Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods

Detalhes bibliográficos
Autor(a) principal: Saraiva, Duarte Filipe Oliveira
Data de Publicação: 2021
Tipo de documento: Dissertação
Idioma: eng
Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
Texto Completo: http://hdl.handle.net/10451/51226
Resumo: Tese de mestrado integrado, Engenharia Biomédica e Biofísica (Engenharia Clínica e Instrumentação Médica) Universidade de Lisboa, Faculdade de Ciências, 2021
id RCAP_942d00ce3c0f278ec2dc0e93f2a27584
oai_identifier_str oai:repositorio.ul.pt:10451/51226
network_acronym_str RCAP
network_name_str Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
repository_id_str 7160
spelling Normative model for the diagnosis of neuropsychiatric disorders using deep learning methodsModelo NormativoRedes Cerebrais FuncionaisPerturbações NeuropsiquiátricasConectividade FuncionalAprendizagem ProfundaTeses de mestrado - 2021Departamento de FísicaTese de mestrado integrado, Engenharia Biomédica e Biofísica (Engenharia Clínica e Instrumentação Médica) Universidade de Lisboa, Faculdade de Ciências, 2021The diagnosis of neuropsychiatric disorders (NPDs) is still exclusively dependent on the analysis of the signs and symptoms of the patients since there are no biomarkers useful for clinical practice. Considering that several signs and symptoms are shared among different NPDs, the diagnosis is sometimes incorrect. Therefore, therapeutic approaches do not always succeed, which has an impact on the quality of life of neuropsychiatric patients. Furthermore, NPDs have a global economic and demographic impact. For this reason, technological solutions, such as DL, have been researched for the optimization of diagnosis, in the non-technological field of neuropsychiatry. However, the most promising studies on the diagnosis of NPDs with deep learning (DL) are based on binary classification, which may not be the most adequate approach to deal with the continuous spectrum of NPDs. Here, a DL-based normative model was developed to investigate functional connectivity abnormalities, that may contribute to the development of a novel diagnostic procedure. This method is here used to evaluate how patients deviate from a normal pattern learned by a group of healthy people. To create and evaluate the normative model, resting-state functional magnetic resonance imaging (rs-fMRI) data from three different databases were used. In order to maximise the balance between the amount and the quality of the data, conditions were defined to restrict the variability of the scan parameters. Subsequently, rs-fMRI data were trimmed to the lowest number of time points presented in the sample (150). Then, standard preprocessing steps were performed, including removal of the first 4 volumes of functional data, motion correction, spatial smoothing, and high pass filtering. Single-session independent component analysis (ICA) was run, and the FSL-FIX tool was used to clean noise and artefacts. The functional images were then registered to the T1-weighted brain extracted structural images, and finally to the Montreal Neurosciences Institute 152 standard space. Dual regression was applied using fourteen resting-state functional brain networks (FBN) previously identified in the literature. The Pearson’s correlation coefficient between the extracted blood oxygen level-dependent (BOLD) time series of each FBN was calculated, and a 14x14 network connectivity matrix was generated for each subject. The second part of the project consisted of the creation and optimization of a normative model. The normative model consisted of an autoencoder (AE) with three hidden layers. The AE was trained only in healthy subjects and was tested in both healthy subjects and neuropsychiatric patients, including schizophrenia (SCZ), bipolar disorder (BD), and attention deficit hyperactivity disorder (ADHD) patients. The hypothesis was that the model would “fail” on reconstructing data from neuropsychiatric patients. To evaluate the model performance, graph theory metrics were applied. Besides, the mean squared error was calculated for each feature (correlation between pairs of FBN) to evaluate which regions were worse reconstructed for each group of subjects. The pipeline for NPDs was tested for a SCZ case study, with the addition of a clustering algorithm. The results of this dissertation revealed that the proposed pipeline was able to identify patterns of functional connectivity abnormality that characterize different NPDs. Moreover, the results found for the two SCZ groups of patients were similar, which demonstrated that the normative model here presented was also generalizable. However, the group-level differences did not withstand individual-level analysis, implying that NPDs are highly heterogeneous. These findings support the idea that a precision-based medical approach, focusing on the specific functional network changes of individual patients, may be more beneficial than the traditional group-based diagnostic classification. A personalised diagnosis would allow for personalised therapy, improving the quality of life of neuropsychiatric patients.Ferreira, Hugo AlexandreRepositório da Universidade de LisboaSaraiva, Duarte Filipe Oliveira2022-02-11T10:53:48Z202120212021-01-01T00:00:00Zinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisapplication/pdfhttp://hdl.handle.net/10451/51226enginfo:eu-repo/semantics/openAccessreponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãoinstacron:RCAAP2023-11-08T16:55:49Zoai:repositorio.ul.pt:10451/51226Portal AgregadorONGhttps://www.rcaap.pt/oai/openaireopendoar:71602024-03-19T22:02:33.862417Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãofalse
dc.title.none.fl_str_mv Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods
title Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods
spellingShingle Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods
Saraiva, Duarte Filipe Oliveira
Modelo Normativo
Redes Cerebrais Funcionais
Perturbações Neuropsiquiátricas
Conectividade Funcional
Aprendizagem Profunda
Teses de mestrado - 2021
Departamento de Física
title_short Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods
title_full Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods
title_fullStr Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods
title_full_unstemmed Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods
title_sort Normative model for the diagnosis of neuropsychiatric disorders using deep learning methods
author Saraiva, Duarte Filipe Oliveira
author_facet Saraiva, Duarte Filipe Oliveira
author_role author
dc.contributor.none.fl_str_mv Ferreira, Hugo Alexandre
Repositório da Universidade de Lisboa
dc.contributor.author.fl_str_mv Saraiva, Duarte Filipe Oliveira
dc.subject.por.fl_str_mv Modelo Normativo
Redes Cerebrais Funcionais
Perturbações Neuropsiquiátricas
Conectividade Funcional
Aprendizagem Profunda
Teses de mestrado - 2021
Departamento de Física
topic Modelo Normativo
Redes Cerebrais Funcionais
Perturbações Neuropsiquiátricas
Conectividade Funcional
Aprendizagem Profunda
Teses de mestrado - 2021
Departamento de Física
description Tese de mestrado integrado, Engenharia Biomédica e Biofísica (Engenharia Clínica e Instrumentação Médica) Universidade de Lisboa, Faculdade de Ciências, 2021
publishDate 2021
dc.date.none.fl_str_mv 2021
2021
2021-01-01T00:00:00Z
2022-02-11T10:53:48Z
dc.type.status.fl_str_mv info:eu-repo/semantics/publishedVersion
dc.type.driver.fl_str_mv info:eu-repo/semantics/masterThesis
format masterThesis
status_str publishedVersion
dc.identifier.uri.fl_str_mv http://hdl.handle.net/10451/51226
url http://hdl.handle.net/10451/51226
dc.language.iso.fl_str_mv eng
language eng
dc.rights.driver.fl_str_mv info:eu-repo/semantics/openAccess
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.source.none.fl_str_mv reponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
instacron:RCAAP
instname_str Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
instacron_str RCAAP
institution RCAAP
reponame_str Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
collection Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
repository.name.fl_str_mv Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
repository.mail.fl_str_mv
_version_ 1799134575130050560

Registros relacionados