Non-convex compressed sensing CT reconstruction based on tensor discrete Fourier slice theorem

Area of Science:

• Medical Imaging
• Computational Imaging
• Radiology

Background:

• X-ray computed tomography (CT) offers high resolution and fast imaging but requires high radiation doses.
• Increased radiation dose elevates cancer risk, especially in younger patients.
• Reducing radiation dose while maintaining image quality is a critical clinical goal.

Purpose of the Study:

• To develop a novel framework combining compressed sensing (CS) theory with X-ray CT.
• To reduce the number of X-ray measurements needed for high-quality CT scans.
• To enhance image reconstruction performance in low-dose CT protocols.

Main Methods:

• Utilized the tensor discrete Fourier slice theorem for data mapping.
• Employed nonuniform random density sampling of Fourier coefficients with uniform projection angle sampling.
• Applied a non-convex compressed sensing model to enhance data sparsity.

Main Results:

• The novel framework enables feasible nonuniform random density sampling.
• The non-convex CS model significantly reduces the required number of measurements.
• Demonstrated superior image reconstruction performance with limited projection data compared to convex models.

Conclusions:

• The presented non-convex CS framework effectively reduces radiation dose in X-ray CT.
• This approach offers a promising solution for improving patient safety in CT imaging.
• Further research can explore optimization of CS parameters for clinical translation.