University of Twente Student Theses


Clustering patients at the emergency department based on their ECG and PPG signals

Bot, M. (2023) Clustering patients at the emergency department based on their ECG and PPG signals.

[img] PDF
Abstract:Early diagnostics at the emergency department (ED) has several benefits, such as reduced mortality and improved patient satisfaction. Studies have shown that machine learning models can determine various patient characteristics, such as age, gender and diabetes, based on their electrocardiography (ECG) or photoplethysmography (PPG) signals, highlighting the potential that more clinical parameters can be determined using these signals. This study aims to uncover which patient characteristics are most suitable to use in classification machine learning models to aid in early diagnostics. Patients enrolled in the Acutelines data-bio bank with ECG and PPG recordings available in the first half hour after admission to the ED were included in this study. The data was cleaned and a 30-second interval of clean data was selected for each patient. The dimensionality of these intervals was reduced using an autoencoder. Consequently, K-means multidimensional time series clustering was performed on the latent features. Four clusters were found. Mortality or ICU admission within 24 or 72 hours, various blood measurements, the likelihood and focus of an infection and SOFA score are parameters that differed significantly between the clusters. It remains unclear, however, which aspects of the ECG and PPG signals specifically form the basis of this clustering.
Item Type:Essay (Master)
Faculty:TNW: Science and Technology
Subject:44 medicine, 50 technical science in general
Programme:Technical Medicine MSc (60033)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page