Investigating causal relationships between cardiovascular signals using effective connectivity assessment measures

In this paper, we apply a linear and nonlinear causal effects identification approach that has previously been shown to be capable of causality predictions in epileptic EEG data to detect such effects in a cardiovascular dataset obtained from a sleep apnea patient, including heart rate, respiration force, and blood oxygen concentration parameters. The method utilized is based on a nonlinear extension of Granger causality that uses nonlinear autoregressive exogenous (NARX) modeling for causality estimations. Furthermore, the dominant direction of causal interactions in the aforementioned cardiovascular dataset has been determined by assessing the predictability improvement (PI) of one signal alone and a signal under the effect of a second signal. The results of estimating the indexes for identifying linear and nonlinear causal effects demonstrate that there are bidirectional nonlinear causal effects between each pair of examined cardiovascular parameters. In addition, PI evaluation findings show that the dominant direction of causal relationships between heart rate and respiration force is from respiration force to heart rate, and blood oxygen concentration is affected by both parameters of heart rate and respiration force, but respiration appears to have a greater role in setting blood oxygen concentration.