Adapted random sampling patterns for accelerated MRI.

Area of Science:

• Medical Imaging
• Image Reconstruction

Background:

• Variable density random sampling patterns are crucial for accelerated MRI.
• Current design methods rely on time-consuming model-based parameter optimization.

Purpose of the Study:

• To develop a novel, practical approach for generating variable density random sampling patterns.
• To eliminate the need for parameter tuning and a priori mathematical models.

Main Methods:

• Utilized power spectra of reference datasets to generate sampling patterns.
• Validated the approach through downsampling experiments and in vivo accelerated measurements.
• Compared performance against established sampling patterns and tested generalization.

Main Results:

• The proposed method achieves image quality comparable to optimized model-based strategies.
• Demonstrated superior results compared to non-optimized model parameters.
• No random sampling pattern outperformed conventional Cartesian subsampling in the tested reconstruction strategy.

Conclusions:

• The power spectrum-based approach offers a practical alternative for designing sampling patterns.
• Image quality is competitive with established methods, especially when compared to unoptimized models.
• Further research may be needed to surpass conventional Cartesian subsampling for certain reconstruction methods.