Pareto navigation instead of re-optimizing

Multi-criteria optimization computes a set of plans where no objective can be improved without making another worse — the Pareto surface — and then lets you navigate between them with sliders. The navigated dose is interpolated from the precomputed plans, and DVH, dose statistics and clinical goals update in real time.

What changes about the work. Ordinary planning is a guess-and-check loop: change a weight, wait for a run, see what moved. Navigation shows you the exchange rate directly — this much rectum for that much coverage — because both plans already exist. You stop asking "can I do better" and start asking "which trade do I want".

Automatic navigation toward a prioritized list of clinical goals exists too, implemented as lexicographic optimization: satisfy goal one, then goal two without breaking one, and so on. That maps cleanly onto a scored challenge, where hard constraints come before weighted metrics.

The catch. The navigated plan is an interpolation. It has to be converted back into a deliverable plan afterwards, and the deliverable version is not exactly the navigated one — check the DVH again after conversion, not before.