Figure 7. Analysis of hemodynamic behavioral correlation states and comparisons between hemodynamic and neural signal representations.

(A) Behaviorally defined locomotion and resting correlation states derived from hemodynamics(Δ[HbT]) forone example mouse, calculated similarlytothe neural states in Figure 3B.

(B) Average coefficients predicted by NNLS models using hemodynamic states in (A) as a basis set (fitting to real-time hemodynamic 10-s moving-window correlation maps) across five mice (n = 54 [onset], 47 [locomotion], and 56 [offset] epochs), mean ± SEM.

(C) Differences between sustained- and initial-rest states; (ii) includes only statistically significant differences (using a Wilcoxon rank-sum test, p < 0.05, Bonferroni corrected).

(D) Negative correlation values between the initial-rest and sustained-rest NNLS coefficients.

(E) Probability density of correlation values between pupil size and NNLS coefficients for initial and sustained rest calculated using kernel density estimation (p < 0.0001, Kolmogorov-Smirnov test).

(F) Comparison of spectral power of neural and hemodynamic signals for the anterior frontal lateral (top) and posterior (visual, bottom) ROIs, over 60-s duration rest periods (n = 182, in five mice, mean ± SEM).

(G) (i) Comparing neural (left) and hemodynamic (right) correlation maps for different frequency bands. Maps show correlations over 10-s windows during sustained-rest averaged over n = 63 epochs in one example mouse. All siganls were zero-mean adjusted and then temporally filtered over different frequency bands from low to high (top to bottom) (see STAR Methods). (ii) Time courses show example neural and hemodynamic signals temporally filtered over the same ranges as in (i) for one sustained-rest epoch extracted from the two ROIs indicated in (F). The r values show Pearson correlation coefficients between the neural and hemodynamic signals.

(H) Euclidean distances between correlation maps for each frequency band in (G) relative to the 0.02–0.25 Hz frequency band for neural (top) and hemodynamic (bottom) measurements.

(I) Euclidean distances between the neural and hemodynamic correlation maps at different frequency bands in (G). In (H) and (I), Kruskal-Wallis tests were used to compare groups (n = 5 mice, p < 0.05). See also Figures S6–S8 and Video S4.