Structure-adaptive CBCT reconstruction using weighted total variation and Hessian penalties.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Cone-beam CT (CBCT) imaging involves radiation exposure, increasing risks of cancer and genetic defects.
• Low-dose CBCT reconstruction commonly uses statistical iterative algorithms with total variation (TV) penalty for noise reduction and edge preservation.
• TV penalty can cause staircase artifacts in smooth image regions, while Hessian penalty blurs edges.

Purpose of the Study:

• To develop a novel structure-adaptive penalty, TV-H, for improved low-dose CBCT reconstruction.
• To combine the strengths of TV and Hessian penalties to overcome their individual limitations.
• To enhance image quality by reducing noise and artifacts while preserving important image structures.

Main Methods:

• Proposed a structure-adaptive TV-H penalty that weights TV and Hessian components based on local image gradients.
• Developed a majorization-minimization (MM) algorithm with a Gauss-Seidel update strategy for efficient optimization.
• Evaluated the algorithm using simulated digital and physical phantoms.

Main Results:

• The TV-H penalty adaptively differentiates image regions (edges, gradual transitions, uniform areas).
• It successfully suppresses noise and preserves edges, similar to TV.
• It significantly improves the representation of gradual intensity transitions compared to TV and Hessian penalties.
• Visual and quantitative experiments demonstrated superior performance of TV-H over TV and Hessian penalties.

Conclusions:

• The proposed TV-H penalty offers a significant advancement in low-dose CBCT image reconstruction.
• This adaptive approach effectively balances noise suppression, edge preservation, and artifact reduction.
• TV-H penalty provides superior image quality for CBCT applications, potentially reducing radiation risks.