When the shared contour is the problem

A challenge hands out one structure set. That makes contouring feel like it is not your problem, and it is exactly why it is.

A wrong contour does not show up as a wrong score. It shows up as everyone converging on the same odd trade-off. If the whole field is fighting one constraint that nobody fights in clinic, the constraint is not necessarily wrong — the structure it is written against might be.

What to check before deciding your plan is the problem.

• Does the organ at risk extend further than you would draw it? A structure with extra volume in a low-dose region reads better than it should; extra volume near the target reads worse.
• Is the convention stated? A cauda equina contour can mean the whole canal, the nerve bundle, or the bundle with the roots — three different D0.03cc values from the same dose.
• Does the target include something you would exclude, or exclude something you would include? Coverage metrics are computed on their volume, not yours.

Then raise it, publicly. Open a data issue on the case rather than emailing anyone. It appears at the top of the case page for everyone still planning, the criteria can be revised and every submission re-scored under the new version, and the reason is recorded where the people whose ranks moved will see it.

This is the failure mode a leaderboard cannot self-correct. A bad plan sinks on its own. A bad contour lifts or sinks everyone together, silently, and only a person looking at the anatomy catches it.