Identification of pretreatment habitat signatures for the prediction of patient outcome in triple negative breast cancer

Motivation: Triple negative breast cancer (TNBC) patients exhibit diverse response to therapy, with ~30% achieving pathological complete response (pCR). Goal(s): In this study, we seek to spatially-resolve heterogeneity of the tumor microenvironment using multiparametric MRI (mpMRI) in TNBC patients undergoing neoadjuvant chemotherapy to predict treatment outcomes. Approach: We employ habitat imaging, clustering mpMRI data to identify physiologically distinct tumor subregions, or habitats. Patients are then defined by tumor habitat composition and clustered to identify common habitat signatures. Associations between habitat signatures and patient outcomes are evaluated. Results: Clustering of patients yielded three habitat signatures with significantly different rates of pCR and recurrence-free survival. Impact: We demonstrate that tumor habitat signatures can differentiate triple-negative breast cancer patients prior to neoadjuvant chemotherapy, identifying those with improved treatment response and long-term outcomes. Clinical translation of this approach could enable patient stratification for therapy escalation/de-escalation and treatment optimization.