Robust Detection of Wide Complex Tachyarrhythmias Using Mobile Cardiac Telemetry Monitors

• Published in: Journal of cardiac failure Vol. 26; no. 10; pp. S77 - S78
• Main Authors: Mahajan, Ruhi, Gambhir, Alok, Adumala, Sameer
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.10.2020
• ISSN: 1071-9164; 1532-8414
• DOI: 10.1016/j.cardfail.2020.09.227

Wide Complex Tachyarrhythmias (WCT) are commonly observed arrhythmias in patients with heart failure. Automatic identification of WCT using mobile cardiac telemetry (MCT) monitors is critical for high-quality patient care due to their potentially lethal nature. There is very limited literature on validating an algorithm's ability in detecting WCTs in near real-time in the outpatient setting. This study demonstrates the efficacy of the ZywieAI® algorithm in detecting episodes of sustained and/or non-sustained WCTs like ventricular tachycardia, ventricular flutter, and ventricular fibrillation. We propose a three-pronged approach to identify WCT episodes in ECG recordings where the heart rate is ≥100 bpm for ≥3 consecutive beats. Firstly, the ECG frequency spectrum is analyzed to identify the time instants of QRS onset-offset and fiducial points in the signal. Secondly, four novel discrete features are extracted from the fiducial points and ectopic beats to get a beat-level assessment. Lastly, a rule-based decision classifier interprets input features to identify a true WCT episode. The ZywieAI® algorithm was evaluated by analyzing 299 WCT episodes in three public PhysioNet databases: MIT-BIH Arrhythmia (MITBA), MIT-BIH Malignant Ventricular Arrhythmia (MITMVA), and CU Ventricular Tachyarrhythmia (CUVT). These annotated databases contain ECG recordings with a wide variety of PQRST morphologies and arrhythmias contaminated with noise, which make the WCT detection a challenging problem. The proposed algorithm identified patients who experienced WCT with 100% accuracy in all three databases. At the individual episode-level, the algorithm had an average gross sensitivity (Se) and specificity (Sp) of 94.8% and 97.6% for MITBA and 94.1% and 89.1% for MITMVA databases, respectively. The episode-level Se for CUVT database was 100%. Furthermore, the Se of the algorithm in identifying any WCT episode with ≥ 10 beats was 100%, 99.8%, and 100% for the respective databases. The results suggest that the ZywieAI® algorithm can robustly identify WCT episodes as short as 0.46 seconds. The proposed algorithm can be used to evaluate real-time ECG signals in remote cardiac monitoring studies, assisting clinicians in the timely diagnosis of patients with most lethal wide complex arrhythmias.