Fig. 3

Sensitivity and specificity of HMM states and dynamics within polysomnography stages. a Fractional occupancies of each of the 19 HMM states computed within the four PSG stages corresponded to the PSG-sensitivity of the whole-brain network states. The coloured bars and error bars show the average and standard error, respectively, across the 18 participants that included all four PSG stages. b PSG-specificity of the HMM states for each of the four PSG stages. Specificity corresponds to the probability of an HMM state occurring within a PSG stage. The bars represent the group average and the error bars the standard error (n = 18). In a and b horizontal lines show significant differences within HMM states, with p values < 0.01 as evaluated through paired t tests and permutation testing. c The mean life times of the 19 HMM states are shown by the bars, representing values averaged across the 18 participants. Error-bars represent the standard error across participants. Each HMM state is coloured according to the probability of finding it within each of the four PSG stages, i.e., their PSG specificity. Note how HMM states with high specificity for N3—and to a lesser extent N2—exhibit longer mean life times. d The dynamics of the HMM transitions were calculated within each of the four PSG stages, in terms of switching frequency (‘Switching’), and e the number of different HMM states visited per time (‘Range of HMM states’). These measures significantly separate the four PSG stages suggestive of a higher dynamical repertoire during wakefulness and N1. In d and e error bars represent standard error across participants and significant differences between PSG stages are denoted by stars: one star: p < 0.05, two stars: p < 0.01 and three stars p < 0.001; all evaluated using paired t tests and permutations. W: wakefulness, N1: N1 sleep, N2: N2 sleep, N3: N3 sleep