Agreement of the performance of equine electrocardiogram recording devices for ECG complexity analysis.

Abstract

BACKGROUND: Non-linear equine electrocardiography (ECG) analysis is an actively developing study area which has the potential to lead to novel, artificial intelligence-based diagnostic tools in equine cardiology. As more ECG recording devices are becoming available, there is a need to ensure results are interchangeable regardless of the equipment used to record the equine ECG. OBJECTIVES: To evaluate the agreement of ECG complexity values obtained using the Televet™ and Equimetre™ systems. STUDY DESIGN: Cross-sectional clinical. METHODS: ECGs were recorded using two devices simultaneously from 37 healthy Thoroughbred horses during routine training. An automated algorithm extracting the ECG segments of acceptable quality extracted 60-second strips with a stable heart rate in the range 30-100 beats per minute. Threshold-crossing, beat detection, and feature detection coarse-graining algorithms were used to annotate the ECG for complexity analysis. Complexity values were corrected to the heart rate using data from 37 horses, and inter-device agreement was evaluated using Bland-Altman plots and Student's t-test using ECG data from 28 horses that provided sufficient data from both devices. RESULTS: The results of complexity analysis obtained with beat detection coarse-graining were independent of the device used at all heart rates. The results obtained with feature detection for heart rates below 75 beats per minute (bpm) and with threshold crossing for heart rates above 75 bpm were significantly different. MAIN LIMITATIONS: The study relied on convenience sampling, and data analysis was constrained by the availability of ECG data in the heart rate range of interest. CONCLUSIONS: The accurate comparison of ECG complexity analysis results requires consideration of differences between recording devices, heart rates and ECG coarse-graining techniques.