Impact: Interative nested habitat analysis for breast DCE-MRI improves benign–malignant classification beyond whole-tumor models, while delivering spatial risk maps and physiology-aligned perfusion summaries. Iterative refinement with early stopping limits overfitting, generalizes across cohorts, and supports reader studies and clinical decision-support integration.

Impact: This study reinforce the role of breast MRI not only in detection and staging, but also as a complementary tool for risk stratification and therapeutic guidance in biologically aggressive subtypes of breast cancer, especially triple-negative tumors.

Impact: Radiomics can indeed improve the diagnostic efficacy of emerging MR cytometry. IMPULSED-derived microstructural parameters and time-dependent ADC measurements provide complementary information, combining both sets of metrics facilitates clinical decision-making.

Impact: This non-invasive model could serve as an alternative to biopsy for assessing PD-L1 status across the entire tumor. It may improve patient selection for immunotherapy, potentially enhancing treatment efficacy and progression-free survival in breast cancer.

Impact: Enables noninvasive breast cancer NAC response prediction via td-dMRI, bridges microstructural MRI parameters with underlying gene expression, and guides precision therapy.

Impact: Important pathological and MRI characteristics could help distinct high/low TILs level, assist the selection of candidates who may benefit from immunotherapy, and also improve the prediction of patients’ survival with different TILs expression.

Impact: Our results suggested whole-tumor histogram parameters of quantitative synthetic MRI (syMRI) could serve as non-invasive biomarkers for identifying HER2-zero, HER2-low, and HER2-oe. The application of syMRI would be beneficial for guiding treatment selection and monitoring HER2 status change during treatment.

Impact: Therapy response at mid-treatment may be determined by a neural network applied to a single multi-parameter mapping sequence, which may help guide treatment strategies for improved outcomes for patients with triple negative breast cancer.

Impact: This study highlights the potential of longitudinal MRI for early prediction of NAC response in breast cancer, providing a feasible non-invasive tool for personalized treatment decisions, and lays the groundwork for future development of multimodal predictive models.

Impact: Accurate early prediction of treatment response can provide a window of opportunity for personalised treatment. This research explores the potential of early non-invasive tumour tracking using imaging and circulating tumour DNA for prediction of neoadjuvant chemotherapy response in breast cancer.

Impact: The proposed AI model enhances detection of breast restricted diffusion on whole-body MRI, improving efficiency and reader confidence, and demonstrating potential to support earlier identification of clinically significant findings.

Impact: Ultrafast DCE-MRI, combined with DWI, enhances diagnostic accuracy and reduces scan times for breast cancer screening. This efficient approach not only improves early detection but also paves the way for future AI-driven MRI screening, optimizing clinical workflows and patient outcomes

Impact: A deep learning based automated pipeline is developed for breast anthropometrical measurements, which can now be used for establishing breast and chest measurements at population level based on their MRI studies. This could aid in better breast coil design.

Impact: Adding tumor histogram parameters of DWI and syMRI with clinic results can significantly improve the performance of ALNM prediction. Therefore, application of DWI and syMRI would be beneficial for preoperatively guiding clinical decision-making in breast cancer patients.

Impact: Quantitative CEST and MT tissue mapping with high-resolution can improve characterization of breast cancer and eliminate the use of a contrast agent.