Fast parallel MR image reconstruction via B1-based, adaptive restart, iterative soft thresholding algorithms...

Area of Science:

• Medical Imaging
• Applied Mathematics
• Computer Science

Background:

• Sparsity-promoting regularization aids compressed sensing in parallel MRI, reducing scan time while maintaining image quality.
• Variable splitting algorithms are state-of-the-art for SENSE-type MR image reconstruction but suffer from complex parameter tuning.
• Majorize-minimize algorithms are slow for shift-variant applications like SENSE reconstruction due to loose Lipschitz constant bounds.

Purpose of the Study:

• To bridge the gap between Lipschitz constants and shift-variant aspects in SENSE-type MR imaging.
• To introduce a novel majorize-minimize method (BARISTA) for faster and more accessible MR image reconstruction.
• To simplify parameter tuning compared to existing variable splitting algorithms.

Main Methods:

• Introduction of majorizing matrices within the range of the regularizer matrix.
• Development of BARISTA, a majorize-minimize method incorporating majorizing matrices.
• Application of momentum acceleration and adaptive momentum restarting to the BARISTA method.

Main Results:

• BARISTA demonstrates faster convergence than state-of-the-art variable splitting algorithms when combined with acceleration techniques.
• The proposed methods simplify parameter tuning, utilizing unitless convergence tolerances.
• BARISTA offers a more accessible and efficient approach to SENSE-type MR image reconstruction.

Conclusions:

• BARISTA effectively addresses the limitations of existing methods in SENSE-type MR image reconstruction.
• The simplified parameter tuning of BARISTA facilitates wider adoption of advanced MR imaging techniques.
• This work advances the field of accelerated MRI by improving reconstruction speed and ease of use.