Parallel MRI with Extended and Averaged Grappa Kernels (PEAK-Grappa): SNR optimized fast dynamic imaging with high acceleration factors

Introduction: In order to reduce total acquisition time or to increase spatiotemporal resolution in dynamic MRI several techniques such as TSENSE [1], TGrappa [2], kt-SENSE, kt-BLAST [3] and kt-Grappa [4] have recently been introduced. The combination of parallel MRI and temporal imaging acceleration permits the use of higher acceleration factors compared to conventional parallel imaging techniques, e.g. SENSE and Grappa. Nevertheless, all techniques suffer from decreased SNR and/or increased temporal blurring with higher acceleration factors. Here we present a novel reconstruction for dynamic MRI which optimally utilizes information along all available spatial directions as well as the temporal dimension. A method called PEAK-Grappa (Parallel MRI with Extended and Averaged Grappa Kernels) is presented which is based on an extended spatiotemporal Grappa kernel in combination with temporal averaging. By including Grappa weight estimation along the time dimension while at the same time exploiting the averaging properties of uncorrelated noise of different acquired time frames, it is possible to maintain or even increase SNR while minimizing temporal blurring. Phantom measurements with reduction factors up to 12 and results from volunteer experiments demonstrate that PEAK-Grappa preserves the dynamics of timeresolved measurements while considerably improving SNR compared to conventional methods.