Compressed-sensing-based content-driven hierarchical reconstruction: Theory and application to C-arm cone-beam...

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Interventional C-arm systems face significant subsampling artifacts in cone-beam CT (CBCT) due to slow rotation and low detector frame rates.
• Analytical reconstruction methods often struggle with these artifacts, impacting image quality in clinical applications.

Purpose of the Study:

• To propose a novel content-driven hierarchical reconstruction method for C-arm CBCT.
• To address and reduce subsampling artifacts using compressed sensing principles.

Main Methods:

• A hierarchical reconstruction strategy prioritizing salient features to mitigate artifact contamination.
• Implementation of various sparsity-promoting schemes including orthogonal matching pursuit, L1-norm minimization, total variation (TV), and nonlinear diffusion (NLD).
• Utilized penalized iterative filtered backprojection algorithms and the alternating direction method of multipliers for optimization.

Main Results:

• Demonstrated fast convergence to accurate solutions for TV-constrained minimization problems on simulated data.
• Clinical C-arm CBCT data showed significant streak artifact reduction and improved image quality using both TV and NLD constraints.

Conclusions:

• A flexible compressed sensing algorithmic framework was developed for C-arm CBCT reconstruction.
• The proposed method effectively handles images not well-approximated by piecewise constant functions, enhancing clinical utility.