A CT reconstruction method based on constrained data fidelity range estimation.

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Parameter selection for regularization terms in CT iterative reconstruction is a significant challenge.
• Transforming reconstruction into constrained optimization is complex due to difficulties in determining and solving constraint ranges.

Purpose of the Study:

• To propose a novel CT reconstruction method addressing the parameter selection problem.
• To accurately estimate constraint ranges for data fidelity terms without parameter tuning.

Main Methods:

• Developed a CT reconstruction method based on a constrained data fidelity term.
• Utilized Taylor expansion to estimate the constraint function distribution and determine the constraint range.
• Employed Douglas-Rachford splitting (DRS) and Projection-based primal-dual algorithm (PPD) for problem splitting and subproblem solving.

Main Results:

• The proposed method accurately estimates the constrained range of data fidelity terms, ensuring reconstruction accuracy.
• Reconstruction experiments using three regularization terms demonstrated stable convergence.
• Reconstruction quality surpassed that of filtered back-projection.

Conclusions:

• The novel CT reconstruction method effectively handles parameter selection for diverse regularization terms.
• The approach provides a robust and accurate solution for CT image reconstruction, outperforming traditional methods.