Motion guided Spatiotemporal Sparsity for high quality 4D-CBCT reconstruction

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Computational Imaging

Background:

• Respiratory motion significantly degrades conventional cone-beam CT (CBCT) image quality, hindering accurate tumor delineation in radiotherapy.
• Four-dimensional CBCT (4D-CBCT) aims to capture organ and tumor motion but is challenged by slow rotation speeds, leading to undersampled data and reconstruction artifacts.

Purpose of the Study:

• To develop a novel framework for reconstructing high-quality 4D-CBCT from undersampled measurements.
• To address noise and streak artifacts inherent in conventional 4D-CBCT reconstruction algorithms.

Main Methods:

• Proposed a Motion guided Spatiotemporal Sparsity (MgSS) algorithm for 4D-CBCT reconstruction.
• Divided CBCT images into 3D blocks (cubes) and tracked their motion across phases using motion field vectors.
• Applied regional spatiotemporal sparsity analysis using higher-order singular value decomposition (HOSVD) on tracked cubes and incorporated it into an image reconstruction cost function.

Main Results:

• The MgSS algorithm demonstrated improved 4D-CBCT image quality in both phantom simulations and real patient data.
• Significantly reduced noise and streak artifacts compared to conventional reconstruction algorithms.
• Enabled more accurate representation of tumor and organ motion.

Conclusions:

• The MgSS algorithm offers a robust solution for reconstructing high-fidelity 4D-CBCT from undersampled data.
• This technique enhances image quality, reduces artifacts, and improves the potential for accurate motion management in radiotherapy.
• MgSS represents a significant advancement in on-board CBCT imaging for radiation oncology.