Metabolic characterization and radiomics-based composite model for breast cancer immune microenvironment types using ¹⁸F-FDG PET/CT.

Abstract

The intricateness of tumor immune microenvironment types (TIMTs) complicates identifying responders to immune checkpoint inhibitors (ICIs). Our purpose was to explore the metabolic characteristics of TIMTs in breast cancer using ¹⁸F-fluorodeoxyglucose (FDG) PET/CT and to establish radiomics-based predictive models for TIMTs. Consecutive 207 breast cancer patients (211 primary lesions), who underwent ¹⁸F-FDG PET/CT examination from Sep 2022 to Aug 2024 in our hospital, were retrospectively reviewed. The programmed death-ligand 1 (PD-L1) and tumor-infiltrating lymphocytes (TILs) were evaluated for TIMTs: TMIT-I (PD-L1-, TILs-), TMIT-II (PD-L1+, TILs+), TMIT-III (PD-L1-, TILs+), and TMIT-IV (PD-L1+, TILs-). The relationship between metabolic parameters (such as maximum standardized uptake value (SUVmax) and tumor-to-liver SUV ratio (TLR)) and TIMTs was analyzed. Then composite predictive models based on radiomics were further developed. TIMT-II represented the highest proportion in HER2+ (14/22, 64%) and triple-negative (17/27, 63%) breast cancer. Most metabolic parameters (such as SUVmax and TLR) exhibited significant differences in TIMT-II vs. -I or TIMT-II vs. -III (P < 0.05). TLR (P = 0.03; OR: 1.1) and Nottingham grade (P = 0.006; OR: 3.1) were independent impact factors of TIMT-II. We further developed a composite model that integrated radiomics, metabolic parameter, and clinicopathological data, which demonstrated promising predictive efficacy for TIMT-II (AUC _{testing set} = 0.86). Metabolic differences existed among different TIMTs, with TIMT-II exhibiting markedly elevated metabolic characteristics. The composite model based on radiomics demonstrated high predictive efficacy for TIMT-II and has the potential to screen ICIs responders.