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

Overview

• This study evaluates how well two equine ECG recording devices, Televet™ and Equimetre™, agree when measuring ECG complexity in horses.
• The goal is to ensure that heart signal complexity results are consistent across different devices, which is important for developing AI-based diagnostics in equine cardiology.

Background

• Non-linear analysis of equine electrocardiograms (ECGs) is a growing field focused on understanding the complexity of heart rhythms in horses.
• ECG complexity analysis has potential applications in creating artificial intelligence (AI) tools to diagnose and monitor equine heart conditions.
• With multiple ECG devices on the market, it is crucial to verify that measurements from different devices are comparable and can be used interchangeably in research and clinical practice.

Objective

• The main objective was to compare the ECG complexity values obtained simultaneously from two different recording devices: Televet™ and Equimetre™.
• The study aimed to assess whether the differences between devices influenced the analysis results.

Methods

• Study Design: A cross-sectional clinical study involving 37 healthy Thoroughbred horses during routine training.
• Data Collection:
  • ECGs were recorded simultaneously using both Televet™ and Equimetre™ devices on each horse.
  • Automated algorithms extracted high-quality 60-second ECG segments with stable heart rates between 30 and 100 beats per minute (bpm).
• ECG Analysis:
  • Three coarse-graining methods were applied to annotate the ECG for complexity analysis:
    • Threshold-crossing
    • Beat detection
    • Feature detection
  • Complexity values were corrected based on heart rate using the data from all 37 horses.
• Statistical Analysis:
  • Agreement between devices was evaluated using Bland-Altman plots to visualize consistency.
  • Student’s t-tests were used to assess statistical significance of differences using data from 28 horses who had sufficient ECG segments from both devices.

Results

• Beat Detection Coarse-Graining:
  • Complexity results were consistent and independent of the device used across all heart rates.
• Feature Detection Coarse-Graining:
  • Results differed significantly between devices for heart rates below 75 bpm.
• Threshold-Crossing Coarse-Graining:
  • Results differed significantly between devices for heart rates above 75 bpm.

Conclusions

• Comparison of ECG complexity between devices is affected by the heart rate range and the coarse-graining technique applied.
• Beat detection coarse-graining provided complexity values that were comparable across devices within the studied heart rate range.
• Results from feature detection and threshold-crossing methods showed device-dependent differences at specific heart rate ranges, indicating these analyses require caution when applied to data from different ECG devices.
• The study’s convenience sampling and limited data availability in some heart rate ranges are limitations that may affect generalizability.
• Overall, the findings emphasize the importance of considering device differences, heart rate ranges, and analysis methods when interpreting equine ECG complexity results to ensure valid and comparable outcomes.