Figure 1.
Spatio-frequency connectivity matrix.
(A) A schematic diagram of spatio-frequency connectivity matrix, with the horizontal axis labeling all the channels and the vertical axis indexing all the frequencies. (B) Frequency distribution of the connectivity between a channel y and the seed channel. It corresponds to one column of the spatio-frequency connectivity matrix. (C) Spatial map of connectivity to the seed channel at a specific frequency point. It is redrawn from one row of the spatio-frequency connectivity matrix.
Figure 2.
(A, C, E) Temporal-frequency characteristics of the simulated dataset. (B, D, F) Spatial characteristics of the simulated dataset. denotes the spontaneous fluctuations,
the global noise 1 (i.e., the systemic physiological noises), and
the global noise 2 (i.e., the motion artifact).
Figure 3.
RSFCdf results of the simulated experiment with contamination of the systemic physiological noise.
(A) An example of the average time course from the two ROIs (shown in Fig. 2B) for the traditional coherence analysis. (B) The mean coherence curve between the two ROIs (i.e., ) over 100 times of stimulations derived from the traditional method. The coherences at frequency >0.25 Hz were not plotted because of very low values. The black line represents the standard deviation. The green area indicates the
(i.e., 0.01∼0.1 Hz), and the yellow and the orange area indicate the frequencies lower or higher than
, respectively. The blue horizontal bar indicates the
. (C) An example of the spatio-frequency connectivity matrix (bottom layer) and an enlargement over the low frequency portion below 0.3 Hz (top layer). (D) The mean spatially weighted coherence curve over 100 times of stimulations derived from the proposed method (i.e.,
). The black line represents the standard deviation. The red horizontal bar indicates the estimated
. (E) ROC curves for RSFCdf determination of the two methods with blue for the traditional method and red for the proposed.
Figure 4.
RSFCdf results of the simulated experiment with contamination of the motion artifact.
(A) Semilog plot of the mean coherence curve between the two ROIs over 100 times of stimulations derived from the traditional method (i.e., ). The black line represents the standard deviation. The green area indicates the
(i.e., 0.01∼0.1 Hz), and the yellow and the orange area indicate the frequencies lower or higher than
, respectively. The blue horizontal bar indicates the
. (B) Semilog plot of the mean spatially weighted coherence curve over 100 times of stimulations derived from the proposed method (i.e.,
). The black line represents the standard deviation. The red horizontal bar indicates the estimated
. (C) ROC curves for RSFCdf determination of the two methods with blue for the traditional method and red for the proposed.
Figure 5.
The robustness of the proposed method to the possible inaccurate spatial template.
(A) The results with the expanded spatial template in which the size of the functional system of interest is overestimated. (B) The
results with the shrunk spatial template in which the size of the functional system of interest is underestimated. (C) The
results with the spatial template which inaccurate estimate the location of the functional system of interest. The dotted line represents the case of template with 10% estimated errors. The dashed line indicates the case of template with 20% estimated errors. And the solid line is for the original template without errors. The horizontal arrow indicates the estimated
range for each case.
Figure 6.
Comparison between the proposed and traditional methods on the real resting-state fNIRS dataset.
(A) Schematics of the fNIRS probe array and the anatomical localizations of the fNIRS channels. The labeling maps with green indicate the channels covering the sensorimotor area. (B) The coherence curve of the bilateral sensorimotor areas across all the subjects derived from the traditional method. The error bar indicates the standard error. The horizontal blue arrow at the bottom indicates the estimated . (C) The group results of the proposed method with four typical frequency-specific connectivity maps. The error bar indicates the standard error across subjects, and the horizontal red arrow indicates the estimated
.