Mitigating Distortion Artifacts in Accelerated-EPI Using an Ensemble of k-t GRAPPA Kernels (EnKT-GRAPPA)

Motivation: EPI suffers from field-inhomogeneity distortions. Employing large inplane acceleration (Rinplane) can mitigate this issue at a cost of increase noise and artifacts, while postprocessing correction can lead to resolution-loss. Goal(s): Develop a reconstruction method based on an ensemble of k-t GRAPPA-kernels (EnKT-GRAPPA) for use on moderately-accelerated EPI, to both fill missing-kspace and reduce distortion. Approach: EnKT-GRAPPA kernels are trained using k-t calibration data to fill missing-kspace and correct for cumulative-phase of field-inhomogeneity in one step, where phase/distortion correction level can be flexibly tuned. Results: For the same distortion mitigation level, EnKT-GRAPPA-reconstructed-images exhibit higher-SNR compared to those from conventional GRAPPA-reconstructed-images of a higher-Rinplane acquisition. Impact: EnKT-GRAPPA enables moderately accelerated EPI to achieve a high level of distortion mitigation while preserving SNR. This method should be useful in many applications such as fMRI, diffusion and perfusion imaging.