Use this link to cite:
http://hdl.handle.net/2183/31919 Aprendizaje automático para el análisis de datos de los pacientes diagnosticados con COVID-19
Loading...
Identifiers
Publication date
Authors
Olañeta Fariña, Daniel
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: El COVID-19 es una amenaza global para los sistemas sanitarios debido a la rápida propagación
del patógeno que lo provoca, el SARS-CoV-2. En tal situación, los clínicos deben
tomar decisiones importantes, en un entorno en el que los recursos médicos pueden ser insuficientes.
En esta tarea, los sistemas de diagnóstico asistido por ordenador pueden ser de gran
utilidad no sólo en la tarea de apoyar la toma de decisiones, sino también para realizar análisis
relevantes que les permitan comprender mejor la enfermedad y las variables que pueden
identificar a los pacientes de alto riesgo. En este contexto, los datos clínicos de los pacientes
recogidos durante la pandemia tienen un gran potencial para ser explotados en el desarrollo
de métodos automatizados para realizar tareas relevantes orientadas al diagnóstico. Con este
propósito, este trabajo propone el uso de varios algoritmos de aprendizaje automático para
estudiar el impacto de determinadas variables clínicas sobre un conjunto de datos recogidos
durante la crisis sanitaria. Para ello, se ha utilizado un conjunto de datos real proporcionado
por el Complejo Hospitalario Universitario A Coruña, diseñado específicamente para los fines
de este estudio. En concreto, los estudios considerados en esta propuesta son útiles para
conocer cuáles son las variables clínicas más relevantes para cada escenario estudiado (riesgo
de muerte u hospitalización de un paciente con COVID-19), así como para medir el rendimiento
de clasificación de los modelos de aprendizaje automático seleccionados. Además, se
proporciona diferentes diagramas representativos mediante estrategias basadas en los árboles
de decisión binarios que ayudan a comprender la toma de decisiones de los modelos de
aprendizaje automático a la hora de estratificar el riesgo de que un paciente sea hospitalizado
o que fallezca debido a la COVID-19.
[Abstract]: COVID-19 is a global threat to healthcare systems due to the rapid spread of the pathogen that causes it, SARS-CoV-2. In such a situation, clinicians must make important decisions, in an environment where medical resources may be insufficient. In this task, computer-aided diagnostic systems can be of great use not only in supporting decision making, but also in performing relevant analyses to better understand the disease and the variables that can identify high-risk patients. In this context, clinical patient data collected during the pandemic have great potential to be exploited in the development of automated methods to perform relevant diagnosis-oriented tasks. To this end, this paper proposes the use of several machine learning algorithms to study the impact of certain clinical variables on a dataset collected during the health crisis. For this purpose, a real dataset provided by the Complejo Hospitalario Universitario A Coruña, specifically designed for the purposes of this study, has been used. Specifically, the studies considered in this proposal are useful to know which are the most relevant clinical variables for each scenario studied (risk of death or hospitalisation of a patient with COVID-19), as well as to measure the classification performance of the selected machine learning models. In addition, different representative diagrams are provided using strategies based on binary decision trees that help to understand the decision-making of the machine learning models when stratifying the risk of a patient being hospitalised or dying due to COVID-19.
[Abstract]: COVID-19 is a global threat to healthcare systems due to the rapid spread of the pathogen that causes it, SARS-CoV-2. In such a situation, clinicians must make important decisions, in an environment where medical resources may be insufficient. In this task, computer-aided diagnostic systems can be of great use not only in supporting decision making, but also in performing relevant analyses to better understand the disease and the variables that can identify high-risk patients. In this context, clinical patient data collected during the pandemic have great potential to be exploited in the development of automated methods to perform relevant diagnosis-oriented tasks. To this end, this paper proposes the use of several machine learning algorithms to study the impact of certain clinical variables on a dataset collected during the health crisis. For this purpose, a real dataset provided by the Complejo Hospitalario Universitario A Coruña, specifically designed for the purposes of this study, has been used. Specifically, the studies considered in this proposal are useful to know which are the most relevant clinical variables for each scenario studied (risk of death or hospitalisation of a patient with COVID-19), as well as to measure the classification performance of the selected machine learning models. In addition, different representative diagrams are provided using strategies based on binary decision trees that help to understand the decision-making of the machine learning models when stratifying the risk of a patient being hospitalised or dying due to COVID-19.
Description
Editor version
Rights
Atribución-NoComercial-SinDerivadas 3.0 España








