Magnetic resonance fingerprinting (MRF) can be used for a fast estimation of quantitative parameters using standard MRI hardware. To date, this method has been demonstrated at 1.5T [1], 3T [2] and 7T [3], with fast gradient echo sequences based on steady state free precession (SSFP). It has recently been shown that B1 inhomogeneity confounds accurate measurements, particularly of T2 [4] using a Cartesian approach for small animal imaging at 4.7T.

Here, we demonstrate a new MRF method including B1+ estimation in a spiral sequence optimised for human brain imaging at 7T. While previous methods have used externally-measured B1+ values to correct the MRF parameter maps [3], we aimed to estimate the B1+ homogeneity directly from the MRF acquisition. We acquired T1, T2 and B1 simultaneously by tailoring the acquisition scheme of gradient-spoiled SSFP spiral MRF and accounting for B1+ effects in the signal model.

To make the sequence more sensitive to the B1+ field, we introduced abrupt changes in flip angle at the end of the MRF train, following [4]. We used alternating blocks of 7 pulses of flip angle 90° followed by 7 pulses of flip angle 0° to exploit the oscillatory behaviour of the transient response, dependent on B1+ but not on T2 [4]. We obtained brain images of healthy subjects using SSFP MRF, acquiring 24 spiral interleaves on a 5mm thick slice (FOV 30 cm, 256x256 matrix) as described in [2], using a Discovery MR950 7T MRI system (GE Healthcare, Milwaukee, WI, USA) equipped with a 2-channel transmit / 32-channel receive head coil (Nova Medical, Wilmington, MA, USA).

The dictionary included T1 values ranging from 100ms to 4s, T2 values ranging from 10 ms to 400 ms and B1+ values ranging from 15% to 150%. Pattern matching was performed as described in [1]. B1+ values derived from our new protocol were compared with maps derived externally using a B1+ mapping method based on the Bloch-Siegert shift (BS) [5] on slices with the same geometry (TR=100 ms, TE=13 ms, Flip Angle=30°, 4ms Fermi pulse, 2Hz off-resonance).

Including a parameter for B1+ in the original MRF scheme from [2] produced substantial errors in the estimation of B1+ and T2 with errors up to 200%. Our new sequence including abrupt changes in flip angle could discriminate between B1+ and T2 effects and generated B1-corrected maps. Phantom results showed a good agreement between BS B1+ mapping and MRF (see Figure 2). Representative data in a healthy volunteer are shown in Figure 3.

We have presented a new protocol for estimating quantitative parameters at 7T. The main novelty of our method is the simultaneous estimation of the transmission B1+ field together with T1 and T2 for 7T neuroimaging. Further work may reduce the number of timeframes required for the acquisition. Such optimisation should take into account both the sensitivity to the quantitative parameters and the influence of aliasing when undersampling.

In conclusion, the method presented here can be used to estimate T1 and T2 maps in the human brain at 7T. The B1+ field can be estimated simultaneously and its effects removed from the estimates directly, without using an external acquisition.