Computational system for heart rate variability analysis

Abstract

The study of heart rate variation (HRV), also called cardiac frequency variability, is used to identify the range of samples where the respiratory sinus arrhythmia occurs and presents information of low frequency (LF) and high frequency (HF) densities. Cardiofrequency meters (Polar WearLink) are widely used heart rate monitors in sports and clinical practice. The sensor registers the values of the R waves (RR) intervals of the electrocardiogram (ECG) to be analyzed by the specialist. This process can introduce errors and deletion of important data due to human error. This work aims at designing a computational system to analyze the non-stationary signals to aid the specialist in cardiovascular diagnosis. The methodology is performed in three steps: (1) pre-processing with a compact support median filter to get rid of the outliers, then passing through a mapping of RR intervals to beats per minute (bpm) to verify maximum frequency, tachycardia and bradycardia in the time domain. To determine the outliers, a threshold is chosen to be applied to the bpm values according to the minimum and maximum values of the normal RR intervals, which depend on genre and age; (2) process with Spectrograms and continuous and discrete Wavelet transforms to improve the diagnosis; (3) visual interpretation of these results by the specialist by verifying that these results are consistent. The results, with the available data from the volunteers, show that the system proved to be consistent to help the specialist improve the accuracy in the cardiovascular diagnosis.