Figure 6.

Training settings of the classifier. In setting A, data from healthy volunteers is appended to the patient dataset to increase its size. In setting B, only patient data is used to train the classifier model. Finally, in setting C, data from the patient population is used to find the most relevant features and build the mapping; then, the model is trained with healthy data only. In all cases, only patient data is employed to generate the predictions. As one would expect, the data used for tests will come from a different set than the one used for training.