Fig. 5.

Results of applying the feature selection method based on MIC to the texture dataset of LBP features extracted from T2W images of subcortical structures (SS) when training the SVM model with cost C = 8. The profile of AUC values obtained for all possible subsets of features according to the MIC ranking is illustrated in (a). The ROC curves provided by the model when using all the features (14 features) and when using the optimal number of features (7 features) is shown in (b).