• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Doctoral Thesis
DOI
https://doi.org/10.11606/T.18.2009.tde-21092009-144642
Document
Author
Full name
André Augusto Spadotto
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 2009
Supervisor
Committee
Pereira, José Carlos (President)
Guido, Rodrigo Capobianco
Maciel, Carlos Dias
Oliveira, Suely Pereira de
Schelp, Arthur Oscar
Title in Portuguese
Análise quantitativa do sinal da deglutição
Keywords in Portuguese
Floresta de caminhos ótimos
OPF
Sinal da deglutição
Transformada Wavelet
Abstract in Portuguese
Neste trabalho, buscou-se compreender a morfologia e os componentes do sinal da deglutição. Na busca desse entendimento diversas técnicas foram empregadas. No intuito de fazer marcações fidedignas em trechos específicos do sinal, o qual foi analisado simultaneamente com a imagem da videofluoroscopia da deglutição, considerado o melhor método atual na avaliação da dinâmica da deglutição. Os parâmetros numéricos utilizados para análise também foram abrangentes e com base em técnicas atuais de processamento de sinais, como emprego de transformada Wavelet. Quanto à classificação dos sinais, foram utilizados classificadores modernos como floresta de caminhos ótimos, máquinas de vetores de suporte, redes neurais artificiais e classificador Bayesiano, dando maior ênfase ao primeiro, por possuir um custo computacional bem menor quando comparado aos outros 3, e consequentemente convergindo mais rapidamente ao resultado. Foram avaliados 84 sinais, divididos em 2 grupos separados pela consistência do bolo alimentar oferecido (líquido e pastoso). Na distinção e/ou caracterização desses tipos foi definido um subconjunto com 4 variáveis que proporcionou uma boa acurácia na separação das classes representantes de cada tipo de bolo alimentar.
Title in English
Quantitative analysis of the swallowing signal
Keywords in English
OPF
Optimum path forest
Swallowing signal
Wavelet transform
Abstract in English
This work proposes to understand the morphology and the components of the swallowing signal. In pursuit of this understanding, a variety of techniques were employed. In order to make reliable markings on specific portions of the signal, the signal was examined simultaneously with videofluoroscopic swallowing, which is considered the best method in the evaluation of swallowing dynamics. The parameters used for numerical analysis were based on current signal processing techniques, such as: Wavelet transform, Optimum path forest, Support vector machines, Artificial neural networks and Bayesian classifier, emphasizing the first technique, due to a much lower computational cost when compared to the previous, and, consequently, the results converged much faster. Eighty four signals, divided into 2 groups separated by the consistency of food bolus offered (liquid and thickened), were evaluated. For distinction and/or characterization of such types, a subset with 4 variables was defined, providing a good accuracy in the separation of these classes representing each type of consistency of the food bolus.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Spadotto.pdf (2.85 Mbytes)
Publishing Date
2009-09-23
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.