Objective: This study aimed to develop a radiomics-based model grounded on the immune response evaluation criteria in solid tumors (iRECIST), in order to predict the immunotherapy combination therapy response in patients with non-small cell lung cancer (NSCLC) and to achieve precise identification of those who may derive clinical benefit at the pre-treatment or early treatment stage. Methods: A total of 75 patients with pathologically confirmed non-small cell lung cancer (NSCLC) who received immunotherapy combination therapy were enrolled in this study. All patients underwent chest CT examination within one month before treatment. Based on the iRECIST, treatment efficacy after 4–6 cycles (12–18 weeks) of immunotherapy combination therapy was assessed, and patients were classified into treatment-sensitive and non-sensitive groups. The patients were randomly divided into a training cohort (n = 50) and a testing cohort (n = 25) at a ratio of 2:1. Clinical feature models and radiomics prediction models were developed separately. The predictive performance and clinical utility of the models were compared using receiver operating characteristic (ROC) curves, the area under the curve (AUC), accuracy (ACC), and decision curve analysis (DCA). Results: After feature extraction and selection, a total of five radiomics features were obtained, including two first-order features and three texture features. After feature fusion and screening, and in order to reduce over-fitting, through three-fold cross-validation, the Naive Bayes was the most stable, with AUC of training set and test set being 0.675 and 0.647, and ACC being 0.720 and 0.720, respectively. Conclusion: Radiomics has certain potential and clinical application value for predicting the efficacy of PD-1 immunotherapy combination therapy in NSCLC patients.