Suppression of false arrhythmia alarms using ECG and pulsatile waveforms

In the context of the 2015 PhysioNet/CinC Challenge we present an algorithm to detect false arrhythmia alarms in the Intensive Care Unit(ICU). Our focus is on life threatening arrythmia alarms: asystole, extreme bradycardia, extreme tachycardia, ventricular tachycardia and ventricular fibrillation or flutter. Our method uses simultaneous ECG and pulsatile waveforms, photoplethysmogram or arterial blood pressure, to detect false alarms. QRS detectors produce for each signal a set of QRS detections that can be used to detect false alarms. Often the conclusions drawn from the results of QRS detectors on different signals are contradictory: some of the signals may be contaminated by noise or simply get lost for a while. Evaluating the signal quality of each waveform is necessary to can decide if we can trust the QRS detections obtained on that waveform. We describe a method to choose in each case which set of QRS detections should be used to conclude if the alarm is true or not. A set of rules is used for each alarm type.