Nonparametric optimization of constrained total variation for tomography reconstruction

Area of Science:

• Medical Imaging
• Computational Science

Background:

• Iterative algorithms using total-variation (TV) optimization are crucial for limited-data X-ray computed tomography (CT) reconstruction.
• Current methods like adaptive-steepest-descent (ASD) require complex, user-defined parameters for balancing TV minimization and projection onto convex sets (POCS).

Purpose of the Study:

• To develop a nonparametric method for constrained TV optimization in X-ray CT reconstruction.
• To eliminate the need for manual parameter selection in iterative CT image reconstruction.

Main Methods:

• A novel nonparametric approach for constrained total-variation optimization is proposed.
• The method automatically adjusts the step-size of TV iteration based on constraint consistency term changes.
• This avoids artificial parameters and simplifies implementation.

Main Results:

• The proposed method eliminates time-consuming parameter optimization required by traditional ASD methods.
• Experimental results on phantom data show the method's flexibility and effectiveness.
• It offers convenient implementation across various applications.

Conclusions:

• The nonparametric constrained TV optimization method offers a simplified and effective alternative for X-ray CT image reconstruction.
• It enhances usability by removing the need for empirical parameter tuning.
• The approach demonstrates significant potential for widespread adoption in medical imaging.