False alarms in intensive care unit monitors: Detection of life-threatening arrhythmias using elementary algebra, descriptive statistics and fuzzy logic

Aims: A false alarm ratio of up to 86 % has been reported in Intensive Care Unit (ICU) monitors. Such a high value can lead to reduced staff attention and patient deprivation. We present a methodfor detection of specific arrhythmias - asystole, extreme bradycardia, extreme tachycardia, ventricular tachycardia and ventricular flutter I fibrillation - in accordance with the "2015 PhysionetlCinC Challenge". Data: The method was trained with the use of 750 records and tested on 500 records from ICUs provided by Physionet. Method: Invalid data segments are detected in each of the channels. Next, QRS complexes and RR intervals are found in all signals using a different QRS detection approach according to the signal source. The RR series obtained are tested for regular heart activity; if this fails, an arrhythmia-specific test is processed. Tests for individual arrhythmias are based on examination of QRS temporal distribution, comparison of heart rate (HR) with known limits, and observation of low-frequency ECG activities. Results: Training-set sensitivity and specificity of 96 % and 89 % were achieved. A hidden test set resulted in a score of 81.39 (real-time event) and 84.96 (retrospective event).