Multiple limited-angles computed tomography reconstruction based on multi-direction total variation minimization

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Accurate computed tomography (CT) reconstruction from incomplete data is crucial for reducing radiation dose and scan time.
• Sparse and limited-angle sampling present challenges in CT reconstruction, with limited-angle sampling often yielding inaccurate images.
• The multiple limited-angles (MLA) sampling scheme offers a balance between technical feasibility and reconstruction complexity, but can introduce shading artifacts.

Purpose of the Study:

• To address shading artifacts in CT images reconstructed from MLA sampling data.
• To develop and evaluate an artifact-suppression model specifically for MLA CT reconstruction.
• To compare the performance of the proposed model against existing methods.

Main Methods:

• A novel artifact-suppression reconstruction model based on multi-direction total variation minimization (MDTVM) was designed.
• The MDTVM model was utilized to solve the optimization problem for CT image reconstruction from MLA data.
• Experiments were conducted using digital phantoms and real projection data.

Main Results:

• CT images reconstructed using the traditional total variation minimization (TVM) method exhibited shading artifacts.
• The proposed MDTVM method demonstrated superior performance in suppressing these directional shading artifacts compared to TVM.
• Experimental results confirmed the effectiveness of MDTVM on both simulated and real-world data.

Conclusions:

• MDTVM is an effective technique for suppressing shading artifacts in CT images reconstructed from MLA sampling.
• The developed method offers improved image quality for CT scans utilizing MLA sampling strategies.
• This work provides a valuable solution for enhancing the accuracy of CT reconstruction in dose-reduced scenarios.