Predictive value of Cardiac Junction Volume for mean heart dose and its clinical applicability in left-sided breast radiotherapy

Background To evaluate the predictive performance of Cardiac Junction Volume (CJV), a contour-based three-dimensional anatomical metric, for mean heart dose (HEARTmean) in left-sided breast radiotherapy, and to assess its potential role as a contour-based screening tool applicable after contouring and prior to beam arrangement within tangential radiotherapy settings. Methods This retrospective study included 200 patients with left-sided breast cancer treated with tangential Intensity-Modulated Radiotherapy. Patients with incomplete contouring or dosimetric data were excluded. The cohort was divided into a derivation set (n = 150) and a validation set (n = 50). CJV was defined as the intersection volume between the heart and isotropically expanded Planning Target Volume (PTV), generated using 0.5 cm and 1.0 cm expansions (CJV-0.5 cm and CJV-1.0 cm). These expansions were used as analytical surrogate volumes to represent anatomically relevant dose-gradient regions for assessing target-heart spatial proximity prior to beam arrangement. Correlation analysis was performed to evaluate the association between CJV and HEARTmean. Predictive performance for identifying HEARTmean > 4 Gy was assessed using receiver operating characteristic (ROC) analysis, with the optimal cutoff determined by Youden’s index. Bootstrap confidence intervals, confusion matrix metrics, and threshold sensitivity analyses were performed. Maximum Heart Distance (MHD) was evaluated as a conventional comparator. A Knowledge-Based Planning (KBP) model was incorporated to assess whether the relationship between CJV and HEARTmean is preserved across different planning strategies. Results CJV-1.0 cm demonstrated a moderate correlation with HEARTmean (r = 0.540, p < 0.001), whereas CJV-0.5 cm showed a weak correlation. ROC analysis showed that CJV-1.0 cm achieved good discriminative performance (AUC = 0.787), with an optimal cutoff of 9.1 cm 3 , yielding a sensitivity of 71.4% and specificity of 76.5%. The Negative Predictive Value (NPV) was high (96.3%), indicating that patients with low CJV were unlikely to exceed the clinically relevant HEARTmean threshold. In the validation cohort, higher CJV values were consistently associated with increased HEARTmean across both manual and knowledge-based planning approaches. Although MHD demonstrated higher predictive performance (AUC = 0.908), CJV provided complementary anatomical information and enabled contour-based assessment after contouring and prior to beam arrangement within tangential radiotherapy settings. Conclusions CJV is a contour-based anatomical metric that reflects the volumetric spatial relationship between the target and the heart. Within tangential radiotherapy settings, CJV-1.0 cm demonstrated moderate predictive performance for HEARTmean. Although MHD showed higher predictive performance, CJV provides complementary anatomical information and can be determined after contouring and prior to beam arrangement. Therefore, CJV may serve as a practical screening tool to identify patients at low risk of exceeding cardiac dose thresholds, rather than as a definitive predictor.