Abnormal Enhanced Gamma Synchronization in the Default Mode Network Associated With Chronic Pain

Figures

• 

  FIGURE 1. Global and local brain activity. (A) Global power spectra for chronic pain patients (CP, orange line) and healthy controls (HC, blue line), shown for frequencies from 1 to 70 Hz with a bandstop filter at 48 to 52 Hz to attenuate line noise. (B) Violin plot of PAF averaged across all channels or cortical areas. (C) Scalp topographical map displaying power differences between the CP and HC groups in delta, theta, alpha, beta, gamma, LF, and HF frequency bands. The color map shows the t‐values comparing CP to HC. The white line indicates channels where power is reduced in CP patients. (D) Visualization of cortical areas with group differences in source space. The displayed nodes represent cortical regions in CP patients with decreased power in the beta, gamma, and HF bands. (E) Regional power analysis of the HF band in electrode space. (F) Regional power analysis of the HF band in source space. Local analysis results were statistically adjusted using the FDR method with significance set at *p < 0.05.

• 

  FIGURE 2. Global and local functional connectivity based on PLV in gamma band. (A) Scalp topographical map of gamma PLV connectivity in electrode space. (B) Brain topographical map of gamma PLV connectivity in source space. Red lines illustrate increased FC, while node colors represent corresponding parcellation regions. (C, D) Gamma PLV inter‐region connectivity in electrode and source spaces, showing significantly elevated PLV connectivity between temporal and occipital regions in chronic pain (CP) group compared to healthy control (HC) group. F, C, L, T, P, and O denote frontal, central, limbic, temporal, parietal, and occipital regions, respectively. Corrections for both global and inter‐region PLV connectivity analyses were performed using the NBS test (5000 permutations, p < 0.05). (E, F) Radar plots of local connectivity in electrode and source spaces across delta, theta, alpha, beta, and gamma bands. The clockwise arrangement is as follows: Delta, theta, alpha, beta, and gamma, repeated for five parcellation regions: Frontal, central (electrode) or limbic (source), temporal, parietal, and occipital. The orange lines indicate the mean FCs values for the CP group, while the blue…

• 

  FIGURE 3. Graph theory analysis of global and local network connectivity. (A) Global graph metrics comparisons between the chronic pain (CP) and healthy controls (HC) groups across different networks. Red diamonds indicate higher values in the CP group (CP>HC), while blue diamonds indicate lower values (CP<HC). (B) Local graph theory measures within the DMN, SAN, and DAN subnetworks. Red nodes represent ROIs where the CP group showed higher values compared to the HC group, indicating increased local connectivity or efficiency, while blue nodes represent ROIs where the CP patients had lower values. gCC, global clustering coefficient; gEff, global efficiency; gSPL, global shortest path length; IstCg, isthmus cingulate; MT, middle temporal; Op, pars opercularis; Or, pars orbitalis; RoMF, rostral middle frontal; S, small‐worldness; SM, supra marginal; Tr, pars triangularis; L, left; R, right.

• 

  FIGURE 4. Dynamic connectivity within the gamma frequency band of the DMN. (A) The two states identified by k‐means clustering and their corresponding cluster centroids. (B) State transition patterns illustrating the time windows in which each subject transitioned between State 1 and State 2. (C) Occurrence frequency and mean dwell time of each state for the entire cohort. (D) Comparative analysis of occurrence frequency and mean dwell time between the chronic pain (CP) and healthy controls (HC) groups, showing no significant differences. (E) State transition probabilities, with no significant group differences observed. (F) Median connectivity matrices for State 1 in the CP and HC groups, along with the results of the NBS permutation test (5000 permutations, p < 0.05).

• 

  FIGURE 5. Machine learning analysis. (A) ROC curve for the SVM model using all features without standardization. (B) ROC curve for the SVM model using all features with standardization. (C) Cross‐validation accuracy as a function of the number of selected features using RFECV. The optimal number of features selected was 51. (D) ROC curve for the SVM model using the 51 RFECV‐selected features with standardization. (E) Box plots comparing classification performance metrics, including accuracy (ACC), sensitivity (Sens), specificity (Spec), and AUC, between models using all standardized features (green) and those using the RFECV‐selected features (red). *p < 0.05; n.s, no significant difference (p > 0.05). (F) Heatmap illustrating the weights of the RFECV‐selected features in the SVM model, with features associated with the DMN (highlighted in blue) playing a significant role in classification. (G) The spatial distribution of the top three features with the highest weights in the SVM model using the RFECV‐selected features. Correlation analysis of pain duration with LOrF.R‐PoC.L (PLV‐Gamma) (H), PoCg.L‐IstCg.L (dFC‐Gamma) (I), and PoCg.L‐CACg.L (dFC‐Gamma) (J). CACg, caudal anterior cingulate;…