Parallel MR image reconstruction based on triple cycle optimization.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Image Reconstruction Algorithms

Background:

• Self-calibration parallel imaging (SC-SENSE) reconstructs images by estimating coil sensitivity matrices.
• This estimation typically requires automatic calibration signal lines (ACSL) from k-space center data.

Purpose of the Study:

• To introduce and evaluate the triple cycle optimization (TCO) method for enhancing SC-SENSE image reconstruction.
• To assess the effectiveness of TCO in reducing noise and artifacts compared to traditional methods.

Main Methods:

• The TCO method iteratively optimizes reconstructed images and sensitivity matrices.
• It uses ACSL-derived sensitivity matrices and reconstructed images as initial inputs for a stable optimization loop.
• k-space data is estimated repeatedly within the cycle to refine both the sensitivity matrix and the image.

Main Results:

• Images reconstructed with TCO exhibit lower noise and fewer artifacts than those from traditional SC-SENSE methods at equivalent sampling densities.
• Combining TCO with variable density sampling yields remarkable improvements, even at substantially reduced sampling rates.

Conclusions:

• The TCO method offers a significant advancement in MRI image reconstruction quality.
• TCO provides a robust approach for improving SC-SENSE performance, particularly in accelerated imaging scenarios.