Open AccessHEART BEAT DETECTION IN NOISY ECG SIGNALS USING STATISTICAL ANALYSIS OF THE AUTOMATICALLY DETECTED ANNOTATIONS / ŠIRDIES DŪŽIŲ NUSTATYMAS IŠ IŠKRAIPYTŲ EKG SIGNALŲ ATLIEKANT AUTOMATIŠKAI APTIKTŲ ATSKAITŲ STATISTINĘ ANALIZĘ
Author(s) -
Andrius Gudiškis
Publication year - 2015
Publication title -
mokslas - lietuvos ateitis
Language(s) - English
Resource type - Journals
eISSN - 2029-2341
pISSN - 2029-2252
DOI - 10.3846/mla.2015.787
Subject(s) - heartbeat , qrs complex , beat (acoustics) , signal (programming language) , computer science , detector , artificial intelligence , pattern recognition (psychology) , noise (video) , signal to noise ratio (imaging) , speech recognition , physics , acoustics , telecommunications , medicine , cardiology , computer security , programming language , image (mathematics)
This paper proposes an algorithm to reduce the noise distortion influence in heartbeat annotation detection in electrocardiogram (ECG) signals. Boundary estimation module is based on energy detector. Heartbeat detection is usually performed by QRS detectors that are able to find QRS regions in a ECG signal that are a direct representation of a heartbeat. However, QRS performs as intended only in cases where ECG signals have high signal to noise ratio, when there are more noticeable signal distortion detectors accuracy decreases. Proposed algorithm uses additional data, taken from arterial blood pressure signal which was recorded in parallel to ECG signal, and uses it to support the QRS detection process in distorted signal areas. Proposed algorithm performs as well as classical QRS detectors in cases where signal to noise ratio is high, compared to the heartbeat annotations provided by experts. In signals with considerably lower signal to noise ratio proposed algorithm improved the detection accuracy to up to 6%