Respiratory Sinus Arrhythmia Estimation: Closing the Gap between Research and Applications

The respiratory sinus arrhythmia (RSA) is a form of cardiorespiratory coupling in which the heart rate accelerates during inhalation and decelerates during exhalation. Its quantification has been suggested as a tool to assess different diseases and conditions. However, whilst the potential of the RSA estimation as a diagnostic tool is shown in research works, its use in clinical practice and mobile applications is rather limited. This can be attributed to the lack of understanding of the mechanisms generating the RSA. To try to explain the RSA, studies are done using noninvasive signals, namely, respiration and heart rate variability (HRV), which are combined using different algorithms. Nevertheless, the algorithms are not standardized, making it difficult to draw solid conclusions from these studies. Therefore, the first aim of this thesis was to develop a framework to evaluate algorithms for RSA estimation. To achieve this, a model of the cardiorespiratory system, in which the RSA strength is easily changed while other modulators of the heart rate are controlled, was proposed. This model offers a framework to evaluate existing and newly developed RSA estimation algorithms in an objective and quantitative manner. It was applied to compare 7 state-of-the-art algorithms, for which recommendations of interpretation were given. The second aim was to assess the robustness of RSA algorithms when calculated using signals from patients and in noisy recordings. This was done by quantitatively evaluating their deterioration in four scenarios. The first one was the effect of cardiac abnormalities present in the recordings. The second one was the occurrence of delays between the respiration and HRV signals. The third one was the change of phase between the signals. The last one was the parametrization of the methods. The estimates were found to be particularly sensitive to abnormalities and changes of phase, but not to delays nor to the parametrization. The third aim was to develop a method to quantify the linear and nonlinear components of the RSA. For this, a machine learning-based approach was proposed. It allowed to identify the dominant form of coupling, and it was applied to a dataset of sleep apnea patients, in which the nonlinear part of the interactions was not significant. The fourth aim was to evaluate the RSA estimates in different applications focusing on ambulatory monitoring. To achieve this, attention was given to the use of HRV, noninvasive respiratory signals and the respiration derived from the electrocardiogram, modalities that can be easily used in mobile setups. In addition, the methods were applied in different datasets, including signals from sleep apnea patients and from healthy volunteers. On each case, the added value of the RSA estimates was demonstrated.\\ This thesis offers tools to improve the understanding of the RSA and might help to close the gap of its use in research and in clinical practice.

Jaar van publicatie:2021

Toegankelijkheid:Open