Auto-contouring is the use of AI to draw treatment-relevant boundaries on medical images such as CT, MRI, PET, or cone-beam CT scans. In cancer care, those boundaries may include the tumor itself, target volumes for radiotherapy, organs at risk, lymph-node regions, surgical landmarks, or structures that should be avoided during treatment.
Why It Matters
Auto-contouring matters because manual contouring is slow, repetitive, and variable across clinicians. In radiotherapy planning, even small differences in how a target or organ is outlined can change the delivered dose. AI can help reduce that workload and make planning more consistent, especially when many structures must be reviewed under time pressure.
Good auto-contouring does not remove the need for expert oversight. It shifts the job from drawing everything from scratch to reviewing, correcting, and approving contours that were generated automatically.
Where AI Fits
Most auto-contouring systems rely on computer vision models trained on large sets of expert-labeled scans. They may work on tumors, organs at risk, or both, and they are often part of broader radiotherapy workflows that include dose planning, adaptation, and quality assurance.
Because the quality of an auto-contour depends heavily on imaging protocol, label consistency, and anatomy shifts, strong systems also depend on reliable ground truth, careful calibration, and explicit handling of uncertainty.
Related Yenra articles: Cancer Treatment Planning, Precision Oncology and Targeted Therapies, Patient Outcome Prediction, and Personalized Medicine.
Related concepts: Computer Vision, Radiomics, Ground Truth, Calibration, and Uncertainty.