A Dempster-Shafer Approach for Corrupted Electrocardiograms Signals

Continuous streaming Electrocardiogram (ECG) data inthe Intensive Care Unit (ICU) is highly susceptible tonoise artifacts and signal corruption. Currently, the publicizedalgorithms for QRS detection do not account forunreliable lead information; waveform detection is typicallycontingent upon information from a single lead;and uncertainty metrics are not provided regarding thedetection accuracy. We propose a cross-correlation fusionmethod for multi-component ECG templates usingDempster-Shafer (DS) Theory. Our experiments usingclinical data were compared to benchmark nonsyntacticdetection algorithms where the detection accuracywas comparable at high signal-to-noise ratio (SNR). However, the fusion approach demonstrated asuperior increase in accuracy when the SNR degraded.Addressing these downfalls for the detection of QRScomplexes and other waveforms has potential to improvepatient risk prediction in the ICU.