Comparison Between Linear and Nonlinear Haar Wavelet for the Detection of the R-peak in the Electrocardiogram of Small Animals

The electrocardiogram is a great resource for diagnosis in humans and animals alike, as well as an important tool in translational medicine and biology. To better explore signal processing advances on small animals’ ECG, this paper compares linear and nonlinear Haar Wavelet Transform performance on R-peak detection as compared to traditional threshold-based algorithms. Original dog, rabbit and mice records were obtained from PhysioZoo platform, as well as the standard ECG processing algorithm. Algorithms performances were compared on mean and maximum position errors on R-peak detection. All wavelet processing algorithms were written in Python. Nonlinear Haar WT mean error figures were 10 times lower than its linear counterpart for mice and almost 100 times lower for dogs and rabbits. The nonlinear Haar WT also outperformed linear Haar WT on the maximum error indicator, but on the same order of magnitude. We suggest that the PhysioZoo platform should be further explored to advance translational medicine in cardiology by equalizing signal processing techniques application among electrocardiographic records of human and small animals alike.