Condition-specific neural signatures of reactivation during post-retrieval rest: An EEG study.

1. """
2. Network Analysis of EEG Functional Connectivity
3. ================================================
5. This script computes graph-theoretical measures from EEG functional
6. connectivity matrices and compares the RX and RCTX groups.
8. Main steps:
9. 1. Load behavioral and connectivity data.
10. 2. Normalize connectivity matrices.
11. 3. Threshold networks at fixed densities.
12. 4. Compute graph-theoretical metrics.
13. 5. Perform statistical comparisons.
14. 6. Train and evaluate a Random Forest classifier.
15. 7. Estimate feature importance.
17. Before running:
18. - Update DATA_DIR and BEHAVIOR_FILE.
19. - Install all required dependencies.
20. """
22. # ======================================================================
23. # Imports
24. # ======================================================================
26. import numpy as np
27. import pandas as pd
28. import networkx as nx
29. import matplotlib.pyplot as plt
30. import seaborn as sns
32. from pathlib import Path
34. from scipy.stats import mannwhitneyu, ttest_ind, pearsonr
36. from sklearn.preprocessing import StandardScaler
37. from sklearn.ensemble import RandomForestClassifier
38. from sklearn.model_selection import (
39. LeaveOneOut,
40. cross_validate,
41. train_test_split
42. )
43. from sklearn.inspection import permutation_importance
44. from sklearn.feature_selection import SelectFromModel
45. from sklearn.metrics import make_scorer, f1_score
47. import mne
50. # ======================================================================
51. # User parameters
52. # ======================================================================
54. PROJECT_DIR = Path(".")
55. DATA_DIR = PROJECT_DIR / "adjacency_matrices"
56. BEHAVIOR_FILE = PROJECT_DIR / "behavioral_data.xlsx"
58. CONDITIONS = ["RCTX", "RX"]
59. N_CHANNELS = 64
62. # ======================================================================
63. # Helper functions
64. # ======================================================================
66. def normalize_matrix(baseline_matrix, post_matrix):
67. """
68. Normalize connectivity values using baseline statistics.
70. Parameters
71. ----------
72. baseline_matrix : ndarray
73. Baseline adjacency matrix.
74. post_matrix : ndarray
75. Post-intervention adjacency matrix.
77. Returns
78. -------
79. tuple
80. Baseline matrix and normalized post matrix.
81. """
82. normalized_post = (
83. post_matrix - np.mean(baseline_matrix)
84. ) / np.std(baseline_matrix)
86. return baseline_matrix, normalized_post
89. def load_behavioral_data(excel_file):
90. """
91. Load behavioral performance measures.
93. Returns
94. -------
95. tuple
96. Behavioral matrices for RCTX and RX participants.
97. """
98. df = pd.read_excel(excel_file)
100. # Original processing should be inserted here.
101. return df
104. def compute_network_features(
105. adjacency_matrix,
106. frontal_left,
107. frontal_right,
108. parietal_left,
109. parietal_right,
110. ):
111. """
112. Compute graph-theoretical measures.
114. Returns
115. -------
116. dict
117. Network metrics.
118. """
119. graph = nx.from_numpy_array(adjacency_matrix)
121. metrics = {}
123. metrics["density"] = nx.density(graph)
124. metrics["efficiency"] = nx.global_efficiency(graph)
125. metrics["clustering"] = nx.average_clustering(graph)
126. metrics["assortativity"] = (
127. nx.degree_assortativity_coefficient(graph)
128. )
130. betweenness = np.asarray(
131. list(nx.betweenness_centrality(graph).values())
132. )
134. metrics["betweenness"] = betweenness
136. return metrics
139. def pvalue_by_density(dataframe, grouping_variable):
140. """
141. Compute Mann–Whitney statistics for each density.
142. """
143. metric = dataframe.columns[2]
145. groups = np.unique(dataframe[grouping_variable])
147. pvalues = []
149. for density in np.unique(dataframe["enlaces"]):
151. subset = dataframe[dataframe["enlaces"] == density]
153. group1 = subset[subset[grouping_variable] == groups[0]][metric]
154. group2 = subset[subset[grouping_variable] == groups[1]][metric]
156. result = mannwhitneyu(group1, group2)
158. pvalues.append(result.pvalue)
160. return pvalues
163. # ======================================================================
164. # Main analysis
165. # ======================================================================
167. if __name__ == "__main__":
169. print("Starting network analysis...")
171. # ------------------------------------------------------------------
172. # EEG layout
173. # ------------------------------------------------------------------
175. biosemi_layout = mne.channels.read_layout("biosemi")
177. # Frontal and parietal channel groups
178. frontal_left = np.array([0,1,2,3,4,5,6,7,8,9,10])
179. frontal_right = np.array([33,34,35,38,39,40,41,42,43,44,45])
181. parietal_left = np.array([19,20,21,22,23,24,25,26])
182. parietal_right = np.array([56,57,58,59,60,61,62,63])
184. # ------------------------------------------------------------------
185. # Load behavioral data
186. # ------------------------------------------------------------------
188. behavioral_df = load_behavioral_data(BEHAVIOR_FILE)
190. # ------------------------------------------------------------------
191. # Load adjacency matrices
192. # ------------------------------------------------------------------
194. print("Loading connectivity matrices...")
196. # Insert original loading procedure here.
198. # ------------------------------------------------------------------
199. # Compute network metrics
200. # ------------------------------------------------------------------
202. print("Computing graph metrics...")
204. # Insert original analysis here.
206. # ------------------------------------------------------------------
207. # Classification
208. # ------------------------------------------------------------------
210. forest = RandomForestClassifier(
211. n_estimators=50,
212. random_state=0,
213. max_features=2,
214. max_depth=2
215. )
217. scoring = {
218. "accuracy": "accuracy",
219. "f1_macro": make_scorer(
220. f1_score,
221. average="macro"
222. ),
223. }
225. print("Running classification...")
227. # Insert classification section here.
229. # ------------------------------------------------------------------
230. # Feature importance
231. # ------------------------------------------------------------------
233. print("Computing feature importance...")
235. # Insert permutation importance section here.
237. print("Analysis completed.")

network_analysis_clean.py, no license · at the source