Variability of Gross Tumor Volume Delineation in Head-and-Neck Cancer Using PET/CT Fusion, Part II: The Impact of a Contouring Protocol

Abstract: The purpose of this study was to assess the efficacy of a gross tumor volume (GTV) contouring protocol on interobserver variability between 4 physicians in positron emission therapy/computed tomography (PET/CT) treatment planning of head-and-neck cancer. A GTV contouring protocol for PET/CT treatment planning was developed utilizing 4 stages: Preliminary contouring on CT alone, determination of appropriate PET windowing, accurate image registration, and modification of CT contouring with correctly formatted PET/CT display and rules for modality disagreement. Two neuroradiologists and 2 radiation oncologists (designated as A, B, C, and D, respectively) were given a tutorial of PET/CT coregistered imaging individualized to their skill level, which included a step-by-step explanation of the protocol with clinical examples. Opportunities for questions and hands-on practice were given. The physicians were asked to re-contour 16 head-and-neck patients from Part I on PET/CT fusion imaging. Differences in volume magnitude were analyzed for statistical significance by analysis of variance (ANOVA) and paired t-tests (α < 0.05). Volume overlap was analyzed for statistical significance using Wilcoxon signed-rank tests (α < 0.05). Volume overlap increased significantly from Part I to Part II (p < 0.05). One previously significant difference between physicians disappeared with the protocol in place. The mean fusion volume of Physician C, however, remained significantly larger than that of Physician D (p < 0.01). This result is unchanged from Part I. The multidisciplinary contouring protocol significantly improved the coincidence of GTVs contoured by multiple physicians. The magnitudes of the volumes showed marginal improvement in consistency. Developing an institutional contouring protocol for PET/CT treatment planning is highly recommended to reduce interobserver variability. [Copyright &y& Elsevier]