LAUSR

Simple Summary Accurate and early selection of patients with advanced non-small-cell lung cancer (NSCLC) who would benefit from immunotherapy is of the utmost clinical importance. The aim of our retrospective multi-centric study was to determine the potential role of CT-based radiomics machine learning models in predicting treatment response and survival in patients with advanced NSCLC treated with immune checkpoint inhibitors. Our delta-radiomics signature was able to identify patients who presented a clinical benefit at 6 months early, with an AUC obtained on an external test dataset of 0.8 (95% CI: 0.65−0.95). Abstract The aim of our study was to determine the potential role of CT-based radiomics in predicting treatment response and survival in patients with advanced NSCLC treated with immune checkpoint inhibitors. We retrospectively included 188 patients with NSCLC treated with PD-1/PD-L1 inhibitors from two independent centers. Radiomics analysis was performed on pre-treatment contrast-enhanced CT. A delta-radiomics analysis was also conducted on a subset of 160 patients who underwent a follow-up contrast-enhanced CT after 2 to 4 treatment cycles. Linear and random forest (RF) models were tested to predict response at 6 months and overall survival. Models based on clinical parameters only and combined clinical and radiomics models were also tested and compared to the radiomics and delta-radiomics models. The RF delta-radiomics model showed the best performance for response prediction with an AUC of 0.8 (95% CI: 0.65−0.95) on the external test dataset. The Cox regression delta-radiomics model was the most accurate at predicting survival with a concordance index of 0.68 (95% CI: 0.56−0.80) (p = 0.02). The baseline CT radiomics signatures did not show any significant results for treatment response prediction or survival. In conclusion, our results demonstrated the ability of a CT-based delta-radiomics signature to identify early on patients with NSCLC who were more likely to benefit from immunotherapy.