Geometric and dosimetric impact of inter-professional contour variability in the upper abdomen for MR-linac

To quantify the geometric accuracy and dosimetric impact of delineation variability for Radiation Therapists (RTs) compared to Radiation Oncologists (ROs) when contouring the organs at risk (OARs) within the upper abdomen on MR imaging.

Eight RT and five RO volunteers contoured six abdominal OARs on MR images for three primary liver cancer cases. The clinical plan was then recalculated, and the study contours were used to generate dose-volume histograms for the OARs. Dice Similarity Coefficient (DSC), Distance to Agreement (DTA) and doses at threshold OAR volumes were used to quantify differences in RT vs. RO contours.

RO STAPLE to RT STAPLE comparisons resulted in DSC ≥0.7 for each patient and each organ. The median DTA surface differences between the RO STAPLE and RT STAPLE were <0.4 cm for all OARs except duodenum and small bowel. The mean dose values for the 8 RT and 5 RO were less than 5% different in 10/24 OAR volumes. RT contours resulted in mean differences >5% hotter than RO contours in 7 OAR volumes, and >5% colder in 7 OAR volumes. Colder OAR volumes were found for duodenum, large and small bowel. Only the stomach 5cc mean dose exceeded the threshold dose constraint for both RO and RT contours. Dose differences between RO STAPLE and individual RT contours were not significantly correlated with either DSC or DSC (ρ < 0.404). There were large inter-patient differences in the dosimetric impact of delineation variability.

Inter-observer delineation variability of OAR in the upper abdomen can be large for both ROs and RTs. The dosimetric impact of those volumetric differences is moderated by proximity of the OAR to the target and the orientation of the plan dosimetry.