Figure 1.
(Left) Locations of the five ROIs (t statistics of the BOLD signal averaged between 4.0 s and 7.0 s after the visual stimulus onset) in each hemisphere.
(Right) Hemodynamic time courses and estimated neuronal activity using hemodynamic deconvolution at five ROIs.
Table 1.
Timing indices and the full-width-half-maximum (FHWM) of the group-average hemodynamic responses in five regions-of-interest.
Figure 2.
The dominant information flow calculated from the difference between two uni-directional Granger estimates among the visual (V), PPC, premotor (PreM), somatosensory (S), and motor (M) cortex ROIs at TR = 0.1 s, 0.5 s, 1 s, and 2 s.
Significant causal modulations (p≤0.05) were shown in thick yellow arrows and connections showing a trend of causal modulation (0.05<p≤0.1) were shown in thin yellow arrows.
Table 2.
The optimal order of the AR model of the time series at five ROI’s.
Table 3.
P-values of all directions of information flow estimated by Granger causality between ROI’s at different sampling rates from 1000 bootstrap iterations.
Table 4.
P-values of Granger causality values between ROI’s at different sampling rates.
Table 5.
P-values of all directions of information flow estimated by conditional Granger causality between ROI’s at different sampling rates from 100 bootstrap iterations.
Figure 3.
The dominant information flow calculated from the difference between two uni-directional Granger estimates among the visual (V), PPC, premotor (PreM), somatosensory (S), and motor (M) cortex ROIs at TR = 0.1 s, 0.5 s, 1 s, and 2 s.
The time series at 0.5 s, 1 s, and 2 s were SINC interpolated such that all time series are of the same length as the 0.1 s time series, which contains the real measured data. Significant causal modulations (p≤0.05) were shown in thick yellow arrows and connections showing a trend of causal modulation (0.05<p≤0.1) were shown in thin yellow arrows. Importantly, only very little improvement is observed in the estimates where the number of observations is artificially increased using the SINC interpolation procedure: the strongest connectivity patterns are, clearly, observed in the 0.1-s condition with only real data.
Figure 4.
The EPI (left column) and InI (right column) time series at the visual and sensorimotor cortices from a representative subject (top panel) and the residual EPI (left column) and InI (right column) time series at the sensorimotor cortex after AR modeling using the sensorimotor cortex time series alone and both sensorimotor and visual cortices (bottom panel).