Network-based approaches for case-control studies diagnostic/investigations

Authors: Normand, Francis
Advisor: Côté, DanielAllard, Antoine
Abstract: This document offers an overview of different methods employing a network-based approach to identify abnormalities in brain networks associated with a certain diagnostic in typical case-control studies. In these methods, networks associated to patients with a condition are compared against healthy control brain networks. Every patient is represented by a single brain network. The methods discussed operate on some property observed on these brain networks. The choice of a property to measure should be meaningful and related to the (thought to be) means of communication in the brain. A review of a few of these possible measurable properties such as centrality measures is included. Four known methods will be presented, namely, a Permutation Network Framework (PNF), the one-way ANOVA, contrast-subgraph and the Network-based statistic (NBS). In addition, a modification/extension to the NBS called the NBS-SNI (NBS-Simultaneous Node Investigation) will be proposed. Some results obtained using the NBS and the novel NBS-SNI methods on various functional datasets associated to case-control studies (such as ADHD200 (attention deficit hyperactivity disorder), ABIDE I (autism), a schizophrenia study and others) will be presented. Moreover, the differences identified between the two groups of brain networks (condition and control) will also be used to make predictions/diagnosis of individuals using NBS-predict (a prediction extension of NBS). For example, prediction accuracies of 70% and 66% were obtained on the complete ABIDE I and ADHD200 datasets, respectively. Also, the results obtained on the complete ABIDE I dataset were further investigated and compared with other work. The features, which manifest in the condition group as hyperconnected or hypoconnected subnetworks that contributed the most to the prediction performance yielded by NBS-predict were extracted. Some characteristics of these subnetworks seemed to be coherent with previously reported findings on the same dataset/condition. Finally, considerations about using network neuroscience in the context of case-control studies are given, along with future outlooks for the proposed NBS-SNI method.
Document Type: Mémoire de maîtrise
Issue Date: 2022
Open Access Date: 6 June 2022
Permalink: http://hdl.handle.net/20.500.11794/73573
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
Description SizeFormat 
38157.pdf117.11 MBAdobe PDFThumbnail
View/Open
All documents in CorpusUL are protected by Copyright Act of Canada.