4D MRI is an in vivo flow imaging modality which has the potential to significantly enhance diagnostics and therapy of cardiovascular diseases. However, current analysis methods demand too much time and expert knowledge in order to apply 4D MRI in the clinics or larger clinical studies. One missing piece are methods allowing to gain a quick overview of the flow data's main properties. We present a line predicate approach that sorts precalculated integral lines, which capture the complete flow dynamics, into bundles with similar properties. We introduce several streamline and pathline predicates that allow to structure the flow according to various features useful for blood flow analysis, such as, e.g., velocity distribution, vortices, and flow paths. The user can combine these predicates flexibly and by that create flow structures that help to gain overview and carve out special features of the current dataset. We show the usefulness of our approach by means of a detailed discussion of 4D MRI datasets of healthy and pathological aortas.