The invention relates to a dynamic sliding-window-like initialization for, for example, iterative VMAT algorithms. Specifically, a dynamic sliding window conversion method is contemplated where typical dynamic VMAT constraints are taken into account to find an optimal set of suitable openings (i.e. binary masks) that can be used as quasi-feasible start initialization for any VMAT algorithm that can refine until a deliverable plan is reached. Here, a multileaf leaf tip trajectory least square constrained optimization is performed to find a set of optimal unidirectional trajectories for all MLC leaf pairs of all arc points. To ensure that a quasi-feasible (or better quasi-deliverable) solution is returned, for example, a maximum dose rate, a maximum gantry speed, a maximum leafs speed, and a maximum treatment time may be enforced.
