Machine learning applied to a Cardiac Surgery Recovery Unit and to a Coronary Care Unit for mortality prediction

Nistal Nuño, Beatriz. (2021). Machine learning applied to a Cardiac Surgery Recovery Unit and to a Coronary Care Unit for mortality prediction Master Thesis, Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática. Departamento de Inteligencia Artificial.

Ficheros (Some files may be inaccessible until you login with your e-spacio credentials)
Nombre Descripción Tipo MIME Size
Nistal_Nuno_Beatriz_TFM.pdf Nistal_Nuno_Beatriz_TFM.pdf application/pdf 1.75MB

Título Machine learning applied to a Cardiac Surgery Recovery Unit and to a Coronary Care Unit for mortality prediction
Autor(es) Nistal Nuño, Beatriz
Abstract Predicting the mortality risk for patients with cardiac disease in Intensive Care Units is essential for effective care planning, where impending deterioration can occur with severe health consequences. Most established severity of illness systems used for prediction of Intensive Care Unit mortality were developed targeted at the general Intensive Care Unit population, based on logistic regression To date, no dynamic predictive tool has been developed targeted at patients in the Cardiac Surgery Recovery Unit and Coronary Care Unit using machine learning. In this research, adult patients at the Cardiac Surgery Recovery Unit and Coronary Care Unit fro m the MIMIC III critical care database were studied. Intensive Care Unit data was extracted during a 5 hour window in addition to a few demographic features to produce 12 hour advan ce mortality predictions. The machine learning models developed were the Tr ee Ensemble of Decision Trees, Random Forest of Decision Trees, XGBoost Tree Ensemble, Naive Bayes, and Bayesian network. The models were compared to six established systems by assessing the discrimination, calibration and accuracy statistics. The main adv antages of these models are that they overcome most limitations of logistic regression, utilized to build the established systems , in addition to being dynamic as opposed to the static traditional systems. The AUROC values for the primary outcome were superior for all t he machine learning models, the accuracy statistics less sensitive to unbalanced cohorts were substantially higher for all the machine learning models and, finally, the Brier score was better for all the machine learning models except the Naïve Bayes over the conventional systems. In conclusion, the discriminatory power of XGBoost and Tree Ensemble were excellent, substantially outperforming the conventional systems. Additionally, the machine learning models showed better performance for the vast majority of accuracy measures. Consequently, the models developed in this work offer promising results that could benefit Cardiac Surgery and Coronary Care Units.
Notas adicionales Trabajo de Fin de Máster. Máster Universitario en I.A. Avanzada: Fundamentos, Métodos y Aplicaciones. Especialidad de Sistemas Inteligentes de Diagnóstico, Planificación y Control. UNED
Materia(s) Ingeniería Informática
Palabra clave Cardiac Surgery
Coronary Care Unit
machine learning
mortality
discrimination
calibration
Editor(es) Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática. Departamento de Inteligencia Artificial.
Director/Tutor Fernández Galán, Severino
García Tahoces, Pablo
Fecha 2021-06-01
Formato application/pdf
Identificador bibliuned:master-ETSInformatica-IAA-Bnistal
http://e-spacio.uned.es/fez/view/bibliuned:master-ETSInformatica-IAA-Bnistal
Idioma eng
Versión de la publicación acceptedVersion
Nivel de acceso y licencia http://creativecommons.org/licenses/by-nc-nd/4.0
info:eu-repo/semantics/openAccess
Tipo de recurso master Thesis
Tipo de acceso Acceso abierto
Notas adicionales "Reproducido con permiso de Springer Nature"

 
Versiones
Versión Tipo de filtro
Contador de citas: Google Scholar Search Google Scholar
Estadísticas de acceso: 190 Visitas, 273 Descargas  -  Estadísticas en detalle
Creado: Fri, 30 Sep 2022, 17:51:29 CET