18F-FDG PET/CT radiomics predicts brain metastasis in I-IIIA resected Non-Small cell lung cancer

Objective: To establish 18F-FDG PET/CT radiomics model for predicting brain metastasis in non-small cell lung cancer (NSCLC) patients.Methods: This research comprised 203 NSCLC patients who had received surgical therapy at two institutions. To identify independent predictive factors of brain metastasis, metabolic indicators, CT features, and clinical fea-tures were investigated. A prediction model was established by incorporating radiomics signature and clinico-pathological risk variables. The suggested model's performance was assessed from the perspective of discrimination, calibration, and clinical application.Results: The C-indices of the PET/CT radiomics model in the training, internal validation, and external validation cohorts were 0.911, 0.825 and 0.800, respectively. According to the multivariate analysis, neuron-specific enolase (NSE) and air bronchogram were independent risk factors for brain metastasis (BM). Furthermore, the combined model integrating radiomics and clinicopathological characteristics related to brain metastasis per-formed better in terms of prediction, with C-indices of 0.927, 0.861, and 0.860 in the training, internal vali-dation, and external validation cohorts, respectively. The decision curve analysis (DCA) suggested that the PET/ CT nomogram was clinically beneficial.Conclusions: A predictive algorithm based on PET/CT imaging information and clinicopathological features may accurately predict the probability of brain metastasis in NSCLC patients following surgery. This presented doctors with a unique technique for screening NSCLC patients at high risk of brain metastasis.