Auxílio diagnóstico para doença de Alzheimer utilizando o modelo de rede neural bioinspirada

• Main Author: Brenner, Gabriel Antonio Stanque
• Publication Date: 2022
• Other Authors: Serenatto, José Hugo, Casimiro, Tatuane Nepomuceno
• Format: Bachelor thesis
• Language: por
• Source: Repositório Institucional da UTFPR (da Universidade Tecnológica Federal do Paraná (RIUT))
• Download full: http://repositorio.utfpr.edu.br/jspui/handle/1/33537
• topic: Alzheimer, Doença de - Diagnóstico; Imagem de ressonância magnética; Redes neurais (Computação); Alzheimer's disease - Diagnosis; Magnetic resonance imaging; Neural networks (Computer science); CNPQ::ENGENHARIAS::ENGENHARIA BIOMEDICA

Due to the increase in life expectancy of the world’s population and as the probability of developing neurodegenerative diseases increases with aging, concern about these diseases is growing. It is estimated that in the next 30 years the number of people with this type of disease will at least double. Among the types of neurodegenerative diseases, Alzheimer’s disease is the most common, which causes memory loss and affects motor capacity. However, currently, there is no cure for the disease, and it is only possible to delay the progression of the disease. Furthermore, as it is a neurological disease, many tests are needed for confirmation the diagnosis, which makes it complex. Therefore, to facilitate this diagnosis, the purpose of this work is to train a neural network that is capable of correctly classifying the level of hippocampal atrophy, one of the biomarkers of Alzheimer’s disease, by comparing data extracted from magnetic resonance images extracted from the Alzheimer’s Disease Neuroimaging Initiative database. Using the Multilayer Perceptron and Koniocortex neural networks, it is intended to evaluate whether the image data corresponds to the data of positive diagnoses for Alzheimer’s disease. The results obtained in the Multilayer Perceptron network were 89.79% correct in the training base and 79.26% correct in the test base with unpublished data. The Koniocortex network presented difficulties in execution and presented similar result patterns for the two diagnoses, which may suggest that the base data could be better defined. Thus, in this work advances are made in research for low-cost diagnostic aid for Alzheimer’s disease.