This study aimed to develop prognostic models of distant metastasis (DM) and bone metastasis (BM) in invasive breast cancer (IBC) incorporating whole-tumor radiomics, intratumoral heterogeneity (ITH) metrics. A total of 1097 patients with IBC from three centers were included and divided into a training cohort (n = 610), a validation cohort (n = 153), and two independent testing cohorts (n = 170 and n = 164). Whole-tumor radiomic features were extracted from two-dimensional regions of interest (ROIs) on mammography and three-dimensional ROIs on the second post-contrast phase of dynamic contrast–enhanced MRI, while ITH metrics were derived from enhancement-based subregions. Predictive performance was assessed for each model, and the resulting risk scores were further used for risk stratification analyses. SHapley Additive exPlanations (SHAP) were used to assess the importance of individual ITH metrics and to quantify the contributions of each component within the combined model. The combined model achieved the highest concordance index and robust risk stratification performance for both DM and BM. SHAP analysis indicated that the radiomics-based risk score ranked as the most influential predictor across both endpoints, while the ITH score consistently showed higher relative importance for BM than for DM. Distance-, shape-, and topology-related ITH metrics were among the most influential contributors across both endpoints. The combined model demonstrated robust prognostic performance, and the differential contribution of the ITH score between DM and BM indicates a more prominent role of imaging-derived heterogeneity in BM.
Sun et al. (Thu,) studied this question.