EEG network analysis to identify cortical pathology in Lennox-Gastaut syndrome
Zhi Ji Wang1, Hye Eun Kwon2, Hui Jin Shin3
1Division of Pediatric Neurology, Department of Pediatrics, Severance Children's Hospital, Yonsei University College of Medicine, Seoul, Republic of Korea; Radio Frequency Bio Research Center, Kwangwoon University, Seoul, Republic of Korea.
Objective:
This study aimed to identify preoperative electrophysiologic biomarkers to localize the epileptogenic zone (EZ) and predict surgical outcomes in patients with Lennox-Gastaut syndrome (LGS) using a graph-theoretical network analysis.
Methods:
We retrospectively analyzed preoperative long-term EEG data from 71 LGS patients in three subgroups: Group A (n = 23), with favorable outcomes after resective surgery (RS); Group B (n = 22), with unfavorable outcomes after RS; and Group C (n = 26), who underwent palliative surgery instead of RS. We analyzed 500-ms EEG windows centered on the point of peak phase coupling preceding interictal generalized sharp-and-slow waves. We computed two sets of biomarkers: (1) the clustering coefficient (CCE) to assess overall network connectivity, and (2) a novel 'Network Dominance Score' (the product of Betweenness Centrality, Node Out-Degree, and Node Out-Strength) to identify the putative ictogenic hub.
Results:
The CCE effectively differentiate the favorable-outcome group (Group A), the unfavorable-outcome group (Group B), and the palliative surgery group (Group C) (Area Under the Curve = 0.915). The Network Dominance Score achieved 100% accuracy in correctly lateralizing the lesion for all patients in the Group A. Conversely, for the Group B, this analysis identified either bilateral or incorrectly localized hubs, demonstrating a clear correlation between our analytical findings and surgical outcomes.
Conclusion:
This graph-theoretical approach offers a promising non-invasive tool to augment conventional presurgical evaluations, improve patient selection for RS, and predict surgical outcomes in individuals with LGS.
