Deep learning‐based whole‐brain B1+‐mapping at 7T

• Published in: Magnetic resonance in medicine Vol. 93; no. 4; pp. 1700 - 1711
• Main Authors: Krueger, Felix, Aigner, Christoph Stefan, Lutz, Max, Riemann, Layla Tabea, Degenhardt, Katja, Hadjikiriakos, Kimon, Zimmermann, Felix Frederik, Hammernik, Kerstin, Schulz‐Menger, Jeanette, Schaeffter, Tobias, Schmitter, Sebastian
• Format: Journal Article
• Language: English
• Published: Hoboken Wiley Subscription Services, Inc 01.04.2025; John Wiley and Sons Inc
• Subjects: 7 tesla; B1+‐mapping; Brain; Brain mapping; Brain slice preparation; Deep learning; Error analysis; Feasibility studies; Imaging Methodology; Machine learning; Neural networks; parallel transmission; Sag; Technical Note
• ISSN: 0740-3194; 1522-2594; 1522-2594
• DOI: 10.1002/mrm.30359

Purpose This study investigates the feasibility of using complex‐valued neural networks (NNs) to estimate quantitative transmit magnetic RF field (B1+) maps from multi‐slice localizer scans with different slice orientations in the human head at 7T, aiming to accelerate subject‐specific B1+‐calibration using parallel transmission (pTx). Methods Datasets containing channel‐wise B1+‐maps and corresponding multi‐slice localizers were acquired in axial, sagittal, and coronal orientation in 15 healthy subjects utilizing an eight‐channel pTx transceiver head coil. Training included five‐fold cross‐validation for four network configurations: NNcxtra$$ {\mathrm{NN}}_{\mathrm{cx}}^{\mathrm{tra}} $$ used transversal, NNcxsag$$ {\mathrm{NN}}_{\mathrm{cx}}^{\mathrm{sag}} $$ sagittal, NNcxcor$$ {\mathrm{NN}}_{\mathrm{cx}}^{\mathrm{cor}} $$ coronal data, and NNcxall$$ {\mathrm{NN}}_{\mathrm{cx}}^{\mathrm{all}} $$ was trained on all slice orientations. The resulting maps were compared to B1+‐reference scans using different quality metrics. The proposed network was applied in‐vivo at 7T in two unseen test subjects using dynamic kt‐point pulses. Results Predicted B1+‐maps demonstrated a high similarity with measured B1+‐maps across multiple orientations. The estimation matched the reference with a mean relative error in the magnitude of (2.70 ± 2.86)% and mean absolute phase difference of (6.70 ± 1.99)° for transversal, (1.82 ± 0.69)% and (4.25 ± 1.62)° for sagittal (NNcxsag$$ {\mathrm{NN}}_{\mathrm{cx}}^{\mathrm{sag}} $$), as well as (1.33 ± 0.27)% and (2.66 ± 0.60)° for coronal slices (NNcxcor$$ {\mathrm{NN}}_{\mathrm{cx}}^{\mathrm{cor}} $$) considering brain tissue. NNcxall$$ {\mathrm{NN}}_{\mathrm{cx}}^{\mathrm{all}} $$ trained on all orientations enables a robust prediction of B1+‐maps across different orientations. Achieving a homogenous excitation over the whole brain for an in‐vivo application displayed the approach's feasibility. Conclusion This study demonstrates the feasibility of utilizing complex‐valued NNs to estimate multi‐slice B1+‐maps in different slice orientations from localizer scans in the human brain at 7T.