Fig. 1From: Prediction of pathologic stage in non-small cell lung cancer using machine learning algorithm based on CT image feature analysisPerformance assessment of prediction model in training and testing sets. a The imbalanced class distribution of NSCLC samples. b The final class distribution of NSCLC samples after equilibrium processing. c Confusion matrix was used to examine whether there is a consistency between the actual and the predicted results in NSCLC cohort. d Receiver operating characteristic (ROC) curve analysis for the prediction of the pathologic stages in NSCLC cohort. The corresponding reference groups are all the other stages patients. e Average precision score of prediction model in NSCLC cohort, micro-average over all classes: AP = 0.60. f Extension of precision-recall curve to multi-classes in NSCLC cohortBack to article page