Analysis of the variability among radiation oncologists in delineation of the postsurgical tumor bed based on 4D-CT

OBJECTIVE This study investigated interobserver and intraobserver variability in radiation oncologists' definition of the tumor bed (TB) after breast-conserving surgery (BCS). RESULTS The TB volume, CVS and number of surgical clips were not significantly related to intraobserver variability. Moreover, no correlation was noted between CT slice thickness and interobserver variability (Δinter, DSCinter) in TB delineation, and no significant difference was noted among the three groups. The TB volume was negatively correlated with Δinter. DSCinter improved significantly with increased TB volume and decreased Δinter. DSCinter also increased significantly in patients with a CVS of 3 to 5 compared with patients with a CVS of 1 to 2. DSCinter was thus positively correlated with the CVS, with a correlation coefficient of 0.451. The use of 7 to 9 surgical clips neither decreased Δinter nor increased DSCinter. MATERIALS AND METHODS Five or more surgical clips were placed at the TB during lumpectomy. The TB was delineated on the end expiration scan. The data were stratified based on the cavity visualization score (CVS), CT slice thickness and surgical clip number. The Dice similarity coefficient (DSC) and inter(intra)observer variability (Δinter and Δintra) in different groups were evaluated and compared. CONCLUSIONS Inter(intra)observer variability in TB delineation was decreased for breast cancer patients implanted with 5 or more surgical clips in the cohort with a higher CVS and a larger TB. The use of more than 6 surgical clips did not significantly improve TB delineation, so 5 to 6 surgical clips are likely adequate to delineate the TB.