Abstract: Spiral magnetic resonance imaging (MRI) data are sampled in the k-space with non-uniform spiral trajectory so that gridding must first be performed to transform the raw data into uniform space before fast Fourier transform (FFT reconstruction) can be used to reconstruct images. In this paper, we compared two classes of gridding algorithms, the large matrix resampling algorithm proposed by Claudia et al and the double-sized gridding algorithm proposed by Jackson et al, with respect to the speed of computation and the quality of resultant images. First, it was shown that, while capable of getting images with similar quality relative to those obtained by the latter, the former algorithm is faster and easier to implement. Secondly, our results showed that data-driven interpolation is more efficient than grid-driven interpolation when the doublesized gridding algorithm is used. Lastly, in the large matrix resampling algorithm, when the efficient-versus-redundancy data ratio surpasses 310∶1, it is more efficient not to average the values of gridding points of the large matrix. The conclusions are helpful to improve the gridding and reconstruction techniques used in spiral MRI.

Key words: MRI gridding, reconstruction, data-driven interpolation, grid-driven interpolation, large matrix resampling