Qualitative Contouring Changes After Completion of Radiation Oncology Resident Clinical Rotations: Secondary Analysis of an International Longitudinal Trial
IJROBP 2026
Abstract
Background: Accurate contouring is essential for high-quality radiation therapy, yet variability remains common during training and beyond in clinical practice. This study qualitatively characterized contouring mistakes across different disease sites before and after site-specific clinical rotations. We hypothesized that residents would exhibit more errors pre-rotation, with improvement following rotation completion.
Methods: Residents completed pre- and post-rotation contouring assessments consisting of three tasks from their respective disease sites. Each resident’s contour was visually compared to an expert reference contour overlaid on the same axial slice. Residents completing their first clinical rotation in Head/Neck (HN), gynecologic (Gyn), or gastrointestinal (GI) disease sites were randomized 1:2 to receive standard contouring education resources (control) or iContour educational priming. iContour is a web-based platform that displays anonymized DICOM images and provides real-time feedback to residents as they contour practice cases.
Results: This trial recruited 31 residents, primarily from the United States (n = 27), with training ranging from PGY2-5 (median PGY-2). Pre- and post-test contours were available for 12 HN, 14 Gyn, and 5 GI residents, totaling 185 contours. For primary-tumor CTV contouring in a nasopharynx case, residents demonstrated undercoverage of the pterygoid muscles (75% pre, 92% post) and overly liberal anterior nasopharyngeal coverage (33% pre, 33% post). Elective HN nodal and OAR contouring demonstrated under-contouring of level Ib (33% pre, 66% post) and V (83% pre, 83% post) cervical lymph nodes, and difficulty identifying the deep parotid lobe (25% pre, 50% post). Notably, residents no longer included the pharyngeal constrictor muscles (50% pre, 0% post) in the post-rotation nodal CTV. In the GI case, residents over-contoured the anal canal anteriorly into the prostate (60% pre, 20% post) and posteriorly into the ischiorectal fossa (40% pre, 40% post). GYN contouring was more accurate, with appropriate coverage of involved common iliac nodes (69% pre, 57% post) and bowel bag (86% pre/post). No correlation was observed between educational priming and contour performance.
Discussion: Resident errors included under-contouring of the deep parotid gland and key nodal subsites in the HN case, and over-contouring of the anal canal into adjacent structures in the GI case. In GYN cases, residents overall demonstrated more proficient contouring with proper lymph node and bowel bag coverage. There were few clear improvement patterns following rotation completion, consistent with previously reported quantitative analyses. These findings highlight predictable site-specific errors that may guide targeted education and feedback for trainees and clinicians.
Matin Yarmand
Ph.D. Alumni at UCSD
Computer Science and Engineering
