Skip to main content
. 2020 Feb 29;9(3):658. doi: 10.3390/jcm9030658

Figure 2.

Figure 2

The testing set classification receiver operating characteristic (ROC) curves showing the best results of (a) extreme gradient boosting (XGB)- generalized fractional anisotropy (GFA), (b) XGB- isotropic value of the orientation distribution function (ISO), (c) XGB- normalized quantitative anisotropy (NQA), (d) logistic regression (LR)-GFA, (e) LR-ISO, and (f) LR-NQA following autoencoder and supervised machine learning analysis, along with the area under curves (AUCs) (AUC: XGB-GFA-1 = 0.92; XGB-GFA-2 = 0.94; XGB-ISO-1 = 0.91; XGB-ISO-2 = 0.89; XGB-NQA-1 = 0.90; XGB-NQA-2 = 0.75; LR-GFA-1 = 0.93; LR-GFA-2 = 0.93; LR-ISO-1 = 0.85; LR-ISO-2 = 0.92; LR-NQA-1 = 0.92; LR-NQA-2 = 0.79).