Quantitative Analysis of Contouring Performance Among Novice Residents and Implications for Future Educational Efforts

Summary: Effective contouring education is vital in radiation oncology residency training, as research links contouring variability to negative patient outcomes. Teaching this skill requires understanding common errors made by novice residents. We conducted a study using an interactive online platform to compare novice resident performance against an expert on real patient cases. We hypothesized this software would identify key mistakes and teaching points for residents before their disease-site-specific training. iContour is a novel web-based platform that is in development. It displays anonymized DICOM data and allows users to input their contours on practice cases. To assess the efficacy of the platform, a randomized study is enrolling residents in their first clinical rotation for either Head/Neck (HN), Gynecologic (GYN), or Gastrointestinal (GI) cancers. At the start of their rotation, participants completed a baseline contouring assessment with three contouring tasks. Tasks included contouring targets (CTVs, GTVs) or organs at risk for a case from their disease site. For this analysis of baseline resident contouring skill, expert contours of the same case were overlaid and compared with the resident contours using the Dice-Sørensen coefficient (DSC). A total of 28 residents (median PGY2, range PGY2-5) from 16 institutions (11 US, 5 international) participated resulting in 83 unique contours. There were 11 residents participating in both the HN and GYN sites, and 6 residents in the GI site. Overall, average DSC scores for resident (vs expert) contours were lowest for the GI disease site and highest for GYN (Table 1). Amongst specific structures, good agreement with expert contours (DSC≥0.8) was seen for the nodal GTV and bowel bag structures in the GYN case. The most difficult structures (DSC< 0.5) were the elective nodal CTVs for HN and GI. Residents had the highest variability in contouring the primary CTV for both GI and HN cases, as well as contouring the parotid gland. These patterns of performance and error in residents who are new to a disease site demonstrate high-yield areas to target with educational efforts during clinical rotations. A randomized trial assessing the efficacy of the iContour platform with or without additional priming education for contouring training is underway. Megan E. Orr (she/her/hers), MD (Presenting Author) - UC San Diego; Kristin H. Daniels, BS (Co-Author) - UC San Diego; Elizabeth A M. Duran, BS (Co-Author) - UC San Diego; Borui Wang, BS (Co-Author) - University of Washington; Matin Yarmand, BS MS (Co-Author) - UC San Diego; Leah A. D'Souza, MD MSc (Co-Author) - Rush University; Nadir Weibel, PhD MSc (Co-Author) - UC San Diego; Erin F. Gillespie, MD MPH (Co-Author) - University of Washington; James Murphy, MD MS (Co-Author) - UC San Diego Health; Michael V. Sherer, MD (Co-Author) - UC San Diego