An analysis framework for the integration of broadband NIRS and EEG to assess neurovascular and neurometabolic coupling

Figures

• 

  Figure 1. Examples of poor quality (A) and good quality (B) ΔHbO2 (red), ΔHHb (blue), and ΔoxCCO (green) signals of two channels from one participant. Data are shown both in the time domain (left) and in the frequency domain (right). Yellow areas mark the visual task blocks.

• 

  Figure 2. Examples of the result of the application of the preprocessing steps to the ΔHbO2 (top), ΔHHb (middle), and ΔoxCCO (bottom) signals of one participant. Panel (A) shows the application of motion artifact correction method, where examples of artifacts are circled in light blue. Panel (B) shows the results of the FIR band-pass filtering. Raw uncorrected ΔHbO2/ΔHHb,/ΔoxCCO are shown in black,; processed ΔHbO2, ΔHHb, and ΔoxCCO are shown in red, blue and green respectively. Yellow areas mark the visual task blocks.

• 

  Figure 3. Grand-averaged responses (mean ± std. error) for the Right (top) and Left (bottom) visual hemifield condition, for each channel. Single-subject responses are estimated through the FIR-GLM and averaged across the group. The yellow area indicates the task block. ΔHbO2 is shown in red, ΔHHb in blue, ΔoxCCO (magnified by a factor of 5 for better visualization) in green. Red circles represent statistically significant ΔHbO2 changes, blue circles statistically significant ΔHHb changes, and green circles statistically significant ΔoxCCO changes (p < 0.005, FDR corrected for multiple comparisons).

• 

  Figure 4. Visual representation of rPWR and rCST for the Right condition. These parameters were computed from the median β-values which were z-scored across channels and then averaged across all participants, for HbO2 and oxCCO (left) and HHb and oxCCO (right) by rotating of 45° along the haemodynamics and metabolism axis, yielding to a rPWR axis (light blue) and rCST axis (orange). bNIRS channels located in the left hemisphere (# 1–8) are indicated by magenta circles and black crosses indicate channels located in the right hemisphere (# 9–16). The map is divided into four quadrants based on the direction and relative magnitude of the chromophores’ changes (greater increase or decrease) across channels.

• 

  Figure 5. t-values activation maps resulting from the group-level one sample t-test against zero performed on the rPWRHbO2 (top) and rPWRHHb (bottom). t-maps are presented for the Right (right) and Left (left) stimulation conditions. Positive t-values correspond to a positive association between HbO2/HHb and oxCCO (i.e., channels with a greater increase in HbO2 and oxCCO/greater decrease in HHb and increase in oxCCO) and negative t-values correspond to a negative association between HbO2/HHb and oxCCO (i.e., channels with a greater decrease in HbO2 and oxCCO/greater increase in HHb and decrease in oxCCO). Small shaded circles represent non-significant channels (p > 0.05), solid big circles represent significant channels (p < 0.05), and magenta circles indicate the significant channels (p < 0.05) surviving the FDR correction.

• 

  Figure 6. Cross-correlations between the EEG measure and the bNIRS signals at selected channels for (A) time window 1–16 s and (B) the time window where the maximum correlation between signals was observed, for the Right and Left Stimulation. For each bNIRS channel and EEG electrode combination, the significant maximum correlation (across subjects) is shown as colored circles with orange representing a strong positive correlation and purple representing a strong negative correlation, as shown by the color bar. On the y-axis, the bNIRS Channels 1–8 are channels over the left hemisphere and Channels 9–16 are channels over the right hemisphere. On the x-axis, EEG electrodes to the left of “Oz” (in the center of the x-axis) are electrodes over the left hemisphere and the electrodes to the right of “Oz” are electrodes over the right hemisphere. The transparent circles show correlations prior to FDR correction and the solid-colored circles indicate significance after FDR correction. The size of the circle represents the corresponding time lag with a positive time-lag indicating that EEG leads the bNIRS response. Note: The reader is advised that in the labelling of EEG channels, “T1” and “T2” do…

• 

  Figure 7. Median cross-correlations between EEG and bNIRS signals, across EEG channels, across bNIRS channels over the right and left hemispheres for all time windows from 1–16 to 10–25 s. (A) Median cross-correlations for Right and Left stimulations over bNIRS channels the right hemisphere, for all time windows. (B) Median cross-correlations for Right and Left stimulations over bNIRS channels the left hemisphere, for all time windows. The median value is shown and the lines represent minimum and maximum correlations at each time window. The correlations that were significant post-FDR correction are indicated with a larger, solid circle.

• 

  Figure 8. Visual stimulation task protocol. Visual stimuli comprise Right and Left hemifield reversing checkerboards spaced out by 18 s rest periods.

• 

  Figure 9. (A) Schematic representation of channel configuration over the visual cortex. Sources are represented by red circles, detectors by blue circles, and the 16 measurement channels by white circles. The EEG channels used for the combined bNIRS-EEG analysis are shown in orange. (B) Locations of the 26 EEG electrodes. The electrodes used for the combined bNIRS – EEG analysis are circled in orange. (C) Picture of the bNIRS probe holder. It is made of 3D printed optode holders and a neoprene band moulded to follow the curvature of the participants’ head.

• 

  Figure 10. Flow chart of the data analysis pipeline for bNIRS (top) and EEG (bottom).

• 

  Figure 11. Procedure for combining the bNIRS and EEG data. (A) An EEG measure (mean gamma power across the task period) is derived from the EEG data that is used to cross-correlate with the bNIRS signals. The data shown here is from a single subject for the purposes of outlining the analysis procedure. (B) The procedure for calculating cross-correlations is demonstrated for the EEG and the NIRS data from time window 5–20 s, for selected time points, for illustrative purposes. For the positive time-lags, the cross-correlation is performed by shifting the NIRS signal forward with respect to the EEG for each time-lag. For example, for time-lag = 1, EEG time point t = 2 is correlated with bNIRS signal time point t = 1. For time-lag = 2, EEG time point t = 3 is correlated with bNIRS signal time point t = 1, i.e. the bNIRS signal is shifted forward for each positive time-lag. For the negative time-lags, the bNIRS signal is shifted back such that at time-lag = -1, EEG time point t = 1 is correlated with bNIRS time point t = 2. For time-lag = − 2, EEG time point t = 1 is correlated with bNIRS time point t = 3, i.e. the bNIRS signal is shifted backwards for each negative time-lag.