Early prediction of neoadjuvant chemotherapy efficacy for mass breast cancer based on dynamic contrast-enhanced magnetic resonance imaging radiomics

Radiomics uses automated algorithms to extract high-order features from images, which can contribute to clinical decisions such as therapeutic efficacy evaluation. We assessed the value of a dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI)-based radiomics model for predicting pathological complete response (pCR) after a second cycle of neoadjuvant chemotherapy (NAC) in patients with mass breast cancer. We retrospectively analyzed data from 149 patients with mass breast cancer who underwent NAC between January 2017 and December 2022. Using DCE-MRI, before NAC and after a second cycle of NAC, the least absolute shrinkage and selection operator and logistic regression (LR) algorithms were applied for feature selection and radiomics modeling. We found significant differences in two clinical imaging features (molecular subtypes, background parenchymal enhancement changes) and two radiomics features. Clinical and radiomics features were employed to build clinical, radiomics, and combined models to predict pCR. The LR model that combined clinical and radiomics features had an area under the curve of 0.811, higher than that for the imaging or radiomics model. Our findings suggest that a combined model based on imaging and radiomics features can improve early prediction of NAC efficacy for patients with mass breast cancer. Anticipating the likelihood of pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) is an important prognostic factor for disease-free and overall survival. A combined model based on clinical features and radiomics improves the predictive performance of early prediction of pCR with NAC in patients with mass-based breast cancer. image