Integrative radiomics clustering analysis to decipher breast cancer heterogeneity and prognostic indicators through multiparametric MRI
Yongsheng He, Shaofeng Duan, Wuling Wang, Hongkai Yang, Shuya Pan, Weiqun Cheng, Liang Xia, Xuan Qi
Abstract
Breast cancer diagnosis and treatment have been revolutionized by multiparametric Magnetic Resonance Imaging (mpMRI), encompassing T2-weighted imaging (T2WI), Diffusion-weighted imaging (DWI), and Dynamic Contrast-Enhanced MRI (DCE-MRI). We conducted a retrospective analysis of mpMRI data from 194 breast cancer patients (September 2019 to October 2023). Using 'pyradiomics' for radiomics feature extraction and MOVICS for unsupervised clustering. Interestingly, we identified two distinct patient clusters associated with significant differences in molecular subtypes, particularly in Luminal A subtype distribution (p = 0.03), estrogen receptor (ER) (p = 0.01), progesterone receptor (PR) (p = 0.04), mean tumor size (p < 0.01), lymph node metastasis (LNM) (p = 0.01), and edema (p < 0.01). Our study emphasizes mpMRI's potential in breast cancer by using radiomics-based cluster analysis to categorize tumors, uncovering heterogeneity, and aiding in personalized treatment strategies.