Assessing Quality of Heart Contouring and Dosimetry in Treatment of Left Sided Breast Cancer Across Cancer Centers within an Integrated Health System

Abstract

Cardiac sparing techniques in the treatment of left-sided breast cancer have led to a significant decrease in mean heart dose (MHD) over time. Inconsistency in contouring can impact accurate evaluation of cardiac dosimetry. Atlas Based Auto-Segmentation (ABS) can be used to reduce this inter-observer variability. We sought to assess quality of heart contouring and dosimetry across six radiation oncology departments within an integrated health system and assess the utility of ABS. We selected at random 60 patients (pts) treated with 3D conformal radiation therapy for left-sided breast cancer in 2017 at six centers within an integrated health system. A similar number of pts treated to left breast, chest wall (CW) or breast with regional lymph nodes (RNI), and CW or breast with RNI + internal mammary chain (IMC) were selected from each center. Doses were 40.05 Gy-50 Gy in 16-25 fractions. To account for different dose/fx regimens, treatment plans were scaled linearly to a dose of 50 Gy. Volumes and dosimetry for “gold standard” heart (GS-H), contoured based on RTOG 1005 and NSABP B51 protocol definitions, clinician contoured heart (C-H), and ABS generated heart (ABS-H) were compared using paired T-test. Volumes and dosimetry were compared across centers using one-way ANOVA. Dice coefficient was used to compare volume of overlap and Hausdorf distance was used to compare the distance between contours. There was no significant difference between the percentage of pts treated to left breast (p=0.97), CW or breast with RNI (p=0.50), and CW or breast with RNI + IMC (p=0.88) across centers. There was no significant difference in MHD (range 1.4 to 2.0; p=0.54), mean D1cc (range 20.5 to 35.2; p=0.22), and mean V5 (range 3.4% to 5.3%; p=0.92) across centers. However, a significant difference in cardiac volumes and dosimetry was observed for GS-H vs. ABS-H and GS-H vs. C-H (Table 1). We identified consistent cardiac dosimetry across six separate departments within an integrated health system. However, we identified heart contouring variability between GS-H and C-H techniques, which translated into statistically significant dosimetric differences, though quantitatively small. While current protocols recommend contouring heart to include the pericardium, C-H commonly excluded the pericardium, which may account for the dosimetric variability compared to GS-H. ABS-H, derived from clinician contours, should be adapted to reflect current protocol definitions.Abstract TU_9_3409: Table 1Comparison of Volume and Dosimetric Parameters Between Contouring MethodsGS-HABS-Hp-valueGS-HC-Hp-valueMean Heart Dose (Gy)1.81.6<0.0011.81.7<0.001Mean D1cc (Gy)26.122.3<0.00126.124.4<0.001Mean V5 (%)4.83.7<0.0014.84.3<0.001Mean V30 (%)0.70.50.0020.70.60.004Mean Dice Coefficient0.920.97Mean Hausdorf Distance0.200.09 Open table in a new tab