• 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.45.2021.tde-19042021-173516
Document
Author
Full name
Taiane Coelho Ramos
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2021
Supervisor
Committee
Fujita, André (President)
Baccala, Luíz Antonio
Oliveira, Leticia de
Sato, João Ricardo
Takahashi, Daniel Yasumasa
Title in Portuguese
Técnicas de clusterização e estratificação de indivíduos para estudo de redes funcionais cerebrais
Keywords in Portuguese
Agrupamento de grafos
Associações cérebro-comportamentais
Densidade espectral
Grafos aleatórios
SPLS
Abstract in Portuguese
Em diversas aplicações de neurociência gostaríamos de entender semelhanças e diferenças entre indiví- duos analisando as conectividades do cérebro. Para isso, propomos duas abordagens: (i) agrupar indivíduos semelhantes ou (ii) posicioná-los em um espectro estratificado de um diagnóstico. Para (i), nós modelamos as redes cerebrais como um grafo e apresentamos dois métodos de clusterização baseados em análise espec- tral: um k-means para grafos do mesmo tamanho e uma abordagem baseada em modelo, chamada gCEM, para grafos de tamanhos diferentes. Para avaliar os pontos fortes e fracos dos métodos, projetamos alguns cenários usando modelos de grafos aleatórios. Também aplicamos a dados reais de compostos químicos e de redes cerebrais funcionais. Para (ii) nós utilizamos uma análise multivariada para encontrar associações cérebro-comportamentais e criar uma estratificação dos indivíduos variando de típicos a atípicos. Aplica- mos um framework de múltiplos holdouts com o método Sparse Partial Least Squares para encontrar essas associações. Utilizamos uma amostra de 349 crianças e adolescentes do conjunto de dados ABIDE II e en- contramos um modo de associação significativo entre dados fMRI e características comportamentais. As variáveis comportamentais mais importantes nesta associação estão relacionadas à capacidade de resposta social e as conexões cerebrais mais importantes estão dentro da rede de modo padrão e entre as redes somatomotora e visual.
Title in English
Clustering techniques and stratification of individuals to study functional brain networks
Keywords in English
Brain-behaviour associations
Graph clustering
Random graph
Spectral density
SPLS
Abstract in English
In several neuroscience applications, we would like to understand the similarities and the differences between individuals in the same group by analysing their brain connectivity. For this, we propose two approaches: (i) to group similar individuals or (ii) to place them in a stratified spectrum of a diagnosis. For (i), we model the brain networks as a graph and present two clustering methods based on spectral analysis: a k-means for graphs of the same size and a model-based approach, called gCEM, for graphs of different sizes. To assess the strengths and weaknesses of the methods, we designed some scenarios using random graph models. We also apply it to real data on chemical compounds and functional brain networks. For (ii), we use a multivariate analysis to find brain-behavioural associations and create stratification of individuals ranging from typical to atypical. We applied a framework of multiple holdouts with Sparse Partial Least Squares method to find these associations. We used a sample of 349 children and adolescents from the ABIDE II data set and found a significant association between fMRI data and behavioural characteristics. The most important behavioural variables in this association are related to social responsiveness. The most important brain connections are within the default-mode network and between the somatomotor and visual networks.
 
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.
Tese_revisada.pdf (8.23 Mbytes)
Publishing Date
2021-04-20
 
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.