Fig. 3: Behavioral correlates of the observed oscillation states.
From: Deciphering neuronal variability across states reveals dynamic sensory encoding
A Face motion energy evaluated as the absolute value of the difference between consecutive frames. B Eye and pupil tracking. Tracking points were identified using a universal tracking model trained in DeepLabCut. C Animal pose estimation. Specific, visible body parts were tracked using a universal tracking model trained in SLEAP. D Example snippet of behavioral changes alongside the animal’s current oscillation state. SH: High-frequency state (green), SI: Intermediate state (blue), and SL: Low-frequency state (pink). E Comparison of the average movement of specific body parts across states (\({p}_{{S}_{H},{S}_{I,L}}\), pupil size: p = 2.8e-15, running: p = 2.0e-17, face motion: p = 6.3e-13, body center: p = 2.6e-18, left forelimb: p = 1.2e-13, left hindlimb: p = 4.9e-14, right hindlimb: p = 3.0e-11, tail start:, p = 3.0e-16, tail end: p = 2.0e-11, n = 25 mice, one-way ANOVA). F, Mutual information (MI) between behavioral variables and the inferred HMM states (mean ± sem, n = 25 mice). All statistical tests were performed using one-way ANOVA. Statistical tests in (E, F) were adjusted for multiple comparisons using the Bonferroni correction (***: p < 0.0001, **: p < 0.001, *: p < 0.05). Source data are provided as a Source Data file.
