Alterations of Cerebral Hemodynamics and Network Properties Induced by Newsvendor Problem in the Human Prefrontal Cortex

Figures

• 

  Figure 1. Experimental protocol and setup. (A) The NP experimental protocol consists of a 30-s initial baseline and 40 trials. Each trial has four phases: up to 20-s NP decision task (Di ), 5-s rest (R2i−1), 10-s feedback (Fi ), and 5-s rest (R2i ) again before starting another block/trial for Di+1 trial. (where i = 1, 2, 3, …, 40) Each block lasts about 30–40 s. (B) A demonstration of the fNIRS optode-setting helmet on a dummy head, holding 25 pairs of light sources (red dots) and 23 detectors (blue dots) with channel numbers marked covering the pre-frontal and frontal regions. (C) The entire optode layout for the whole-head helmet; the holes within the dashed frames were not used to hold any optode in this study while the other holes are color-matched to those shown in (B) on the forehead.

• 

  Figure 2. A schematic flow chart for dynamic graph theory analysis from the 1st trial to 5th trial. Ri denotes the rest period during the ith trial. Di denotes the decision period during the ith period. Rest (black lines) and decision periods (green lines) of three trials were extracted and concatenated into two time series separately in a three-trial moving window overlapping one trial (see that R3 and D3 are used twice for three-trial windows). Each concatenated time series were used to calculate PCC for the dynamic connectivity at each respective time. Then, correlation matrices from multiple three-trial windows were averaged for both rest (blue lines) and decision (red lines) periods, respectively. The final averaged correlation matrices (labeled as R-Rest and D-Decision) underwent Fischer r-to-z transform to calculate network parameters in GRETNA or Gretna for each subject.

• 

  Figure 3. Behavioral scores showing (A) an average of profit per trial and (B) an average reaction or decision-making time acquired from the LM and HM groups. “***”indicates the statistical significance between the two groups at p < 0.001 (p = 1.9 × 10−13). The reaction time did not show a significant difference. Outliers are marked by “+”.

• 

  Figure 4. Topographic β1 maps in response to brain stimulations by the 40-trial NP from all subjects (n = 27) without considering the treatment level. (A) Averaged topographic beta (or β1) map derived from Δ[HbO] during the decision-making phase. Dashed lines mark estimated regions of the three pf-ROIs: R-DLPDC, L-DLPDC, and OFC+FPA. (B) Group-averaged, area-summed β1 values for decision-making relative to baseline for OFC+FPA (p < 0.015), R-DLPFC and L-DLPFC (p < 0.008), respectively. “*” denotes p < 0.05 (corrected). Error bars: standard error of the mean.

• 

  Figure 5. Topographic β1 maps of Δ[HbO] in response to the 40-trial NP from groups with (A) HM (n = 13) and (B) LM (n = 14) treatment. Group-averaged, area-summed β1 values from groups with HM and LM decision-making tasks in (C) the R-DLPFC (p < 0.015), (D) the L-DLPFC, and (E) the OFC+FPA, respectively. “*” denotes p < 0.05 (corrected). Error bars: standard error of the mean.

• 

  Figure 6. Correlations between behavior performance (i.e., profit values) and area-summed β1 values of Δ[HbO] from the entire region of (A) R-DLPFC and (B) L-DLPFC. The data within the red boxes were derived from the LM group (n = 14). The solid lines are a linear fit across all the subjects (n = 27 without separating HM and LM tasks) with R2 value of 0.41 and 0.01, respectively, for both DLPFC areas. A significant linear correlation exists only for R-DLPFC at p < 0.014.

• 

  Figure 7. Comparison of averaged dynamic functional network properties derived from GTA during NP decision (red) and rest (blue) phases for all subjects (n = 27): (A) Eloc, (B) Cp, and (C) Lp considering sparsity ranging from 0.05 to 0.5, and (D) three respective network metrics quantified by the area under the curve (AUC) within the sparsity of 0.05–0.5 for the decision (red) and rest (blue) phases. In all panels, the error bars are the standard deviation. The * indicates the significant difference between the decision and rest phase for each network metric at p < 0.01.