This poster is part of the arrhythmia detection project: the development of a statistical model for the detection and classification of cardiac arrhythmias in sports horses, carried by the EQUIMETRE tool. This first work enables to formalize the problem of arrhythmia detection and thus to translate it mathematically and propose a first solution.
The detection of arrhythmias requires a segmentation of ECGs as a beat affected by an arrhythmia is significantly longer than a healthy beat. However, the segmentation algorithm makes errors, beats are missed, and others are added, these errors are respectively confused with arrhythmias and healthy beats. To solve this problem, we try to consider other variables, in addition to the elapsed time.
To show the impact of these variables, statistical tests are performed on simulations. We show that with a perfect segmentation, we could detect arrhythmias using only the elapsed time, with segmentation errors, the detection is bad but with the new variables, the detection is improved.
To make these simulations, assumptions had to be made but they are not necessarily close to reality, the future work will be to change or relax these assumptions by using test statistics adapted to the problem.
Keywords: arrhythmia detection, cardiac arrhythmia, heart, racehorse equine veterinarian, equine researcher