Dynamic evaluation of induced epilepsy in rats: a bayesian network perspective

• Ano de defesa: 2022
• Autor(a) principal: Tsukahara, Victor Hugo Batista
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Bayesian networks; Complex system; Electroencephalography; Eletroencefalografia; Epilepsia; Epilepsy; Redes bayesianas; Sistema complexo; Stochastic volatility; Volatilidade estocástica
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/18/18153/tde-19122022-123219/

Epilepsy is one of the most common neurological disorders worldwide. Recent findings on it suggest that the brain is a complex system based on a network of neurons whose interactions result in an epileptic seizure, which is currently considered an emergent property. Based on such a modern view, network physiology has emerged to address how brain areas coordinate, synchronize and integrate their dynamics during sound health and afflicted conditions.The objective of this thesis is to present an application of (Dynamic) Bayesian Networks (DBN) to model Local Field Potentials (LFP) based on recordings of rats induced to epileptic seizures and arcs evaluated using an analytical threshold approach. A dynamic network model was constructed from data using the Bayesian Network method, either by considering the delay of communication among brain areas recorded in this study or not. To such an end, the Multivariate Stochastic Volatility method was employed to identify the lag among Local Field Potentials and K2 Score so as to compare the models. Results also showed that the DBN analysis has captured the dynamic nature of brain connectivity across ictogenesis, and that there is a significant correlation to neurobiology derived from pioneering studies employing techniques of pharmacological manipulation, lesion, and modern optogenetics. The arcs evaluation under the proposed approach was consistent with previous literature. Moreover, it provided exciting novel insights, such as a discontinuity between forelimb clonus and generalized tonic-clonic seizure (GTCS) dynamics. Dynamic Bayesian Network depicted the evolution of rats\' brains from resting-state until the generalized tonic-clonic seizure. Multivariate Stochastic Volatility captured the lag among brain areas, and better results were yielded after its application on the DBN model.