Please use this identifier to cite or link to this item:
http://hdl.handle.net/10609/127526
Title: | Exploratory analysis of a biological database (DEXA) and application of Machine Learning models to detect osteoporosis in HIV-positive patients |
Author: | Regué Alsina, Adrià |
Tutor: | Perez-Alvarez, Nuria |
Others: | Maceira, Marc |
Abstract: | Osteoporosis incidence is notoriously larger in the HIV-positive population. For this reason, DEXA analysis (bone densitometry tests) are conducted as a control technique. This work focuses on studying a real DEXA database, retrieved from HIV+ patients doing medical checkups in the Lluita contra la SIDA Foundation, in Badalona. Medical databases often suffer from strong correlations between variables. For this reason, the first chapter of the study has been destinated to purify the database and discover said relationships, via correlation plots and more innovative techniques such as graphical models (GGMs and MGMs). Also, a dimensionality reduction analysis has been executed using principal components. This first part of the study corroborated the relevance of the gender variable. All the subsequent analysis has been conducted separately for men and women. Also, graphical models suggested that vertebral variables have a rather weak importance in determining the minimum T-score (and thus, predicting osteoporosis). The second part of the study has focused on generating a predictive model with the ability to diagnose osteoporosis without using its classical indicator variables. After modelling with various Machine Learning algorithms (Random Forests, SVMs, k-NNs), a classificatory model has been generated, reporting a sensitivity and specificity of ~80%. |
Keywords: | HIV DEXA dimensionality reduction graphical models machine learning |
Document type: | info:eu-repo/semantics/masterThesis |
Issue Date: | Jan-2021 |
Publication license: | http://creativecommons.org/licenses/by-nc-nd/3.0/es/ |
Appears in Collections: | Trabajos finales de carrera, trabajos de investigación, etc. |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
DEXAmarkdown.pdf | 372,83 kB | Adobe PDF | View/Open | |
aregueaTFM0121memory.pdf | Memory of TFM | 3,9 MB | Adobe PDF | View/Open |
Share:
This item is licensed under a Creative Commons License