Quantitative DCE-MRI or fast T₁ mapping using variable flip-angle approaches requires reliable information on the B₁⁺ field, especially at inhomogeneous fields. B₁⁺ mapping methods are generally limited in terms of SNR or time efficiency. The present work explores the feasibility of using one generic coil-specific B₁⁺ distribution (coil template) to estimate B₁⁺ fields for breast imaging at 7T, while circumventing B₁⁺ mapping.

FDTD simulations for five healthy female volunteers (1-5) representing a wide range in breast anatomies presented in previous work were used to assess theoretical inter-subject differences in B₁⁺.¹ Simulated B₁⁺ fields were aligned on coil position using multi-resolution intensity-based rigid image registration using the mutual information similarity metric and B-spline interpolation in elastix.² One volunteer’s simulation was excluded, since data was not available for full inclusion of the coil. To rule out differences in absolute RF power settings in constructing the template, all simulations were multiplied by a scaling factor relative to volunteer 5 (arbitrarily chosen): $$$Scale(i)=median_{all\ r}(\frac{Simulation_5(r)}{Simulation_i(r)})$$$, where i is the volunteer number and r the position in the simulation. The average of the remaining four scaled simulations was computed and served as a coil-specific B₁⁺ template.

Using the same coil used in the simulations,³ a 3D B1 map (DREAM⁴), 3D T₁-weighted gradient echo images (FFE) at four flip-angles (2°, 3°, 11° and 17°) and a 2D single-slice Look-Locker measurement using a water-selective adiabatic inversion pulse were obtained unilaterally from three healthy female volunteers (6-8) aged 23 - 30.⁵ To assess the proposed method’s feasibility in vivo, rigid registration was applied between the measured B₁⁺ map of the subject and the coil-specific B₁⁺ template, using masks to exclude regions where the mapping failed and where the simulated B₁⁺ < 20% or > 100% of the nominal angle. The registered template was scaled using: $$$Scale=median_{r\in M}(\frac{Map(r)}{Template(r)})$$$, where M represents Map>70%.

As a preliminary analysis, T₁ maps were calculated for these volunteers using the DESPOT1 method, without B₁⁺ information or incorporating B₁⁺ information from either the measured map or the registered template.^{6, 7} An independent reference T₁ map was calculated from the Look-Locker real channel data by a three-parameter non-linear least-squares fit.

Figure 1 shows the range in breast anatomies included in the simulations by means of their fat-suppressed T₁w scans.

Figure 2 shows a comparison between the designed coil-specific template and subject-specific simulations. Volunteer 2 showed least agreement; the mean difference was 0.87% of the nominal flip angle with a standard deviation of 3.92%.

Figure 3 shows a comparison between the measured B₁⁺ map and the scaled and registered template. Volunteer 7 showed least agreement; the mean difference was 2.87% of the nominal flip angle with a standard deviation of 11.94%.

As a visual illustration, Figure 4 shows the T₁ maps generated for volunteer 8 with the DESPOT1 method using no B₁⁺ information, the measured B₁⁺ map and the registered template respectively, alongside the Look-Locker based reference.

The agreement between B₁⁺ simulations of different volunteers is better than the accuracy of popular B₁⁺ mapping methods,⁸ which indicates that B₁⁺ inhomogeneity in the breast at 7T when using a quadrature setup is mostly determined by the transmit coil, and that variations between subjects contribute to this only slightly. This finding can most likely be explained by the dielectric properties of fat, which have a modest effect on the RF wavelength and, by extent, cause smooth B₁⁺ fields in fat-dominated areas such as the breast. This opens up the possibility to estimate B₁⁺ using a coil-specific template, provided this template can be reliably positioned and scaled.

When the position and scaling of the template are obtained directly from a B₁⁺ measurement (by rigid registration and scaling as described above), the accuracy of the registered and scaled template is excellent, although the standard deviation of the difference with the measurement is higher mainly due to the noisy nature of the B₁⁺ map.

Preliminary results of T₁ mapping with DESPOT1 show that both T₁ maps that incorporate B₁⁺ information, either from the DREAM map or the registered template, show fair agreement with the Look-Locker based reference.

Simulations show that inter-subject differences in B₁⁺ fields of the breast at 7T are comparable to the accuracy of popular B₁⁺ mapping methods reported in literature when using a quadrature setup. This opens up the possibility of using one generic B₁⁺ map, instead of subject-specific mapping.