Network Dynamics in Disorders of Consciousness: A Graph Theoretical Perspective

Brain network analysis plays a crucial role in understanding the complex topological organization of EEG in patients with disorders of consciousness (DOC). This study explores the connectivity patterns in DOC patients by constructing undirected weighted functional networks using amplitude envelope correlation, debiased weighted phase lag index (dwPLI), and coherence. The networks were analyzed using graph theoretical metrics, namely clustering coefficient, path length, modularity, smallworldness, betweenness centrality, and node degree to assess resting-state brain activity. The relationship between the network features and the Glasgow outcome scale–extended scores at three months was investigated using canonical correlation analysis (CCA). Results revealed that unresponsive wakefulness syndrome patients exhibited higher clustering coefficients than those in a minimally conscious state across all three network models. CCA identified a strong correlation of 0.91 in the dwPLI model, highlighting its prognostic significance. The ability of graph metrics to differentiate between DOC groups was tested using the XGBoost classifier, achieving an accuracy of 73%. These findings demonstrate that brain network analysis, grounded in functional connectivity measures and graph theory, provides valuable insights into the neural dynamics underlying DOC and offers a promising approach for predicting clinical outcomes.