Fig. 3

Macroaverage curves of these ML models in different MRI sequences. The performance of XGBoost model on T1-weighted (D), T2-weighted (E) and contrast-enhanced T1-weighted (F) was better than that of SVM model (A-C) and LR model (G-I) constructed with the corresponding features. The average AUC values of the XGBoost model on T1-weighted, T2-weighted, and contrast-enhanced T1-weighted sequences were 0.8523, 0.9312, and 0.9560, respectively; for the SVM model, the corresponding values were 0.8460, 0.9247, and 0.9381; and for the LR model, the values were 0.8200, 0.9220, and 0.9285, respectively