Spoke pulse design in magnetic resonance imaging using greedy minimax algorithm

Spoke RF pulse design in MRI requires joint optimization of the k-space trajectory and RF pulse weights. This design task is often modelled as a sparse approximation problem with a cost function evaluating the l₂ norm of the excitation error, which can be approximately solved using the orthogonal matching pursuit (OMP) algorithm. However, l₂ optimization does not strictly regulate a maximum deviation between excitation and desired patterns, and may leave bright or dark spots in the image. In this paper, we model the pulse design problem as a sparse approximation problem with an l_∞ norm cost function, and propose a greedy algorithm for solving this new problem. Simulation results demonstrate that our algorithm can produce improved spoke RF pulses (reduced maximum error) compared to l₂ optimization.