Accelerated MRI Reconstruction With Separable and Enhanced Low-Rank Hankel Regularization.

Area of Science:

• Medical Imaging
• Computational Imaging
• Biomedical Engineering

Background:

• Magnetic resonance imaging (MRI) is crucial for clinical diagnosis but is limited by long acquisition times.
• Sparse sampling accelerates MRI acquisition but necessitates robust image reconstruction from undersampled data.
• Structured low-rank methods offer robustness but suffer from high computational and memory costs due to large Hankel matrices.

Purpose of the Study:

• To develop an efficient and accurate MRI image reconstruction method from sparse k-space data.
• To overcome the computational and memory limitations of traditional structured low-rank methods.
• To enhance reconstruction quality by incorporating self-consistency and virtual coil priors.

Main Methods:

• Proposed a separable model constructing multiple small Hankel matrices from k-space rows and columns to reduce computational load.
• Introduced self-consistency of k-space and virtual coil prior to mitigate reconstruction errors caused by the separable model.
• Demonstrated the model's applicability to imaging scenarios with exponential parameter characteristics.

Main Results:

• The separable model significantly reduced computational time compared to traditional methods.
• Incorporating self-consistency and virtual coil priors improved reconstruction accuracy without substantial computational increase.
• Achieved the fastest computational speed for parameter imaging reconstruction, requiring only 4% of state-of-the-art runtime for parallel imaging.

Conclusions:

• The proposed separable structured low-rank method offers a significant speed-up for MRI reconstruction.
• The integration of self-consistency and virtual coil priors enhances image quality while maintaining computational efficiency.
• This approach provides a promising solution for accelerating MRI acquisition and improving diagnostic capabilities.