Background and purpose: The purpose of this paper was to explore the factors associated with non-small cell lung cancer (NSCLC) patient’s recurrence situation based on CT image features. Methods: A hundred and fifty-seven sets of data collected in NSCLC radiogenomics database were used in the experiment. The lung tumors were segmented

and image features were extracted. Independent samples t test was used to perform a univariate analysis. And logistic regression model was used to obtain the significant factors associated with NSCLC recurrence. Z-score normalization and synthetic minority over-sampling technique (SMOTE) methods were used to analyze data. Finally

random forest

K-nearest neighbor (KNN)

support vector machine (SVM)

decision-tree and leave-one-out cross validation were used to train classifier and test the validity of results. Results: The independent samples t test showed that Variance

Energy

Relative message

Add-entropy and Coarseness were related to NSCLC recurrence (P0.05). And the logistic regression analysis showed that Energy and Add-entropy were significantly correlated with NSCLC recurrence (P0.05). Furthermore

the classification results revealed that the best accuracy was 82.7% and the maximum area under curve (AUC) was 0.891. These two features could make a well prediction for NSCLC patient’s recurrence. Conclusion: Energy and Add-entropy were the factors significantly associated with NSCLC recurrence.