Graph-based sinogram denoising for tomographic reconstructions.

Area of Science:

• Medical Imaging
• Image Processing
• Computational Science

Background:

• Low-dose constraints in tomographic reconstruction lead to noisy and incomplete data.
• Sinogram denoising is a critical preprocessing step for improving low-dose Computed Tomography (CT) reconstructions.
• Traditional filtering methods may not fully exploit signal structures in sinograms.

Purpose of the Study:

• To propose a novel sinogram denoising algorithm utilizing signal processing on graphs.
• To evaluate the effectiveness of the proposed graph-based denoising method for low-dose CT.
• To compare the performance of the denoised reconstructions against standard methods.

Main Methods:

• Development of a novel denoising algorithm based on graph signal processing principles.
• Application of the algorithm to sinograms from various phantoms.
• Testing the denoised sinograms with analytical filtered back-projection (FBP) and iterative reconstruction methods (ART, SIRT).

Main Results:

• The proposed graph-based denoising algorithm effectively reduces noise in sinograms.
• Denoised sinograms consistently minimized error measures across different reconstruction techniques.
• Reconstructions using denoised sinograms showed improved accuracy compared to standard reconstructions.

Conclusions:

• Graph-based sinogram denoising is a promising approach for enhancing low-dose CT.
• The novel algorithm improves the performance of both analytical and iterative reconstruction methods.
• This method offers a significant advancement in achieving accurate and less noisy CT images under data-limited conditions.