High-Accelerated Parallel Imaging With the Inherent Local Feature in PE-xSPEN MRI

To achieve highly accelerated high-resolution imaging with PE-xSPEN that efficiently utilizes the spatiotemporal encoded signals to physically distinguish contributions from distinct spatial locations prior to k-space acquisition, enabling the differentiation of aliased localized information in k-space.

In the conventional k-space, undersampling along the phase-encoding dimension results in aliasing and overlapping replicas of the object along the $y$ spatial direction. The modulated hyperbolic phase arising from the PE-xSPEN encoding induces coupling between the $y$ and $z$ dimensions. So that for a given k-space location, the signal is predominantly contributed by local voxels. While undersampling along the PE dimension still causes aliased spatial replicas, these replicas remain distinguishable even without multi-channel coils after applying the traditional inverse Fourier transform along the undersampled dimension. Combining with parallel imaging, this approach enhances unfolding efficiency compared to traditional Fourier-encoded signals.

Combining with GRAPPA, the undersampled signals along the $y$ dimension can be reconstructed with fewer residual aliasing in PE-xSPEN compared to the traditional Fourier-encoded signals. The performance gain is demonstrated by MRI simulation, phantom tests, and in vivo brain experiments on 3T scanner.

The proposed PE-xSPEN GRAPPA framework utilizes hyperbolic phase modulation to control aliasing artifacts and generate $k_z$ -dependent coil response, reducing artifacts and noise amplification.