Detection of exercising ectopic atrial and ventricular beats using non-linear analysis of clinically normal electrocardiograms at rest or low-intensity exercise
V Alexeenko, H Tavanaeimanesh, F Stein, J Gold, L Hughes, M Mccue, C Marr, S Durward-Akhurst, K JeevaratnamAbstract
Background
Cardiac arrhythmias in healthy athletic horses are common. In horses, like in humans, arrhythmias that develop during exercise lead to poor athletic performance and are thought to lead to exercise-associated sudden death. Due to the intermittent nature and potential high risk associated with these arrythmias, early detection of high-risk horses is an important goal of cardiovascular diagnostics. We have demonstrated that non-linear methods assessing the disorderliness of equine electrocardiograms (ECG), using brief, artifact-free recordings of normal sinus rhythm ECG might be used to identify horses with past episodes of paroxysmal atrial fibrillation (PAF).
Purpose
We hypothesised that ECG complexity-based techniques might be productive in identification of horses exhibiting intermittent ectopic atrial and ventricular heart rhythm abnormalities triggered by pathophysiological mechanisms different from PAF.
Methods
In a convenience prospective cross-sectional study, ambulatory ECGs were recorded using the Televet 100 or II devices from 91 Thoroughbred or Standardbred racehorses during routine training sessions. During these sessions the heart rate exceeded 100 bpm for 15-20 minutes and was in the 60-100 bpm range for the rest of the time. Horses classified as cases had at least five premature depolarizations and/or at least one episode of complexity, including couplets, triplets, and/or supraventricular or ventricular tachycardia during high intensity exercise (heart rate > 180 beats per minute) or immediately after exercise; horses classified as cases had less than five premature complexes before, during, and after exercise. Acceptable quality 60-second ECG strips with stable heart rate (20-120 beats per minute) were identified automatically. Disorderliness of the ECG was estimated using Lempel-Ziv’76 and ’78, and Titchener complexity, and Shannon, sample, and approximate entropy algorithms, using several data preprocessing algorithms to highlight the role of individual fiducial points within the ECG waveform. Numerical values obtained by complexity estimation algorithms were corrected to the heart rate.
Results
We found that for optimal diagnostic performance recordings of 60-100 beats per minute should be used. The best-performing algorithm used Lempel-Ziv ’76 complexity, with R peak and ends of S and T peaks as fiducial points. The receiver operating curve analysis has demonstrated the area under curve of 0.86 for this combination, indicating acceptable differentiation between cases and controls, with sensitivity of 85.7% (confidence interval, CI 59.8-100%), specificity 83.3% (CI 73.9%-92.8%), positive predictive value of 37.5% (CI 13.8- 61.2%) and negative predictive value of 98.0% (CI 94.2-100%).
Conclusions
Our findings suggests that non-linear analysis could be a useful screening tool to identify racehorses that should undergo exercising ECGs to determine the frequency and severity of exercising cardiac arrhythmias.