Annotating ECG signals using echo state networks as a time series classifier

Lange, Lennard de (2022) Annotating ECG signals using echo state networks as a time series classifier. Bachelor's Thesis, Artificial Intelligence.

Preview	Text BachProj16052022.pdf Download (319kB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (118kB)

Abstract

The variability of the interval between two heartbeats, known as the heart rate variability, has been found to be an indicator of many physiological conditions. The methodology of analysis of it consists recording a raw ECG signal and annotating the individual Q-, R-, or S-, peaks within the signal, and then through a discrete Fourier transform on the interbeat interval the power spectrum analysis is executed. The stage of annotating the individual peaks has become partially automated, but still requires manual labour. Threshold based algorithms, such as PreCar , partially automate the annotation of the individual peaks, but such algorithms can be faulty when faced with artifacts in the signal. This paper sets out to assess whether an echo state network is a suitable method to fully automate this process, in comparison to the threshold based algorithm. It was found that the F1-scores of the echo state network (M = 0.715, SD = 0.054) were significantly lower than the F1-scores of the threshold based algorithm (M = 0.886, SD = 0.135). Therefore it is concluded that the tested echo state network is not a more accurate tool in automatically annotating ECG signals than PreCar.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Cnossen, F.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	27 May 2022 07:46
Last Modified:	27 May 2022 07:46
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/27083

Actions (login required)

View Item