VVBP-Tensor in the FBP Algorithm: Its Properties and Application in Low-Dose CT Reconstruction.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Filtered Backprojection (FBP) is the standard for CT image reconstruction.
• Lower radiation doses challenge FBP's image quality.
• Existing FBP improvements focus on preprocessing, filtering, or postprocessing.

Purpose of the Study:

• To analyze stacked view-by-view backprojections (VVBP-Tensor) within the FBP algorithm.
• To develop a novel denoising algorithm for low-mAs CT imaging.
• To evaluate the effectiveness of the proposed method against current techniques.

Main Methods:

• Introduced a sorting operation to the VVBP-Tensor to reveal object structures.
• Analyzed VVBP-Tensor properties: self-similarity, sparsity, and noise statistics.
• Developed the VVBP-tSVD algorithm using tensor singular value decomposition for denoising.

Main Results:

• Sorted VVBP-Tensor exhibits object-similar structures and inter-slice correlations.
• VVBP-tSVD demonstrated superior noise reduction compared to preprocessing, postprocessing, and iterative reconstruction methods.
• Effective performance was validated on physical phantoms and clinical data at low mAs levels.

Conclusions:

• VVBP-Tensor is a viable target for enhancing FBP reconstruction quality.
• VVBP-tSVD is an effective algorithm for noise reduction in low-mAs CT imaging.
• This work offers a new perspective on FBP algorithm optimization.