Altered neural avalanche spreading in people with drug-resistant epilepsy

• Published in: NeuroImage (Orlando, Fla.) Vol. 311; p. 121188
• Main Authors: Sancetta, B.M., Matarrese, M.A.G., Ricci, L., Lanzone, J., Lippa, G., Nesta, M., Zappasodi, F., Brunetti, M., Di Lazzaro, V., Tombini, M., Assenza, G.
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.05.2025; Elsevier Limited; Elsevier
• Subjects: Adult; Avalanche analysis; Avalanche transition matrices; Avalanches; Biomarkers; Convulsions & seizures; Drug dosages; Drug resistance; Drug Resistant Epilepsy - physiopathology; Drug-resistant epilepsy; EEG; Electroencephalography; Electroencephalography - methods; Epilepsy; Female; Graph theory; Humans; Hypotheses; Male; Middle Aged; Nerve Net - physiopathology; Neural networks; Phase transitions; Seizures; Spreading; Young Adult; Graph theory; Drug-resistant epilepsy; Avalanche transition matrices; Avalanche analysis
• ISSN: 1053-8119; 1095-9572; 1095-9572
• DOI: 10.1016/j.neuroimage.2025.121188

•Avalanche spreading proprieties in drug-resistant epilepsy exhibit a more regular network topology (higher local cliquishness and low global integration).•Avalanche transition matrices accurately differentiate people with epilepsy according to clinical outcome (drug-resistant or not).•The characteristics of resting-state EEG activity of drug-resistant epilepsy do not deviate from those systems working at criticality. To characterize a peculiar “EEG endophenotype” of drug-resistant epilepsy (DRE) through the graph theory characterization of avalanche spatiotemporal spreading properties. We performed avalanche analysis and computed avalanche transition matrices (ATMs) on 19-channel scalp EEG of 120 people with epilepsy (60 DRE and 60 non-DRE) who assumed two anti-seizure medications, comparing such results with a group of 40 healthy subjects (HS). Network topologies of ATMs were characterized through graph theory metrics. We performed an analysis of variance to compare aperiodic metrics between HS, DRE and non-DRE. Logistic regression was performed to test and compare the ability of graph theory metrics on ATM and clinical features to correctly discriminate the PwE group according to the clinical outcome (DRE or non-DRE). DRE exhibited a peculiar altered avalanche spreading as proved by the higher mean betweenness centrality, the longer characteristic path length and the lower small-world index (more regular and less plastic network topology) of ATMs than non-DRE and HS (p-values from <0.001 to 0.05). Graph metrics on ATMs significantly improved the yield of detecting DRE and contributed the most to the model accuracy (0.83) than clinical features. Resting-state EEG activity of HS and PwE did not deviate from the characteristics of a system operating at criticality. ATMs detect alterations of resting-state networks peculiar to the DRE condition. These findings could open new scenarios for the future identification of promising biomarkers of DRE through scalp EEG.