Figure 2.

Overview of selected features. The machine learning model was fitted on the depicted grey matter features from randomly subsets of the training data. The selection included auditory but also various non-auditory brain areas (left panel). Restriction to the most frequently (i.e. in at least 99% of the samples) selected regions revealed that the selection pattern was highly consistent. This pattern consisted of anatomical features of lateral and medial prefrontal cortex, opercular cortex and postcentral gyrus, occipitotemporal cortex, superior temporal cortex, pallidum, thalamus and cerebellum (right panel). Colours indicate the frequency of being selected.