Researchers developed and validated a CT-based radiomics model to predict pathological grade of non-small cell lung cancer (NSCLC) in 800 patients in a retrospective study. The whole tumor volume (WTV) was delineated by expanding 3 mm to the gross tumor volume (GTV) from the non-contrast CT scan, and habitat subregions were identified by K-means clustering. They used a two-stage binary classification model: Model-1 distinguished grade 3 from grades 1–2 and Model-2 distinguished grade 1 from grade 2. The models were based on logistic regression with features from WTV radiomics, habitat radiomics and clinical features. The Clf Habitat model achieved an AUC of 0.89 and 0.87 on the test set, a specificity of 0.73 for both models, and a BACC of 0.78 and 0.79. The combined Clf Total model performed best with AUC of 0.91 and 0.88, specificity of 0.84 and 0.77, and BACC of 0.82 and 0.81. The multimodal model provides robust performance and high specificity to support personalized treatment.