Cluster analysis of quantitative ultrafast DCE-MRI for prediction of breast cancer response to neoadjuvant chemotherapy

Motivation: Early prediction of pathologic complete response (pCR) enhances personalized breast cancer treatment. Goal(s): This study aims to measure kinetic parameters from pre-neoadjuvant chemotherapy (NAC) ultrafast dynamic contrast-enhanced (DCE)-MRI using k-means clustering (KMC) to predict pCR. Approach: Fifty-six patients undergoing NAC were enrolled. Tumor and normal parenchymal voxels were divided into five clusters based on maximum enhancement rate. Ipsilateral/contralateral (I/C) background parenchymal enhancement kinetics (kBPE) parameters were compared between pCR and non-pCR patients. A logistic regression model incorporating clinical features. Results: MRI tumor kinetics, and kBPE I/Cs achieved an AUC of 0.94, with 0.91 sensitivity and specificity, for predicting breast cancer response to NAC. Impact: K-means clustering analysis of ultrafast DCE-MRI is a stable technique to effectively predict treatment response in breast cancer patients prior to neoadjuvant chemotherapy, which facilitates personalized therapy adjustments and can improve clinical outcomes through individualized treatments.