Characterizing Tumor Habitats in Breast Cancer: A DKI Approach to Predicting Estrogen Receptor Expression

Motivation: This study assessed estrogen receptor (ER) expression using a non-invasive method based on diffusion kurtosis imaging (DKI) habitat analysis to reveal tumor microenvironment characteristics. Goal(s): We aimed to investigate the relationship between tumor habitats and ER expression, focusing on microstructural complexity and water diffusion. Approach: The predictive model employed segmentation and unsupervised clustering of DKI data, highlighting spatial variations within tumor subregions. Results: We identified four distinct tumor subregions, with significant differences in mean kurtosis (MK) and mean diffusivity (MD), indicating that the model effectively predicted ER expression based on microenvironment heterogeneity. Impact: This study demonstrates that diffusion kurtosis imaging can non-invasively predict estrogen receptor expression by characterizing tumor microenvironments, revealing significant relationships between tumor habitats and their microstructural properties.