Contouring is where the planning pathway slows down.
Before any dose can be calculated, every organ at risk has to be drawn. It is careful, repetitive, expert work — and it sets the pace of the whole department.
A radiation oncology plan cannot begin until the structures exist. That single dependency makes manual delineation the rate-limiting step in the pathway: not the linac, not the planning system, not the clinician's judgement — the drawing.
It is also the step where consistency is hardest to hold. The same scan contoured by two clinicians produces two different structure sets, and the same clinician at the end of a long list does not necessarily contour as they did at the start.
Hours per patient
A full set of organs at risk is drawn slice by slice, structure by structure. For head & neck cases, nodal levels are the most variable and most time-intensive step in the entire plan.
Inter-observer variability
Two clinicians, one scan, two different structure sets. That variation does not stay in the contour — it propagates directly into the dose distribution and into any comparison between patients or sites.
Contouring gates the queue
Planning cannot start until structures are complete, so delineation capacity — not treatment capacity — determines how many patients move through the department each week.
Tools tied to a workstation
Conventional automation arrives as a TPS plugin: one vendor, one version, one machine — and a clinician physically present to initiate every single case.
Semi-automatic is still manual
Most existing tools need a seed point or an initial contour before they will run. Every structure still requires manual initialisation, so the clinician's time scales with the structure count.
Generic models, tuned to nobody
Pre-trained on public or pooled datasets, the same model is shared across every institution — reflecting no single site's patient population, scanner protocols or delineation conventions.
Where this goes