University of Twente Student Theses


Machine learning classification of circulatory shock on the ICU using arterial waveforms

Mulder, M.P. (2020) Machine learning classification of circulatory shock on the ICU using arterial waveforms.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Abstract:Around one third of the patients submitted to the intensive care unit are in circulatory shock, a life-threatening state in which the circulatory system fails to deliver oxygen to the tissues. Early diagnosing and differentiating between the types of shock (cardiogenic, distributive, obstructive and hypovolemic) is vital. In this study arterial blood pressure waveform features are used as input data for machine learning classification techniques to diagnose the presence of circulatory shock and differentiate between shock types. A total of 484 patients are included in this study, of whom 293 have no signs of shock and 191 patients are diagnosed with shock. The random forest model reached an accuracy of 72.64% and an area under the receiver operator characteristic curve of 0.76 in identifying shock. Unfortunately, the classification of shock type was not possible with this data. Based on real world medical data, it is possible to detect circulatory shock with machine learning, but the accuracy is not high enough for implementation in clinical practice right now. Further data collection and more precise labelling is needed to increase the model's performance. In the future, machine learning algorithms could help diagnose hemodynamic unstable patients, maybe even before circulatory shock is present.
Item Type:Essay (Master)
Amsterdam UMC (location AMC), Amsterdam, The Netherlands
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