A New k-Space Model for Non-Cartesian Fourier Imaging

Area of Science:

• Medical Imaging
• Signal Processing
• Computational Science

Background:

• Model-based approaches are popular for Fourier imaging reconstruction, incorporating physical constraints and machine learning priors.
• The conventional voxel-based model, while widely used, suffers from high computational costs, slow convergence, and artifacts.
• Existing methods face limitations like undesirable approximation, wrap-around, and nullspace issues.

Purpose of the Study:

• To reexamine the limitations of the traditional voxel-based model for Fourier imaging reconstruction.
• To propose a new Fourier-domain basis expansion model to overcome existing and newly identified issues.
• To improve image quality and computational efficiency in Fourier imaging reconstruction.

Main Methods:

• Developed a novel Fourier-domain basis expansion model for image reconstruction.
• Compared the new model against the standard image-domain voxel-based approach.
• Evaluated the model's performance in non-Cartesian Magnetic Resonance Imaging (MRI) reconstruction.

Main Results:

• The proposed Fourier-domain model demonstrates improved resilience to limitations of the voxel-based approach.
• Results show enhanced image quality with reduced artifacts.
• The new model offers reduced computational complexity, leading to faster computations and improved convergence.

Conclusions:

• The Fourier-domain basis expansion model represents a significant advancement over traditional voxel-based methods.
• This new approach enhances both image quality and computational efficiency in Fourier imaging.
• The findings are particularly relevant for applications like non-Cartesian MRI reconstruction.