A heart atlas for breast radiation therapy and the influence of delination education on both intra and interobserver variability

Abstract

We developed a heart atlas for breast radiation therapy and evaluated the influence of education on intra and inter-observer similarity, and cardiac dose reporting. The data of 16 left breast cancer patients were analyzed. Eight observers delineated heart and cardiac subunits [left (LCA) and right (RCA) coronary arteries, left anterior descending artery (LAD), bilateral atrium and ventricles] before the education. A radiologist and radiation oncologist developed the atlas and delineated the gold standard (GS) volumes. Observers repeated the delineation after education. RT plans were made for pre/post-atlas contours. The similarity was assessed by Dice (DSC) and Jaccard (JSC) similarity coefficient indices. The absolute difference rate was calculated for the dose analysis. The inter-observer similarity increased in heart and all subunits. The intra-observer similarity showed a heterogeneous distribution. The absolute difference rate in dose reporting was statistically significant for the bilateral atrium, right ventricle, LAD, LCA+LAD, RCA's maximum doses (p < 0.05). The maximum dosereporting differences from the GS decreased from 16.9 to 8.9% for LAD (p = 0.011); from 14.8 to 9.3% for LCA+LAD (p = 0.010). The cardiac atlas reduces the intra-interobserver differences and improves dosereporting consistency. The first intra-observer similarity analysis was made in our study and revealed the need for repeated education to increase the consistency.