Publicação
Classification of dementias based on brain radiomics features
| Resumo: | Neurodegenerative diseases impair the functioning of the brain and are characterized by alterations in the morphology of specific brain regions. Some of the main disorders include Alzheimer's, Parkinson's, and Huntington's diseases, and the number of cases increases exponentially since ageing is one of the main risk factors. Trying to identify the areas in which this type of disease appears is something that can have a very positive impact in this area of Medicine and can guarantee a more appropriate treatment or allow the improvement of the quality of life of patients. With the current technological advances, computer tools are capable of performing a structural or functional analysis of neuroimaging data from Magnetic Resonance Images(MRI). Therefore, Medical Informatics uses these techniques to create and manage medical neuroimaging data to improve the diagnosis and management of these patients. MRI is the image type used in the analysis of the brain area and points to a promising and reliable diagnostic tool since it allows high-quality images in various planes or strategies and MRI methods are fundamental diagnostic tools in clinical practice, allowing the diagnosis of pathologic processes such as stroke or brain tumours. However, structural MRI has limitations for the diagnosis of neurodegenerative disorders since it mainly identifies atrophy of brain regions. Currently, there is increased interest in informatics applications capable of monitoring and quantifying human brain imaging alterations, with potential for neurodegenerative disorders diagnosis and monitoring. One of these applications is Radiomics, which corresponds to a methodolog ythat allows the extraction of features from images of a given region of the brain. Specific quantitative metrics from MRI are acquired by this tool, and they correspond to a set of features, including texture, shape, among others. To standardize Radiomics application, specific libraries have been proposed to be used by the bioinformatics and biomedical communities, such as PyRadiomics, which corresponds to an open source Python package for extracting Radiomics of MRIs. Therefore, this dissertation was developed based on magnetic resonance images and the study of Deep Learning (DL) techniques to assist researchers and neuroradiologists in the diagnosis and prediction of neurodegenerative disease development. Two different main tasks were made: first, a segmentation, using FreeSurfer, of different regions of the brain and then, a model was build from radiomic features extracted from each part of the brain and interpreted for knowledge extraction. |
|---|---|
| Autores principais: | Carvalho, Sofia Manuela Gomes de |
| Assunto: | Machine Learning Dementia classification Magnetic resonance imaging Radiomics Brain morphological features Neurodegenerative diseases ADNI Classificação de demências Imagens ressonância magnética Radiomics Features morfológicas cerebrais Doenças neurodegenerativas |
| Ano: | 2021 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Neurodegenerative diseases impair the functioning of the brain and are characterized by alterations in the morphology of specific brain regions. Some of the main disorders include Alzheimer's, Parkinson's, and Huntington's diseases, and the number of cases increases exponentially since ageing is one of the main risk factors. Trying to identify the areas in which this type of disease appears is something that can have a very positive impact in this area of Medicine and can guarantee a more appropriate treatment or allow the improvement of the quality of life of patients. With the current technological advances, computer tools are capable of performing a structural or functional analysis of neuroimaging data from Magnetic Resonance Images(MRI). Therefore, Medical Informatics uses these techniques to create and manage medical neuroimaging data to improve the diagnosis and management of these patients. MRI is the image type used in the analysis of the brain area and points to a promising and reliable diagnostic tool since it allows high-quality images in various planes or strategies and MRI methods are fundamental diagnostic tools in clinical practice, allowing the diagnosis of pathologic processes such as stroke or brain tumours. However, structural MRI has limitations for the diagnosis of neurodegenerative disorders since it mainly identifies atrophy of brain regions. Currently, there is increased interest in informatics applications capable of monitoring and quantifying human brain imaging alterations, with potential for neurodegenerative disorders diagnosis and monitoring. One of these applications is Radiomics, which corresponds to a methodolog ythat allows the extraction of features from images of a given region of the brain. Specific quantitative metrics from MRI are acquired by this tool, and they correspond to a set of features, including texture, shape, among others. To standardize Radiomics application, specific libraries have been proposed to be used by the bioinformatics and biomedical communities, such as PyRadiomics, which corresponds to an open source Python package for extracting Radiomics of MRIs. Therefore, this dissertation was developed based on magnetic resonance images and the study of Deep Learning (DL) techniques to assist researchers and neuroradiologists in the diagnosis and prediction of neurodegenerative disease development. Two different main tasks were made: first, a segmentation, using FreeSurfer, of different regions of the brain and then, a model was build from radiomic features extracted from each part of the brain and interpreted for knowledge extraction. |
|---|