Loading...
Publication
Aplicação de machine learning na identificação precoce da doença de Crohn e estratificação da sua gravidade com base em anomalias metabólicas.
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.61 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Atualmente, com a quantidade crescente de dados disponíveis nos mais variados setores, tem emergido a necessidade de procurar abordagens tecnológicas capazes de analisar grandes volumes de dados. Assim, a Machine Learning (ML), um ramo da Inteligência Artificial (AI), tem-se tornado cada vez mais popular, permitindo aplicar algoritmos a conjuntos de dados com o intuito de prever resultados ou identificar relações entre as informações do dataset. As técnicas de ML têm vindo a ganhar popularidade do ramo da medicina, permitindo prever e monitorizar a evolução de determinadas doenças. Este estudo pretende aplicar estratégias de ML a dados analíticos de casos com doença de Crohn, bem como de casos com doenças hereditárias do metabolismo, explorando o potencial dos perfis metabólicos como biomarcadores e a sua utilidade na estratificação
de doentes. O estudo progrediu em várias etapas: pré-preparação dos dados; aplicação de modelos de classificação, avaliando a capacidade preditiva de diferentes algoritmos na distinção de diagnósticos; implementação de clustering no conjunto completo de amostras, visando a identificação de diferentes perfis metabólicos; e aplicação de clustering restrito a doentes com Crohn, analisando a heterogeneidade interna nesta população e sugerindo a existência de grupos mais predispostos a complicações, como o desenvolvimento de fístulas perianais. Os resultado alcançados reforçam a importância da combinação da análise metabolómica
com técnicas de ML, tanto na identificação de padrões metabólicos, como no potencial da utilização de biomarcadores no apoio do diagnóstico e acompanhamento da doença. Este trabalho contribui, assim, para evidenciar o papel da metabolómica e da Inteligência Artifical no avanço da medicina de precisão aplicada à Doença de Crohn.
Currently, with the growth of available data in a wide variety of sectors, there is an increasing demand for technological approaches capable of analysing substantial volumes of data. ML, a branch of AI, has become increasingly popular, making it possible to apply algorithms to data sets in order to predict results or identify relationships between information in the dataset. ML techniques have seen a surge in popularity within the medical field, with applications ranging from disease progression monitoring to predictive modelling. The objective of this study is to implement ML methods on clinical data from patients diagnosed with Crohn’s disease as well as patients with hereditary metabolic disorders, with the aim of investigating the potential of metabolic profiles as biomarkers and their efficacy in patient startification. The study progressed in several stages: firstly, data was prepared for analysis; secondly, classification models were applied, and the predictive capacity of different algorithms was evaluated in different diagnoses; thirdly, clustering was implemented on the complete set of samples to identify different metabolic profiles; and finally, clustering was applied exclusively to Crohn’s disease patients, analysing the internal heterogeneity in this population and studying the possibility of the existence of subgroups with a higher predisposition to complications, such as the development of perianal fistulas. The findings underscore the significance of integrating metabolomic analysis with machine learning methodologies, both in discerning metabolic patterns and in the prospective utilisation of biomarkers to facilitate diagnosis and disease monitoring. This work thus contributes to highlighting the role of metabolomics and artificial intelligence in advancing precision medicine applied to Crohn’s disease.
Currently, with the growth of available data in a wide variety of sectors, there is an increasing demand for technological approaches capable of analysing substantial volumes of data. ML, a branch of AI, has become increasingly popular, making it possible to apply algorithms to data sets in order to predict results or identify relationships between information in the dataset. ML techniques have seen a surge in popularity within the medical field, with applications ranging from disease progression monitoring to predictive modelling. The objective of this study is to implement ML methods on clinical data from patients diagnosed with Crohn’s disease as well as patients with hereditary metabolic disorders, with the aim of investigating the potential of metabolic profiles as biomarkers and their efficacy in patient startification. The study progressed in several stages: firstly, data was prepared for analysis; secondly, classification models were applied, and the predictive capacity of different algorithms was evaluated in different diagnoses; thirdly, clustering was implemented on the complete set of samples to identify different metabolic profiles; and finally, clustering was applied exclusively to Crohn’s disease patients, analysing the internal heterogeneity in this population and studying the possibility of the existence of subgroups with a higher predisposition to complications, such as the development of perianal fistulas. The findings underscore the significance of integrating metabolomic analysis with machine learning methodologies, both in discerning metabolic patterns and in the prospective utilisation of biomarkers to facilitate diagnosis and disease monitoring. This work thus contributes to highlighting the role of metabolomics and artificial intelligence in advancing precision medicine applied to Crohn’s disease.
Description
Keywords
Matabolomics Metabolic Profiles Artifical Intelligence Machine Learning Classification Clustering Biomarkers Perianal Fistulas Crohn’s Disease Doença de Crohn Fístulas perianais Perfis metabólicos Inteligência artificial Machine learning