A compressed sensing based approach on Discrete Algebraic Reconstruction Technique.

Area of Science:

• Medical Imaging
• Computational Science
• Image Reconstruction

Background:

• Discrete tomography (DT) offers improved reconstruction with fewer projections compared to continuous methods.
• The Discrete Algebraic Reconstruction Technique (DART) leverages prior knowledge of object densities for reconstruction.
• Existing DART methods can be further optimized for enhanced accuracy and robustness.

Purpose of the Study:

• To improve the Discrete Algebraic Reconstruction Technique (DART) for discrete tomography.
• To enhance DART by integrating total variation minimization and a novel segmentation procedure.
• To evaluate the performance of the enhanced DART algorithm under challenging imaging conditions.

Main Methods:

• Combined DART with an initial total variation minimization (TvMin) phase for a refined initial guess.
• Incorporated a segmentation procedure to estimate threshold values, minimizing projection error and total variation (TV) simultaneously.
• Conducted simulation experiments to compare the enhanced DART with the original DART.

Main Results:

• The enhanced DART algorithm demonstrated improved accuracy and robustness compared to the original DART.
• Performance was validated under conditions of limited projections, limited view, and noisy projections.
• The combined approach effectively minimized both projection error and total variation.

Conclusions:

• The proposed enhanced DART method offers a more accurate and robust solution for discrete tomography reconstruction.
• The integration of TvMin and adaptive segmentation significantly improves DART's performance, especially in limited data scenarios.
• This technique shows promise for applications requiring high-fidelity reconstruction from sparse or noisy tomographic data.