method (IR). Analytical methods reconstruct images through analytical formulas, such as filtered backprojection (FBP) in 2D and FDK in 3D, which can be either mathematically exact or approximate. In contrast, IR is often based on the discrete forward model of X-ray transform and formulated as a minimization problem with certain image regularization, so that the reconstructed image corresponds to the minimizer obtained after iterations. Each method has its own advantages. This work is to investigate the fused analytical and iterative reconstruction method (AIR).
Methods: As a 2D example of AIR, FBP is incorporated into tensor framelet (TF) based IR. Specifically, AIR is formulated as a L1-type optimization problem. The data fidelity term consists of the discrete X-ray transform with the preconditioner that is specific to FBP. The image regularization term is via TF. The formulated minimization problem is efficiently solved through split Bregman method with GPU-accelerated X-ray transform and its adjoint.
Results: FBP, IR, and AIR were compared using experimental data that were acquired from a resolution phantom and a contrast phantom with TomoTherapy megavoltage CT. The reconstruction was performed with the regular fully sampled case (100% data) and the low-dose undersampled case (25% data). Quantitative contrast and resolution evaluations were computed, including the full width at half maximum (FWHM) and the contrast-to-noise ratio (CNR). The experimental results suggest that AIR provides better image resolution than FBP or IR, and the reconstructed image contrast from AIR is comparable to that from IR, and better than that from FBP.
Conclusion: AIR is proposed with a 2D example of FBP-based TF-regularized IR. Its superiority in terms of the resolution and the contrast for either regular or low-dose scan has been established through experimental studies.
Funding Support, Disclosures, and Conflict of Interest: This work is partially supported by NIH/NIBIB grant EB013387.