Accelerated barrier optimization compressed sensing (ABOCS) reconstruction for cone-beam CT: phantom studies.

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Compressed Sensing (CS) enables accurate CT reconstruction from undersampled data, crucial for dose reduction in radiation therapy (e.g., CBCT).
• Conventional CS reconstruction methods are often computationally inefficient or require inconsistent parameter tuning, limiting clinical adoption.

Purpose of the Study:

• To develop a computationally efficient optimization algorithm for CS-based CT image reconstruction.
• To overcome the limitations of existing CS reconstruction methods, specifically addressing computational inefficiency and parameter tuning inconsistencies.

Main Methods:

• Proposed Accelerated Barrier Optimization for Compressed Sensing (ABOCS) algorithm.
• Utilized a TV optimization framework with a data fidelity constraint, converted using a logarithmic barrier method.
• Employed gradient projection with adaptive Barzilai-Borwein step-size selection for accelerated convergence.

Main Results:

• ABOCS directly estimates data fidelity tolerance, ensuring consistent reconstruction across different noise levels and objects.
• Achieved significant computational efficiency improvement, outperforming ASD-POCS by over an order of magnitude in MATLAB.
• Demonstrated consistent performance using the same parameters in digital and physical phantom studies.

Conclusions:

• ABOCS offers a fast-converging and parameter-consistent solution for CBCT reconstruction.
• The proposed method addresses key limitations of existing CS-based algorithms, showing promise for clinical applications.
• Phantom studies validate the attractive features of ABOCS for efficient and reliable image reconstruction.