Directional view interpolation for compensation of sparse angular sampling in cone-beam CT

Area of Science:

• Medical Imaging
• Image Processing
• Computational Science

Background:

• Sparse angular sampling in flat detector cone-beam computed tomography (CT) leads to streak artifacts.
• Interpolation of additional views is a proposed solution to mitigate these artifacts.

Purpose of the Study:

• To investigate the practicality of angular interpolation for 3-D sinogram data in cone-beam CT.
• To develop and evaluate a novel shape-driven directional interpolation algorithm.

Main Methods:

• A novel shape-driven directional interpolation algorithm utilizing a structure tensor approach was developed.
• The algorithm was applied to angular interpolation of 3-D sinogram data.
• Quantitative evaluation was performed on simulated and clinical cone-beam CT datasets of the human head.

Main Results:

• The developed shape-driven directional interpolation method significantly outperformed conventional scene-based interpolation schemes.
• Directionally interpolated views effectively reduced streak artifacts and noise in cone-beam CT images.
• A minor trade-off of slight image blur was observed with the use of interpolated views.

Conclusions:

• The novel shape-driven directional interpolation algorithm is a practical and effective method for reducing artifacts in sparse-view cone-beam CT.
• This approach enhances image quality by minimizing streak artifacts and noise, making it suitable for clinical applications.
• The method offers a favorable balance between artifact reduction and image fidelity, despite a slight introduction of blur.