MRI at ultra-high field (UHF) has the major advantage of providing an increased signal-to-noise ratio (SNR) and different contrast compared to lower field strengths. However, employing the sequences at such high field strength is challenging due to specific absorption rate (SAR) restrictions and detrimental physical effects such as B1 and B0 inhomogeneities amongst others. In this work, we suggest to use the recently proposed trueCISS method [1] to acquire an on-resonant bSSFP image at ultra-high field and to synthetically generate contrasts that cannot be practically obtained at UHF due to physical or safety constraints.

After written consent was obtained, a healthy volunteer was imaged on a 9.4T research system (Siemens, Erlangen, Germany) with a maximum gradient strength and slew rate of 70 mT/m and 200 T/m/s respectively, using a custom-built head coil [2] for RF transmission and reception (16 transmit/31 receive channels) in combination with a recently proposed highly undersampled phase-cycled 3D balanced Steady-State Free Precession (bSSFP) prototype trueCISS sequence [1] (1x1x1mm³, TR 4.26 ms, TE 2.13 ms, flip-angle 15°, 16 phases $$$\phi$$$ = 0°, 45°, 90°, ..., 315°, 8-fold undersampling, TA 2:40). First, the phase-cycled images were reconstructed using a sparse iterative reconstruction. Subsequently, the bSSFP signal-model (cf. equation below with $$$M_0$$$ equilibrium magnetization, $$$\Lambda$$$ relaxation time ratio $$$T_1/T_2$$$, $$$\alpha$$$ flip-angle, $$$\Delta\phi$$$ local phase offset relating to field inhomogeneity) was fitted voxel-wise onto these images, effectively estimating three parameter maps: $$$M_0$$$, $$$\Lambda$$$, and $$$\Delta\phi$$$. The resulting parameter maps were used to synthesize an on-resonant bSSFP signal image by applying the forward signal model with $$$\phi - \Delta\phi$$$ = 0°.

$$M = M_0 \left\lvert\frac{2\sin\alpha \cos\left(\frac{\phi - \Delta\phi}{2}\right)}{1 + \cos\alpha + 2\cos\left(\phi - \Delta\phi\right) + \left(4\Lambda - \cos^2\left(\phi-\Delta\phi\right)\sin^2\left(\frac{\alpha}{2}\right)\right)}\right\rvert$$

For comparison, Fig. 1a shows two slices of a reconstructed on-resonant trueCISS image for $$$\alpha$$$ = 15° inside a human brain measured at 3T [1]. Fig. 1b shows two similarly located slices in an acquisition done at 9.4T with $$$\alpha$$$ = 15°. The ultra-high field images show improved contrast in the red nucleus (blue arrow), substantia nigra (red arrow) and globus pallidus (white arrows) compared to the 3T image. Due to severe SAR limitations when scanning at ultra-high field strengths, it is in practice impossible to measure with high flip angles (ie. $$$\alpha$$$ > 15°) and short TR. In the present case, however, since the flip angle is an independent parameter within the signal model and all other parameters are known from the fitting procedure, it is possible to synthetically generate an on-resonant bSSFP image of any flip angle. To illustrate this, a synthetic on-resonant bSSFP image for a flip angle of 70° was generated and is presented in Fig. 1c for the same slices as before. Lastly, Fig. 1d shows the same two slices reconstructed using the maximum bSSFP signal computed voxel-wise: $$$M_\max\lvert_{\theta = \theta_{opt}} \approx \frac{1}{2}M_0\Lambda^{-1/2}$$$ where $$$\theta_{opt} \approx \cos^{-1}\left(\frac{\Lambda-1}{\Lambda+1}\right) $$$ [3].

It should be noted that the trueCISS acquisition presented above only requires about the same scan time as a single fully-sampled bSSFP image and thus not only offers high quality images in a time-efficient manner, but also delivers genuine bSSFP contrast images that can be used as basis for further quantification methods. One issue with the current approach is the stability of the fitting procedure due to the low number of phase-cycles acquired. While increasing the number of phase-cycles greatly improves the robustness of the method, it is time-consuming but can be partially compensated by further increasing the undersampling of K-space.

We have shown that trueCISS imaging is able to provide artifact-free bSSFP images at ultra-high field strength. Moreover, the forward signal model can be used to generate synthetic images, which would otherwise be impossible to measure directly on the scanner due to either safety or physical constraints.