What to contour on a 4D scan

A free-breathing lung tumour moves 5–30 mm in the superior–inferior direction, and AAPM TG-76 treats displacement above 5 mm as the point where respiratory compensation should be considered rather than ignored. A 4D scan measures that motion; it does not decide what to do about it.

Three answers, and they are not equivalent.

• ITV from all phases. Contour the GTV on each phase and union them. Safe, and known to be generous: an ITV overestimates the margin breathing actually requires, because it assumes the tumour spends equal time at every position and it does not.
• MIP. A maximum-intensity projection collapses the phases into one image, and for a lung tumour it shows the full excursion, because tumour is dense and lung is not. That is exactly why it fails where the tumour touches something of similar density — chest wall, mediastinum, diaphragm, atelectasis. Against those, MIP silently draws the smaller volume. Check it against the individual phases before trusting it, and stop trusting it when the tumour abuts.
• Mid-ventilation. Contour on the phase closest to the time-weighted mean position and apply a systematic-error margin instead. Smaller volume for the same coverage probability, more work to justify.

What to write down. The measured excursion in each axis, which phases you contoured on, and — if you used MIP — whether the tumour abutted anything. A write-up that says "ITV from 4D-CT" and nothing else cannot be compared with anyone else's, because those three approaches produce different volumes on the same scan.

On this platform. A seeded 4D-Lung case ships the structure set its collection published, contoured on one phase. If your plan depends on a different motion assumption than the one the contours encode, say so — the criteria score the dose you delivered to the contours provided, and the argument about which contour was right belongs in the write-up.