Room: AAPM ePoster Library
An RF pulse sequence is designed and implemented onto scanner to produce 3D sparse sampling chemical-shift-imaging (CSI) for hyperpolarized 13C metabolic MRI. The 3D CSI is reconstructed with a 4D compressed sensing algorithm.
The sparse sampling patterns are firstly generated with higher density around the center of the 3D k-space. The acquisition starts from the center of the k-space and roll out to the edge of the 3D k-space for optimum MR signal efficiency. The acquisition of k-space voxels is encoded according to the sparse pattern. A variable-flip-angle scheme is implemented to compensate the loss of high frequency signal. At reduction rate of 8 or 16 (R=8 or 16), a CSI matrix of 16x16x16 can be completed in 40 and 20 seconds respectively which are within the hyperpolarized MR signal time frame (typically T1=20s to 40s). The CSI reconstruction of the sparsely acquired data is accomplished with a 4D compressed sensing algorithm with adaptive stepping regularization weighting. The sequence is implemented on a 13C enabled GE MR750W scanner and tested with a phantom of four chambers filled with C1-labeled 13C metabolites lactate, alanine, formic acid and bicarbonate respectively.
The sequence has operated successfully on the scanner in real-time sparse acquisition with reduction rates R=2 to 32 (50% to 3.125% sampling). The reconstructed CSI is of high quality. At R=16 (6.25% sampling) the normalized root-of-mean-square-error nRMSE=2.34x10^-3 and the structural-similarity-index SSIM=0.977. Even at R=32, it produced an nRMSE=3.66x10^-3 and SSIM=0.949. Both acquisition time and image quality reached our goals and made the hyperpolarized 3D metabolic MR CSI possible.
The CSI with full-length Free-Induced-Decay (FID) acquisition can cover a wider spectrum band comparing to partial FID methods. Sparse phase encoding CSI acquisition combining with 4D compressed sensing is the practically possible method for hyperpolarized metabolic 3D MR CSI.