Maskininlärningsmodell för prognosbedömning efter hjärtstopp
(2026) EEML05 20261Division for Biomedical Engineering
- Abstract
- Cardiac arrest is a life-threatening condition in which the heart stops pumping blood to the body, leading to loss of oxygen supply to vital organs. Despite cardiopulmonary resuscitation (CPR), early defibrillation and advances in emergency care, many survivors still suffer from neurological injury due to brain damage from loss of oxygen.
Predicting neurological outcome, or even the chances of survival, within 72 hours remains highly challenging. This might lead to inefficient misallocation of resources in emergency care units. At the same time, relatives are left uncertain of the outcome for a long time, which causes stress and psychological impact.
The aim of this project was to create a machine learning model for outcome... (More) - Cardiac arrest is a life-threatening condition in which the heart stops pumping blood to the body, leading to loss of oxygen supply to vital organs. Despite cardiopulmonary resuscitation (CPR), early defibrillation and advances in emergency care, many survivors still suffer from neurological injury due to brain damage from loss of oxygen.
Predicting neurological outcome, or even the chances of survival, within 72 hours remains highly challenging. This might lead to inefficient misallocation of resources in emergency care units. At the same time, relatives are left uncertain of the outcome for a long time, which causes stress and psychological impact.
The aim of this project was to create a machine learning model for outcome prediction within 24 hours after cardiac arrest. Two different machine learning models were tested based on their ability to analyze complex data and predict an outcome; Random Forest and Logistic regression. They were trained on a database of 191 patients with cardiac arrest, where approximately 50\% survived. The database included variables such as age, if CPR was performed, pH and electroencephalography measurements (EEG). The models were developed and analyzed according to accuracy, precision, recall and F1-score. This resulted in a prediction model of two steps; the first model is based on Random Forest and predicts the chances of survival, while the second model is based on Logistic regression and predicts the level of neurological complications. The models identified EEG and pH as some of the most important variables to predict the outcomes. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9237302
- author
- Glitterstam, Lydia LU and Liljeholm, Caroline LU
- supervisor
- organization
- alternative title
- Machine learning model for outcome prediction after cardiac arrest
- course
- EEML05 20261
- year
- 2026
- type
- M2 - Bachelor Degree
- subject
- keywords
- Maskininlärning, Artificiell intelligens, AI, Machine learning, Random Forest, Logistisk regression, Linjär regression, Prediktionsmodell, Prediktionsbedömning, HJärtstopp, Cardiac arrest, Utfallsprediktion
- language
- Swedish
- id
- 9237302
- date added to LUP
- 2026-06-23 12:43:15
- date last changed
- 2026-06-23 12:43:15
@misc{9237302,
abstract = {{Cardiac arrest is a life-threatening condition in which the heart stops pumping blood to the body, leading to loss of oxygen supply to vital organs. Despite cardiopulmonary resuscitation (CPR), early defibrillation and advances in emergency care, many survivors still suffer from neurological injury due to brain damage from loss of oxygen.
Predicting neurological outcome, or even the chances of survival, within 72 hours remains highly challenging. This might lead to inefficient misallocation of resources in emergency care units. At the same time, relatives are left uncertain of the outcome for a long time, which causes stress and psychological impact.
The aim of this project was to create a machine learning model for outcome prediction within 24 hours after cardiac arrest. Two different machine learning models were tested based on their ability to analyze complex data and predict an outcome; Random Forest and Logistic regression. They were trained on a database of 191 patients with cardiac arrest, where approximately 50\% survived. The database included variables such as age, if CPR was performed, pH and electroencephalography measurements (EEG). The models were developed and analyzed according to accuracy, precision, recall and F1-score. This resulted in a prediction model of two steps; the first model is based on Random Forest and predicts the chances of survival, while the second model is based on Logistic regression and predicts the level of neurological complications. The models identified EEG and pH as some of the most important variables to predict the outcomes.}},
author = {{Glitterstam, Lydia and Liljeholm, Caroline}},
language = {{swe}},
note = {{Student Paper}},
title = {{Maskininlärningsmodell för prognosbedömning efter hjärtstopp}},
year = {{2026}},
}