Motivation: Identifying optimal responders to neoadjuvant chemotherapy (NAC) in breast cancer before treatment is challenging but crucial for optimizing therapeutic strategies. Goal(s): Our goal was to evaluate if combining pretreatment DCE-MRI and ADC maps, with or without clinical factors, can predict treatment response in breast cancer. Approach: This retrospective study employed a rigorous feature selection process to identify key clinical and radiomic features from semi-automatically segmented tumor regions on pretreatment MR images, followed by multivariate logistic regression to assess models' predicitve performance. Results: DCE-MRI and ADC independently provided valuable insights, with their combination enhancing predictive performance. Integrating clinical features further improved model performance. Impact: This study demonstrates the potential of combining pretreatment DCE-MRI, ADC maps, and clinical factors in predicting NAC repsponse in breast cancer, offering a non-invasive approach to guide personalized treatment strategies, ultimately improving patient outcomes and reducing unnecessary interventions for non-responders.
Tao et al. (2025) studied this question.