Classificação de Janelas de Eletroencefalograma com e sem Crises Epilépticas.

• Ano de defesa: 2021
• Autor(a) principal: NASCIMENTO, Davi Costa
• Orientador(a): BARROS FILHO, Allan Kardec Duailibe
• Banca de defesa: BARROS FILHO, Allan Kardec Duailibe , QUEIROZ, Jonathan Araújo , SERRA, Ginalber Luiz de Oliveira , TOMAZ, Carlos Alberto Bezerra
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Maranhão
• Programa de Pós-Graduação: PROGRAMA DE PÓS-GRADUAÇÃO EM ENGENHARIA DE ELETRICIDADE/CCET
• Departamento: DEPARTAMENTO DE ENGENHARIA DA ELETRICIDADE/CCET
• País: Brasil
• Palavras-chave em Português: Epilepsia; Eletroencefalograma; Aprendizado de máquina; Estatística
• Palavras-chave em Inglês: Epilepsy; Electroencephalogram; Machine Learning; Statistics
• Área do conhecimento CNPq: Engenharia de Software
• Link de acesso: https://tedebc.ufma.br/jspui/handle/tede/3246

Epilepsy is a chronic neurological disorder, which enhances the occurrence of seizures in its victims. This clinical picture causes daily disorders of patients, that in some cases, the occurrence of seizures can promote injuries, trauma or even sudden death. A previous diagnosis, as well as an adequate treatment, allows the patients a more comfortable life and even the disappearance of seizures. Therefore, there is a need to develop technologies and methodologies that can streamline and simplify the diagnosis of epilepsy, as well as new options for the treatment and monitoring of patients. This work aims to develop a generalized model for the classi cation of electroencephalogram segments with the presence or absence of epileptic seizures. The methodology adopted in the extraction of features consists in the calculation of the statistical moments of second, third and fourth order. These features were extracted from 1-second windows of les from the CHB-MIT database (Children's Hospital Boston). The classi cation of the feature vectors obtained consisted of two stages, the rst from the statistical moments, the second with the rotation of the same vectors, using the principal component analysis (PCA). The proposed classi cation model obtained for the feature vectors and for the components coming from the PCA, accuracy of 86.4% and 94.6%, respectively, both using the k-NN algorithm. It was believed that with appropriate adaptations and improvements, the model can be embedded in a device, for classi cation of windows in real time.