AI maps immune cells in breast tumors to predict treatment response
Researchers used machine learning to analyze immune cell patterns in breast cancer biopsies, finding that the location and density of immune cells can predict which patients will respond to chemotherapy. The discovery could help oncologists personalize treatment and potentially avoid unnecessary interventions in patients unlikely to benefit.
Originaltitel: Machine learning-based spatial characterization of tumor-immune microenvironment in the EORTC 10994/BIG 1-00 early breast cancer trial
<p>Breast cancer (BC) represents a heterogeneous ecosystem and elucidation of tumor microenvironment components remains essential. Our study aimed to depict the composition and prognostic correlates of immune infiltrate in early BC, at a multiplex and spatial resolution. Pretreatment tumor biopsies from patients enrolled in the EORTC 10994/BIG 1-00 randomized phase III neoadjuvant trial (NCT00017095) were used; the CNN11 classifier for H&E-based digital TILs (dTILs) quantification and multiplex immunofluorescence were applied, coupled with machine learning (ML)-based spatial features. dTILs were higher in the triple-negative (TN) subtype, and associated with pathological complete response (pCR) in the whole cohort. Total CD4+ and intra-tumoral CD8+ T-cells expression was associated with pCR. Higher immune-tumor cell colocalization was observed in TN tumors of patients achieving pCR. Immune cell subsets were enriched in TP53-mutated tumors. Our results indicate the feasibility of ML-based algorithms for immune infiltrate characterization and the prognostic implications of its abundance and tumor-host interactions.</p>