Supervised methods to classify body composition in HIV-infected patients
Since it was first diagnosed in the early 1980s, the research on AIDS has exponentially improved along the years. Nowadays, in fields like biostatistics where biology meets advanced statistics and mathematics, the study lines have turned and another approach can be made to handle scientific problems...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2019 |
| País: | España |
| Institución: | Universitat Oberta de Catalunya (UOC) |
| Repositorio: | O2, repositorio institucional de la UOC |
| OAI Identifier: | oai:openaccess.uoc.edu:10609/98146 |
| Acceso en línea: | http://hdl.handle.net/10609/98146 |
| Access Level: | acceso abierto |
| Palabra clave: | machine learning body composition HIV aprendizaje automático composición corporal VIH aprenentatge automàtic composició corporal Bioinformatics -- TFM Bioinformàtica -- TFM Bioinformática -- TFM |
| Sumario: | Since it was first diagnosed in the early 1980s, the research on AIDS has exponentially improved along the years. Nowadays, in fields like biostatistics where biology meets advanced statistics and mathematics, the study lines have turned and another approach can be made to handle scientific problems. Due to the successful improvements that research has provided to the quality of life of HIV-positive individuals, their life expectancy is approximately normal, but they develop ageing related diseases: their bones turn to be more fragile, their muscles get weaker, and their fat mass may be abnormally distributed. The aim of this project is to find machine learning supervised classification methods to detect ageing diseases in HIV-infected individuals, and discuss if these methods provide valuable information regarding the variables of the dataset and how they relate to each other. Thus, a significative ('No information rate' test p-value < 0.05) Neural Network model is computed using R and in particular the package mlr, to successfully classify a lean mass disease associated with the ageing process, using bone and fat data samples as explicative variables. The topology and weight distribution of this model's network provides information about the most relevant variables, which may be of clinical interest. Given the elevated accuracy and positive performance parameters of said classification method, it is safe to say that the assumption that a lean mass ageing disease could be predicted by the bone and fat tissue variables is validated, and consequently the goals of this project are achieved. |
|---|