Detección de la mortalidad temprana en pacientes con traumatismo craneoencefálico mediante técnicas de Machine Learning
Loading...
Download
Official URL
Full text at PDC
Publication date
2022
Defense date
2022
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
Los traumatismos craneoencefálicos continúan siendo un grave problema de salud para la sociedad, además de que no se ha conseguido reducir su tasa de mortalidad. Dada sus complejas afecciones, el tiempo de respuesta hospitalario tras sufrir la lesión es clave para salvar la vida de los pacientes, dado que puede llegar a producir en los individuos una mortalidad temprana.
En el presente estudio se fija como objetivo la predicción de la mortalidad temprana en pacientes que ingresan con la patología de traumatismo craneoencefálico mediante la aplicación algoritmos de Machine Learning, lo que puede ayudar a que el equipo médico disponga de suficiente tiempo para revertir la situación.
Para este caso, se puede observar a lo largo del estudio que la aplicación de inteligencia artificial a diferentes características y síntomas de los pacientes no modifica de forma notable la predicción que aporta un modelo de regresión logística, con un área bajo la curva ROC de 0.9274 en un intervalo de confianza al 95% de: 0.8966-0.9582. Por lo que, dada su estructura menos compleja, nos permite seleccionarlo como mejor modelo para predecir que suceda un fallecimiento temprano.
Además de la elección del mejor modelo, se analizan las diferentes características de los pacientes, pudiendo obtener conclusiones sobre distintas pruebas y evaluaciones médicas como que la reacción pupilar o la evaluación inicial del motor score, son esenciales para poder determinar el estado y la probabilidad de la mortalidad temprana tras sufrir un traumatismo craneoencefálico.
Traumatic brain injuries continue to be a serious health problem for society, and their mortality rate has not yet been reduced. Given the complexity of their conditions, the hospital response time after injury is key to saving patients' lives, as it can lead to early mortality in individuals. This study aims to predict early mortality in patients admitted with traumatic brain injuries through the application of Machine Learning algorithms, which can help the medical team to have enough time to reverse the situation. In this case, it can be observed throughout the study that the application of artificial intelligence to different characteristics and symptoms of the patients does not notably modify the prediction provided by a logistic regression model, with an area under the ROC curve of 0.9274 with a 95% confidence interval of 0.8966-0.9582. Therefore, given its less complex structure, it allows us to select it as the best model for predicting early death. In addition to the choice of the best model, the different characteristics of the patients are analysed, being able to obtain conclusions about different tests and medical evaluations, such as the pupillary reaction or the initial evaluation of the motor score, which are essential to be able to determine the state and probability of early mortality after suffering a traumatic brain injury.
Traumatic brain injuries continue to be a serious health problem for society, and their mortality rate has not yet been reduced. Given the complexity of their conditions, the hospital response time after injury is key to saving patients' lives, as it can lead to early mortality in individuals. This study aims to predict early mortality in patients admitted with traumatic brain injuries through the application of Machine Learning algorithms, which can help the medical team to have enough time to reverse the situation. In this case, it can be observed throughout the study that the application of artificial intelligence to different characteristics and symptoms of the patients does not notably modify the prediction provided by a logistic regression model, with an area under the ROC curve of 0.9274 with a 95% confidence interval of 0.8966-0.9582. Therefore, given its less complex structure, it allows us to select it as the best model for predicting early death. In addition to the choice of the best model, the different characteristics of the patients are analysed, being able to obtain conclusions about different tests and medical evaluations, such as the pupillary reaction or the initial evaluation of the motor score, which are essential to be able to determine the state and probability of early mortality after suffering a traumatic brain injury.