Quickly finding a needle in a haystack: a new automated cardiac arrhythmia detection software for preclinical studies

Florence Koeppel,David Labarre,Philippe Zitoun

DOI: https://doi.org/10.1016/j.vascn.2012.04.008

Abstract:Introduction: The occurrence of drug-induced arrhythmias in safety pharmacology or toxicology studies is a primary safety concern. The risk assessment requires an accurate knowledge of background arrhythmia incidence and frequency in the test system/species, as well as a rigorous evaluation of the effects of the potential new medical entities on the electrocardiogram (ECG). However, the direct assessment of arrhythmia in ECG recordings is a time-consuming effort and is rarely achieved due to lack of suitable automated tools. A new software application named ARR30a was developed for fast automated detection in preclinical studies, for the five major arrhythmia types, namely sinus pauses, atrial beats, junctional beats, ventricular beats and type 2 atrio-ventricular blocks (AV-blocks II). The purpose of this study was to characterize the performance of ARR30a in large and small animal species. Methods: Detection sensitivity and predictivity were evaluated on a database of 84 ECG recordings representative of each animal species and experimental protocols typically used in efficacy, safety pharmacology and toxicology studies. Automated arrhythmia detection was compared with manual analysis. Results: In large animals such as dogs, non-human primates and pigs, ARR30a sensitivity reached 90.6%, 82.2% and 78.0% for ventricular beats, AV-blocks II and junctional beats with predictivity of 83.4%, 94.4% and 93.5%, respectively. Significantly lower sensitivity was observed in rats for junctional beats due to challenging problems of detection for low amplitude P-waves. Robustness to noise was assessed by adding increasing noise levels to ECG signals and showed no significant impact on arrhythmia detection at moderate noise levels. Processing time for a 24 hour recording was approximately 4 min for dogs and 6 min for rats on a 3 GHz processor. Discussion: This newly validated ECG arrhythmia detector ARR30a allows evaluating all major ECG signal abnormalities and enhances the quantification of arrhythmia incidence in all major laboratory animal species. The mark editor RME10a enables manual validation of the automated analysis and refinement of the arrhythmia classification.