Regularized sensitivity encoding (SENSE) reconstruction using Bregman iterations.

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Image Reconstruction

Background:

• Parallel imaging techniques like sensitivity encoding (SENSE) enhance MRI speed but face signal-to-noise ratio (SNR) degradation due to ill-conditioning, especially at high acceleration factors.
• Existing regularization methods mitigate ill-conditioning but can introduce artifacts at high accelerations due to data inconsistency from heavy regularization.

Purpose of the Study:

• To introduce and evaluate Bregman iteration as a novel regularization method for SENSE reconstruction in parallel MRI.
• To address the limitations of fixed regularization functions in existing methods, aiming to reduce artifacts and improve image quality.

Main Methods:

• The proposed method utilizes Bregman iteration, which adaptively updates the regularization function based on Bregman distance at each iteration.
• A discrepancy principle is employed as the stopping criterion for the iterative reconstruction process.

Main Results:

• Bregman iteration successfully reduces noise and aliasing artifacts in SENSE reconstructions.
• The method preserves sharp edges, outperforming Tikhonov regularization, and recovers fine structures, surpassing total variation (TV) regularization.
• Adaptive regularization prevents the excessive data inconsistency seen with heavy fixed regularization.

Conclusions:

• Bregman iteration offers superior performance for SENSE regularization in parallel MRI compared to existing methods.
• This adaptive approach effectively balances artifact reduction, noise suppression, and preservation of image details, particularly at high acceleration factors.