Uma metodologia para extrair e avaliar padrões em imagens de ressonância magnética funcional na dimensão temporal

• Ano de defesa: 2013
• Autor(a) principal: Prata, Marlon Santos
• Orientador(a): Montesco, Carlos Alberto Estombelo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Sergipe
• Programa de Pós-Graduação: Pós-Graduação em Ciência da Computação
• País: BR
• Palavras-chave em Português: Computação; Ressonância magnética; Processamento de imagens; Eletromagnetismo; Dimensão temporal; Análise de componentes independentes
• Palavras-chave em Inglês: Image processing; Signal analysis; Magnetically resonance; Temporal dimension; Independent component analysis
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://ri.ufs.br/handle/riufs/3346

This work presents a methodology to extract and evaluate signals in functional magnetic resonance imaging (fMRI) in the temporal dimension. It is assumed that it is possible to separate the signals from brain activations of a given protocol of other signals such as breathing, heartbeat, involuntary eye movements, and others. This paper proposes a methodological way, get through simulated and controlled measure the efficiency of the Model of Independent Component Analysis (ACI) for fMRI images to separate these signals experiments. To validate the experiments it was necessary to generate simulated data. The data generated were formed for three (3) signs that did not have a Gaussian distribution, in an array of temporal dimension 80 x 80 x 64. Within this set of signs have been added three (3) signal blocks of size 5 x 5 x 64 that simulated activations of FRH protocols. After this process, an array of mixture was added so that the signals could not be identified. Only after the data gave mixed-if the process of completing the pre-processing with the bleaching and centering of the variables to the model following ACI was performed to separate the signals that were mixed, thereby finding the estimated component signals. Different amounts of Gaussian signals were added until no more would be possible to extract the component which would correspond to a signal FRH testing the theory that ACI has efficiency in extracting components of non-Gaussian data. The model was run in real signals, where a volunteer performing a hearing protocol, the results data of each slice extracted resonance are evident throughout this work, in some slices was possible to extract the expected component with a degree of correlation between the acceptable component and signal protocol FRH, as in other slices could not perform the extraction of these signals. As a final result a statistical correlation map for each slice resulting from the estimated component and the raw data was generated, the signals were evaluated on the assumption of acceptance only to greater than 0.72 correlation with statistical significance at 95% confidence.