Self-calibrated B₁ ⁺ and B₀ Field Inhomogeneities Estimation (SAFE) in 3D Magnetic Resonance Fingerprinting

Area of Science:

• Biomedical Imaging
• Medical Physics
• Artificial Intelligence in Medicine

Background:

• Magnetic Resonance Fingerprinting (MRF) enables quantitative imaging but is sensitive to B1+ and B0 field inhomogeneities.
• Accurate field mapping is crucial for reliable MRF-based tissue parameter quantification.
• Existing field calibration methods can be time-consuming and require dedicated hardware.

Purpose of the Study:

• To develop SAFE, a self-calibrated framework for estimating B1+ and B0 field inhomogeneities directly from conventional MRF acquisitions.
• To improve the quantification accuracy of MRF by correcting for field imperfections.
• To enable calibration-free field mapping within standard MRF protocols.

Main Methods:

• SAFE employs a two-step approach using physics-informed image markers derived from MRF data.
• A deep learning network maps these markers to quantitative B1+ and B0 field maps.
• The framework was trained and validated on a multi-site, multi-vendor MRF brain dataset (N=358) at 3T.

Main Results:

• SAFE achieved <3% normalized-root-mean-square-error against gold-standard field calibration scans (N=32).
• Effective correction of T1 and T2 biases was demonstrated in healthy volunteers and patients.
• Tissue quantification accuracy was validated on a large patient cohort (N=86).
• SAFE showed adaptability to unseen MRF sequences without additional training data.

Conclusions:

• The SAFE framework combines physics-informed markers and deep learning for calibration-free B1+ and B0 field inhomogeneity estimation.
• This approach is applicable to whole-brain MRF at 3T.
• SAFE enhances the robustness and accuracy of quantitative MRF imaging.