University of Twente Student Theses

Login

Out-of-hospital ventricular fibrillation waveform analysis : towards a multi-lead defibrillator for in-field detection of acute myocardial infarction

Sluijs, K.M. van der (2020) Out-of-hospital ventricular fibrillation waveform analysis : towards a multi-lead defibrillator for in-field detection of acute myocardial infarction.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Embargo date:31 December 2022
Abstract:Globally, 34.7 per 100,000 inhabitants suffer from out-of-hospital cardiac arrest (OHCA) each year, of which approximately 30% presents with a cardiac rhythm of ventricular fibrillation (VF). Less than 20% survives; the only available therapy is defibrillation, aimed at achieving a return of spontaneous circulation. The most frequent underlying cause of VF is acute myocardial infarction (AMI). In-field identification of AMI could individualise resuscitation care and improve survival. The paddle electrocardiogram (ECG) recorded by the defibrillator provides unique, patient-specific information in an early phase of the arrest. Quantified using ventricular fibrillation waveform characteristics, the VF-waveform on the ECG can be analysed in real-time. Research suggests that VF-morphology is affected by old myocardial infarction (OMI) and even more by AMI. Studies have therefore focused on detecting OMI as a surrogate for AMI. A proof-of-concept machine learning study applying waveform analysis on 12-lead ECGs of VF induced for implantable cardioverter-defibrillator testing showed that lead II could identify OMI and that detection improved when twelve leads were used. This method lacked an optimisation of input features and did not combine specific leads. The first aim of this thesis was to investigate the effect of established feature selection methods on the ability of support vector machines to discriminate between patients with (n=137) and without (n=105) OMI. Models of lead II, twelve leads and lead II + V1 reached an area under the curve (AUC) of 0.58, 0.83 and 0.76 respectively. The results showed that feature selection and additional leads improved the detection of OMI. The main limitation was that VF was induced electrically, indicating that these methods require validation in the out-of-hospital setting with spontaneous VF and AMI. The second aim was to assess the performances of similar models based on the paddle ECG in a real-world OHCA cohort to discriminate between patients with (n=62) and without (n=40) AMI. Models based on a single VF segment of the resuscitation reached AUCs of 0.74 and 0.72. A model including both segments had an AUC of 0.76. Incorporating the evolution of the VFwaveform over time and including organisation-related measures led to an acceptable detection of AMI using the paddle ECG. Concluding, this thesis demonstrated that selecting relevant features of multiple ECG leads and segments improves the detection of myocardial infarction. Clinical implementation of multi-lead models that detect AMI during the arrest is the next step to facilitate individualised OHCA treatment and improve survival.
Item Type:Essay (Master)
Faculty:TNW: Science and Technology
Subject:44 medicine
Programme:Technical Medicine MSc (60033)
Link to this item:https://purl.utwente.nl/essays/84036
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page