Preoperative Prediction of Spread Through Air Spaces in Lung Cancer Using PET/CT Radiomics and Peritumoral Microenvironment Features

Background/Objectives: Spread through air spaces (STAS) represents an aggressive invasion pattern in lung cancer and is associated with adverse oncologic outcomes. However, STAS is conventionally identified only after surgical resection, highlighting the need for reliable preoperative, noninvasive predictive approaches. Methods: In this retrospective study, patients who underwent surgical resection for lung cancer and had available preoperative ^18F-FDG PET/CT imaging were analyzed. Radiomic features were extracted from both intratumoral and peritumoral regions, the latter intended to reflect tumor microenvironment–related characteristics. Radiomic-only and clinicoradiomic models integrating clinical variables were developed using feature selection and multivariable modeling strategies, and their performance was evaluated using discrimination, calibration, and decision curve analyses. Results: Radiomic features derived from intratumoral metabolism and peritumoral tissue heterogeneity were associated with the presence of STAS. Integration of radiomic features with clinical parameters resulted in improved predictive performance compared with clinical models alone. The combined clinicoradiomic model demonstrated acceptable discrimination, calibration, and clinical utility across a range of threshold probabilities. Conclusions: Preoperative prediction of STAS in lung cancer is feasible using PET/CT-based radiomic analysis incorporating both intratumoral and peritumoral features. This noninvasive approach may provide biologically relevant information beyond anatomy-based assessment and supports further prospective validation of radiomic and clinicoradiomic models for STAS-oriented risk stratification.