Classificação de lateralidade hemisférica de linguagem de linguagem em imagens de ressonância magnética funcional utilizando aprendizado de máquina

• Autor(a) principal: Morais, Mariana Pastro
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da PUC_RS
• Texto Completo: http://tede2.pucrs.br/tede2/handle/tede/9334
• topic: Ressonância Magnética Funcional; Aprendizado de Máquina; Neurolinguagem; Lateralidade Hemisférica Cerebral; Functional Magnetic; Resonance; Machine Learning; Neurolanguage; Cerebral Hemispheric Laterality; ENGENHARIAS

Although the cerebral language area has its anatomical location defined by Wernicke’s area and Broca’s area, some patients affected with brain tumors have their activation topographies altered due to a phenomenon called language translocation. Functional Magnetic Resonance Imaging (fMRI) is a diagnostic imaging method that highlights language areas, contributing to the correct surgical planning of tumor removal. Despite the detection of brain activation areas by visual inspection being the most used at the clinic, some studies have been using Machine Learning techniques, such as Support Vector Machine, Logistic Regression, Decision Trees, and others. This work aims to evaluate the application of a technique for classification language area laterality to be using Machine Learning. The activation coefficients of the areas of interest in fMRI images were extracted using the AFNI software. A Multilayer Perceptron artificial neural network was set up in order to determine the probability of the language area are allocated in a certain cerebral hemisphere. The performance of the neural network was assessed by means of statistical comparison with the Laterality Index and the visual analisys. The results showed that MLP RNA has the potential to be an auxiliary method for classifying cerebral hemispheric laterality in the language area.