Fig 2. Significant transfer learning accuracy between physiological and pharmacologically-induced states of unconsciousness (N3 sleep and LoC), but not between them and pathological states of unconsciousness (UWS).

(A) Average across subject FC differences for N3 vs. wake (left), LoC vs. wake (center), and UWS vs. wake (right), together with anatomical renderings of the top (red) and bottom (blue) 5% functional connections associated with the largest difference between states in absolute value. (B) Nodes in the diagram represent different brain states (N3, LoC, and UWS) and the arrows between them indicate that a machine learning classifier trained to distinguish the source state from W presented significant transfer learning accuracy when distinguishing the target state from W (p<0.05, random label shuffling with 1000 iterations, Bonferroni corrected).