Quantitative susceptibility mapping (QSM) provides novel insights into tissue composition complementary to established contrasts and yields 10-fold-increased grey matter/white matter contrast compared to magnitude imaging. As susceptibility effects scale with the magnetic field QSM highly profits from scanning at ultra-high field¹.

Data are conventionally acquired by Cartesian 3D spoiled gradient-echo (GRE) sequences. However, high spatial resolutions come along with extensive measurement times, which can be problematic in terms of head motion. Providing a significantly higher sampling efficiency, Cartesian echo-planar-imaging (EPI) represents a fast alternative to GRE but its application is limited by the gradient performance, geometric distortions and signal dropouts.

The purpose of this study was to perform QSM at 7T at an isotropic resolution of 1 mm. To achieve a high sampling efficiency, minimize distortions and maximize the motion robustness, we propose a non-Cartesian 3D EPI sequence with a paddlewheel-shaped readout scheme^2,3.

The planes-on-a-paddlewheel (POP) trajectory is realized by rotating 2D EPI readout planes about the phase encoding axis (Fig. 1a). Single planes on this paddlewheel are sampled at each excitation, with the slab selection performed along the rotation axis.

The concept was evaluated in a healthy volunteer (26, female). Measurements were performed on a 7 T whole body scanner (Siemens Healthcare, Germany) with a gradient strength of 70 mT/m, slew rate of 200 mT/m/s and a 32 channel Tx/Rx head array (Nova Medical, USA). Images were obtained at an isotropic resolution of 1 mm after 3^rd order shimming using the following acquisition parameters: FOV = 212 x 212 x 108 mm³, 330 projections, TR = 47 ms, TE = 24 ms, ES = 1.0 ms, flip angle = 13°, ramp sampling, 10 measurements. A homogeneous azimuthal distribution of the planes was achieved by employing an interleaved radial projection order (Fig. 1b). For the correction of gradient delays and Nyquist ghosting three non-phase-encoded navigator echoes were used. Parallel imaging with undersampling factor R_PE = 3 was applied along the phase encoding (PE) direction.

Post ramp sampling regridding, gradient delay and Nyquist ghost correction was performed, and GRAPPA was employed to reconstruct the missing phase encoding lines with the weight sets determined from a separate 3-fold segmented full resolution reference scan. Gridding of the non-Cartesian data was performed using the non-uniform fast Fourier transform (NUFFT) software⁴. For susceptibility mapping, a single-step QSM method using total generalized variation⁵ (TGV) was employed for each coil channel phase data individually. The brain mask for QSM was derived from the root-sum-of-squares-combined magnitude data. Finally, QSM images were obtained by calculating the mean of all channels. All QSM images were reconstructed using the same regularization parameters for the TGV-QSM algorithm.

For comparison, conventional imaging was performed using a 3D multi-echo GRE sequence with identical spatial resolution (FOV = 212 x 212 x 120 mm³, TR = 29 ms, 7 echoes, TE₁= 4.36 ms, ES = 2.86 ms, flip angle = 13°, GRAPPA with R^PE = 3 and 16 additional calibration lines).

The images obtained by the 3D POP EPI sequence are free from reconstruction artifacts (Fig. 2a) and smaller anatomical features of the brain can be delineated well. Both images and susceptibility maps (Fig. 2b) are comparable to the corresponding 3D GRE data. As can be seen in Fig. 2c, the image geometry is not significantly altered with respect to the 3D GRE measurement. The superimposed contours (red) were calculated from the corresponding 3D GRE images after applying an affine co-registration algorithm. No deviations can be identified within the axial plane while slight distortions are found along the phase encoding axis (arrows). Figure 3 summarizes susceptibility values obtained in three sub-cortical regions of interest. Mean and standard deviation are not affected by the number of 3D POP EPI measurements and are comparable to the values from 3D GRE.High resolution quantitative susceptibility mapping was successfully performed at ultra-high field using 3D POP EPI. With less than 16 seconds per volume, the proposed non-Cartesian readout scheme is considerably faster than standard 3D multi-echo GRE imaging. The radial nature of the readout is beneficial in terms of motion robustness and provides high flexibility for further undersampling. Due to the smaller FOV in the phase encoding direction and by employing parallel imaging with an acceleration factor of R_PE = 3, low geometric distortions were achieved, also in this most problematic dimension. The low variance of the susceptibility values for multiple measurements indicates a high reproducibility of the technique.