Figure 2
From: Brain imaging signatures of neuropathic facial pain derived by artificial intelligence
CTN versus HC classification model performance (A) and analysis of variance of feature weights (B). The areas under the ROC curves are 0.93 and 0.99 for the RF and LR, respectively (p < 0.05). The RF model prioritized DTI measures, whereas LR was agnostic regarding imaging modality. CTN classical trigeminal neuralgia, HC healthy control, ROC receiver operator curve, LR logistic regression (red ROC), RF random forest (blue ROC), DTI diffusion tensor imaging.
