Integrative analysis of DCE-MRI radiomics and genomics reveals complementary characterization of treatment response–associated tumor vascular heterogeneity in breast cancer

Dynamic contrast-enhanced MRI (DCE-MRI) and vascular-related gene expression characterize tumor vasculature the blood vessel network that supports tumor perfusion and drug delivery from macroscopic imaging and microscopic molecular perspectives. We aimed to investigate the complementary vascular heterogeneity captured by these modalities. Paired pre-treatment DCE-MRI and mRNA expression from 932 patients with high-risk stage II/III breast cancer were analyzed. Imaging and gene subgroups were identified separately by k-means clustering of repeatable radiomic features and vascular-related genes. The complementarity of imaging and gene subgroups was assessed by examining their association with each other, comparing their distributions across receptor-defined and PAM50 subtypes, pathway enrichment patterns, associations with pathological complete response (pCR) after adjustment for HR/HER2 status, and distant recurrence-free survival (DRFS) stratification. Three imaging subgroups (image subgroup 1 -subgroup 3: 24.8%, 34.3%, and 40.9%) and three gene subgroups (gene subgroup 1 -​subgroup 3: 33.6%, 29.2%, and 37.2%) were identified, with a weak association between them (Cramér’s V = 0.068; p = 0.073). The imaging subgroups showed weak associations with receptor-defined and PAM50 subtypes (Cramér’s V = 0.073 and 0.115), whereas gene subgroups showed stronger associations (Cramér’s V = 0.306 and 0.418). Distinct pathway enrichment patterns were observed between imaging and gene subgroups, with metabolic pathways and stromal signaling associated with imaging subgroups, and DNA repair, oxidative metabolism, and immune signaling associated with gene subgroups (FDR q < 0.05). Imaging and gene subgroups showed different pCR association patterns, with image subgroup 2 associated with higher pCR (adjusted OR = 1.409; LR p = 0.027), while gene subgroup 1 and gene subgroup 2 were associated with lower and higher pCR, respectively (adjusted OR = 0.597 and 2.217; LR p < 0.003). Imaging subgroups showed higher prognostic discrimination than gene subgroups (C-index = 0.624 vs. 0.589; overall log-rank p = 0.008 and 0.015, respectively). DCE-MRI radiomic and vascular genomic subgroups capture complementary dimensions of tumor vascular heterogeneity. Their distinct biological and clinical associations support the potential value of integrating imaging- and gene-based vascular phenotypes for patient stratification in breast cancer.