Machine learning methods for cross-sectional and longitudinal study of abnormal body fat distribution in HIV-infected individuals
Lipodystrophy is an alteration of body fat distribution associated to HIV and its pharmacological treatment, which can be a risk factor for other health problems, so its identification and prediction may contribute to improve the quality of life of these patients. The goal of this work was to apply...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2021 |
| País: | España |
| Institución: | Universitat Oberta de Catalunya (UOC) |
| Repositorio: | O2, repositorio institucional de la UOC |
| OAI Identifier: | oai:openaccess.uoc.edu:10609/127227 |
| Acceso en línea: | http://hdl.handle.net/10609/127227 |
| Access Level: | acceso abierto |
| Palabra clave: | DEXA/DXA HIV machine learning aprenentatge automàtic Bioinformatics -- TFM Bioinformàtica -- TFM Bioinformática -- TFM |
| Sumario: | Lipodystrophy is an alteration of body fat distribution associated to HIV and its pharmacological treatment, which can be a risk factor for other health problems, so its identification and prediction may contribute to improve the quality of life of these patients. The goal of this work was to apply machine learning (ML) methods to a real dataset containing DXA-derived measures of bone mineral density, lean mass and fat mass from a sample of HIV-infected patients, with repeated measures, aiming to develop tools for identifying and predicting the evolution of lipodystrophy. First, correlational methods and PCA were used to examine data structure, and results showed high correlations among variables, with 6 principal components explaining more than 90% of the original variance contained in 58 variables. ML models showed, cross-sectionally, a very precise classification performance of lipodystrophy cases when variables quantifying fat mass or percentage were included in the models, but poor performance if prediction was based on other body tissues. To incorporate the longitudinal structure, linear mixed models and a combined approach (MEml, Mixed Effects machine learning) were used. Both methods showed good predictive capacity. MEml models allow, in addition, the longitudinal prediction of lipodystrophy. Results highlight the potential of ML methods for classification and prediction of body tissue distribution alterations in the context of HIV infection. |
|---|