Convolution kernel decision scheme for reconstruction in k-t/k space
Huang F, Duensing G
In this work, a novel reconstruction algorithm is introduced. The method is based on allowing the convolution kernel to be selected on the basis of prior acquisitions or from ACS lines within the acquired data. Because of the size of the convolution kernel, temporal resolution of the images reconstructed by k-t GRAPPA may be reduced. To improve the temporal resolution, it is proposed that the choice of the convolution kernel should be guided by the convolution matrix generated with acquired data. Experiments showed that the proposed method could generate better results than k-t GRAPPA with a fixed, small kernel definition.