Tumor neoangiogenesis is an important hallmark of cancer progression, triggered by alternating selective pressures from the hypoxic tumor microenvironment. Non-invasive, non-contrast-enhanced multiparametric MRI combining blood-oxygen-level-dependent (BOLD) MRI, which depicts blood oxygen saturation, and intravoxel-incoherent-motion (IVIM) MRI, which captures intravascular and extravascular diffusion, can provide insights into tumor oxygenation and neovascularization simultaneously. Our objective was to identify imaging markers that can predict hypoxia-induced angiogenesis and to validate our findings using multiplexed immunohistochemical analyses. We present an in vivo study involving 36 female athymic nude mice inoculated with luminal A, Her2+, and triple-negative breast cancer cells. We used a high-field 9.4-tesla MRI system for imaging and subsequently analyzed the tumors using multiplex immunohistochemistry for CD-31, PDGFR-β, and Hif1-α. We found that the hyperoxic-BOLD-MRI-derived parameter ΔR2* discriminated luminal A from Her2+ and triple-negative breast cancers, while the IVIM-derived parameter fIVIM discriminated luminal A and Her2+ from triple-negative breast cancers. A comprehensive analysis using principal-component analysis of both multiparametric MRI- and mpIHC-derived data highlighted the differences between triple-negative and luminal A breast cancers. We conclude that multiparametric MRI combining hyperoxic BOLD MRI and IVIM MRI, without the need for contrast agents, offers promising non-invasive markers for evaluating hypoxia-induced angiogenesis.

Objective: We evaluated whether a deep-learning model could predict the response to neoadjuvant chemotherapy (NAC) in breast cancer using the pre-treatment B-mode ultrasound. Methods: This retrospective study included 245 female patients (253 lesions) treated with NAC between 2017 and 2019. Lesions were categorized as complete response (CR; 103) or non-CR (150) based on postoperative pathology. We trained ResNet18-based models using pre-treatment B-mode ultrasound images (Image) and clinical features. Three training strategies were evaluated: training from scratch (SC); transfer learning (TL); and domain-specific pretraining (USP). Predictive performance was assessed using descriptive statistics. Results: The best-performing model (USP Image) achieved 0.76 accuracy (specificity: 0.80; sensitivity: 0.72), significantly outperforming all other models, including those that used additional clinical features ($p<0.05$). USP improved performance across most model types compared to SC and TL, highlighting the value of domain-specific pretraining. Clinical features added value with SC or TL, but not with USP, suggesting that pretrained models can extract the most relevant information directly from images. Grad-CAM analysis revealed that non-CR predictions focused on the tumor and posterior shadowing—features linked to chemoresistant subtypes. CR predictions focused mainly on more heterogeneous, peritumoral regions. Conclusions: This finding underscores ultrasound’s potential as a low-cost, accessible tool for predictive oncology in personalized, AI-driven treatment planning.