Compressed sensing based cone-beam computed tomography reconstruction with a first-order method.

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Cone-beam computed tomography (CBCT) is crucial for medical imaging.
• Image reconstruction from undersampled or noisy data presents a significant challenge.
• Reducing radiation dose in CBCT is a primary clinical and research goal.

Purpose of the Study:

• To develop a robust method for reconstructing high-quality CBCT images from undersampled and noisy projection data.
• To address the limitations of traditional algorithms in handling sparse and noisy CBCT measurements.
• To investigate the application of compressed sensing techniques for dose reduction in CBCT.

Main Methods:

• Reconstruction formulated as a compressed sensing problem using l1 norm minimization.
• Statistically weighted least-squares constrained by CBCT projection data.
• Incorporation of noise characteristics to weight projection measurements.
• Total-variation norm minimization solved with Nesterov's first-order method.

Main Results:

• The proposed method achieves fast convergence with low memory requirements.
• High-quality CBCT images are successfully reconstructed from sparse and noisy projection data.
• Demonstrated effectiveness through digital and experimental phantom studies.
• Sparse sampling and reduced X-ray tube current (noisy data) correlate with dose reduction.

Conclusions:

• Compressed sensing significantly outperforms traditional algorithms for sparse and noisy CBCT data.
• The developed method enables high-quality image reconstruction with reduced radiation exposure.
• This approach offers a promising solution for dose-efficient CBCT imaging.