Comparison of parameter optimization methods for quantitative susceptibility mapping.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Science

Background:

• Quantitative Susceptibility Mapping (QSM) relies on minimizing functionals with data fidelity and regularization terms.
• A critical weighting parameter balances these terms, but finding the optimal balance is challenging.
• Existing methods like L-curve maximum curvature can be slow, unreliable, and lead to over-regularization.

Purpose of the Study:

• To develop and evaluate novel methods for optimizing the balance between data fidelity and regularization in QSM.
• To address the limitations of traditional parameter selection techniques in QSM reconstruction.

Main Methods:

• Proposed two alternative approaches: L-curve log-log domain inflection point search and frequency component comparison of QSM reconstructions.
• Compared these novel methods against conventional L-curve and U-curve approaches.

Main Results:

• The proposed methods showed better correlation with RMS error, high-frequency error norm, and structural similarity metrics.
• The L-curve inflection point method resulted in less over-regularization and lower errors compared to traditional methods.
• Frequency analysis produced more visually appealing QSM results, albeit with higher RMS error.

Conclusions:

• The developed methods offer a robust framework for parameter optimization in QSM reconstruction using variational penalties.
• The L-curve inflection point search yielded near-optimal results for standard QSM settings.
• Frequency analysis, potentially enhanced with correction factors and fast search algorithms, shows promise for broader QSM applications.