Real-time Fusion of ECG and SpO2 Signals to Reduce False Alarms

Abstract

The aim of this study was the reduction of false arrhythmia alarms and improvement in the accuracy of calculated heart rates (HR) in real-time patient monitors by the fusion of information from ECG and SpO2 signals. These signals were analyzed independently to derive features for use in the fusion analysis. Information regarding the detection and classification of QRSs, signal quality index (SQI), HR and arrhythmia alarms were obtained from the ECG. Pulses, signal quality, pulse rate (PR) and hemodynamic parameters were obtained from the SpO2. Independent results from each signal were confirmed by fusing features from both. When an arrhythmia alarm was triggered, corresponding SpO2 features were checked to determine whether the alarm should be accepted or rejected HR and PR reliability was estimated using signal quality while QRSs and SpO2 pulses were matched to exclude spurious beats due to motion or other artifacts. Thus, calculated HR and PR were more accurate and false HRs from noisy ECG signals could be supplanted by the pulse rate if noise free. Adult, pediatric and neonate ECG signals were collected to make up training and test datasets. False alarm suppression was over 50% for all arrhythmia calls while it was over 60% for life threatening alarms. False HR/PR was reduced by 80%.