Automated Identification of Paced Beats in Holter ECG

0551341 - ÚPT 2022 RIV US eng C - Conference Paper (international conference)
Plešinger, Filip - Viščor, Ivo - Smíšek, Radovan - Halámek, Josef - Bulková, V. - Nejedlý, Petr - Ivora, Adam - Matejková, M. - Leinveber, P. - Jurák, Pavel
Automated Identification of Paced Beats in Holter ECG.
Computing in Cardiology Conference. In: 2020 Computing in Cardiology (CinC 2020). New York: IEEE, 2020, (2020), č. článku 67. ISBN 978-1-7281-7382-5. ISSN 2325-8861.
[Computing in Cardiology 2020. Rimini (IT), 13.09.2020-16.09.2020]
R&D Projects: GA TA ČR(CZ) FW01010305
Institutional support: RVO:68081731
Keywords : ECG * EKG * signal processing * signal analysis * cardiac pacing
OECD category: Medical engineering
https://ieeexplore.ieee.org/document/9344294

Background: Identification of paced beats is a necessity for Holter ECG monitoring reports. However, lower sampling frequencies make the pacing stimuli hard to identify. In this study, we present a method to distinguish between paced and non-paced beats. Method: One-hour ECG recordings (158 patients, single lead, 250 Hz) were recorded during usual daily activities. A total of 44,918 QRS complexes were detected and marked as paced (19,004) or non-paced (25,914). This dataset was split (60%, out-of-patient) to training and testing datasets. Three features based on amplitude envelopes in two frequency bands were used to build a logistic regression model. An additional external dataset (2,193 recordings with 16,941 QRS) was assessed at a different facility and was used for the cross-database test. Results: The model showed the test Fl-score of 0.93, cross-database test shown Fl-score of 0.924. Conclusion: The presented method recognizes paced and non-paced heartbeats in lower sampling frequency, even if it can hardly be visually observed in the raw signal.
Permanent Link: http://hdl.handle.net/11104/0326763