Meeting Banner
Abstract #4245

Data Decoupling for Efficient Auto-Calibrating Parallel Imaging for Arbitrary Cartesian K-Space Sampling: Application to Highly-Accelerated 3D Cardiac Cine MRI

Peng Lai1, Anja C.S Brau1

1Global Applied Science Laboratory, GE Healthcare, Menlo Park, CA, United States

This work develops a novel source data decoupling method for computationally efficient auto-calibrating parallel imaging for arbitrary Cartesian sampling and validates its performance on 3D cardiac cine MRI with different acceleration factors and different coils. Our computation analysis shows that the proposed method can dramatically reduce calibration time compared to conventional autocalibrating parallel imaging methods, especially with large numbers of coil channels and synthesis patterns. Our invivo results on cardiac cine MRI also show the decoupling method preserves image quality of conventional kt GRAPPA and provides more accurate reconstruction and higher SNR than GRAPPA operator based methods due to its optimal combination of syntheses from different source data groups. The proposed data decoupling method is promising for online reconstruction for 3D cardiac cine MRI and other static 3D MRI applications.

Abstract PDF E-Poster Video

Keywords

accelerated acceleration accurate acquisition adaptive address among analyzed another artifacts assess assuming auto available basis better black calibrating calibration cardiac central challenge challenging chart cine coil coils collected combination combine combining complete complex comprised computation considerably correlation covariance cycle dataset datasets decoupled decoupling degrade density develop develops dots dramatically efficiency efficient errors especially exactly expensive fashion feasibility final find finds fitting flow fully furthermore gating generate generated generates ghosting global grog highly improvement included indicated instance interleaved inversion involves kernel laboratory larger like local matrix motion much multiplications neighborhood neighbors noise novel operator optimal outer parallel park pattern patterns potential potentially preparation proposed pseudo quality random recombine reconstructed reconstruction reconstructions recovered reduce reduces reduction regard regular regularization removing repeated resolution sampling science significantly sliding solid sorted source space static step suboptimal substantially syntheses synthesis synthetic table target targets temporal theory timer trivial typical uniform validate variable visible volunteers width window words