The Effect of Dental Artifacts, Contrast Media, and Experience on Interobserver Contouring Variations in Head and Neck Anatomy

Abstract

To investigate interobserver variability in the delineation of head-and-neck (H&N) anatomic structures on CT images, including the effects of image artifacts and observer experience. Nine observers (7 radiation oncologists, 1 surgeon, and 1 physician assistant) with varying levels of H&N delineation experience independently contoured H&N gross tumor volumes and critical structures on radiation therapy treatment planning CT images alongside reference diagnostic CT images for 4 patients with oropharynx cancer. Image artifacts from dental fillings partially obstructed 3 images. Differences in the structure volumes, center-of-volume positions, and boundary positions (1 SD) were measured. In-house software created three-dimensional overlap distributions, including all observers. The effects of dental artifacts and observer experience on contouring precision were investigated, and the need for contrast media was assessed. In the absence of artifacts, all 9 participants achieved reasonable precision (1 SD < or =3 mm all boundaries). The structures obscured by dental image artifacts had larger variations when measured by the 3 metrics (1 SD = 8 mm cranial/caudal boundary). Experience improved the interobserver consistency of contouring for structures obscured by artifacts (1 SD = 2 mm cranial/caudal boundary). Interobserver contouring variability for anatomic H&N structures, specifically oropharyngeal gross tumor volumes and parotid glands, was acceptable in the absence of artifacts. Dental artifacts increased the contouring variability, but experienced participants achieved reasonable precision even with artifacts present. With a staging contrast CT image as a reference, delineation on a noncontrast treatment planning CT image can achieve acceptable precision.