Efficient sample density estimation by combining gridding and an optimized kernel

The reconstruction of non-Cartesian k-space trajectories often requires the estimation of nonuniform sampling density. Particularly for 3D, this calculation can be computationally expensive. The method proposed in this work combines an iterative algorithm previously proposed by Pipe and Menon (Magn Reson Med 1999;41:179-186) with the optimal kernel design previously proposed by Johnson and Pipe (Magn Reson Med 2009;61:439-447). The proposed method shows substantial time reductions in estimating the densities of center-out trajectories, when compared with that of Johnson. It is demonstrated that, depending on the trajectory, the proposed method can provide reductions in execution time by factors of 12 to 85. The method is also shown to be robust in areas of high trajectory overlap, when compared with two analytical density estimation methods, producing a 10-fold increase in accuracy in one case. Initial conditions allow the proposed method to converge in fewer iterations and are shown to be flexible in terms of the accuracy of information supplied. The proposed method is not only one of the fastest and most accurate algorithms, it is also completely generic, allowing any arbitrary trajectory to be density compensated extemporaneously. The proposed method is also simple and can be implemented on parallel computing platforms in a straightforward manner.