| US 7,394,252 B1 | ||
| Regularized GRAPPA reconstruction | ||
| Fa-Hsuan Lin, Brookline, Mass. (US) | ||
| Assigned to The General Hospital Corporation, Boston, Mass. (US) | ||
| Filed on May 03, 2007, as Appl. No. 11/743,739. | ||
| Int. Cl. G01V 3/00 (2006.01) | ||
| U.S. Cl. 324—309 [324/307] | 20 Claims |
| 1. A method for generating a fully sampled k-space data set corresponding to each of a plurality of receiver channels in a
parallel MRI system having a plurality of corresponding receiver coil elements, the fully sampled k-space data sets corresponding
to an image of a scanned subject using the parallel MRI system, the method comprising:
(a) performing an undersampled GRAPPA scan of the subject using the set of receiver coil elements and corresponding receiver
channels to obtain a reduced k-space data set,
(b) obtaining a plurality of autocalibration samples in k-space for each receiver channel,
(c) calculating a reconstruction kernel from the plurality of autocalibration samples and the reduced k-space data set,
(d) formulating a Tikhonov regularization framework to allow a tradeoff between reliance on a conventional GRAPPA reconstruction
which uses the reconstruction kernel and a replication of the prior k-space data from the plurality of autocalibration samples,
with the tradeoff determined by an optimal regularization parameter; and
(e) reconstructing missing k-space data to obtain a reconstructed k-space data set which together with the reduced data set
fully samples each channel of k-space, the missing k-space data being reconstructed using one of a regularized GRAPPA reconstruction
using the Tikhonov regularization framework and a conventional GRAPPA reconstruction, wherein for locations in the reconstructed
k-space data set which correspond to locations of autocalibration samples included in the plurality of autocalibration samples,
a regularized GRAPPA reconstruction is performed, and wherein conventional GRAPPA reconstructions are performed if no prior
k-space data is available.
|