Reduction of false critical ECG alarms using waveform features of arterial blood pressure and/or photoplethysmogram signals

To address the Physionet/Computing in Cardiology (CinC) Challenge 2015 [1], this article presents practical algorithm to reduce false critical ECG alarms using waveform features of arterial blood pressure (ABP) and/or photo-plethysmogram (PPG). The ABP/PPG pulse features are extracted on a beat-by-beat basis in real-time manner. For each detected pulse (or forced detection), 5 event feature indicators (EFIs), which correspond to the 5 critical ECG arrhythmia alarms, are generated from a group of pulses prior to the current pulse. At the time of a critical arrhythmia alarm, the corresponding EFI values of those ABP/PPG pulses prior to or around the alarm time are checked for adjudicating (accept/reject) this alarm. As an extension or option, available electrocardiogram (ECG) signals are included in a way similar to ABP/PPG processing. Quantitative results on the Challenge training and test datasets are presented. The algorithm is practical owing to its real-time processing mechanism and computational efficiency.