Quantifying cardiac parasympathetic and sympathetic function based on a weighted-principal component regression method

A quantitative evaluation of autonomic cardiovascular control is important in understanding basic pathophysiological mechanisms or for patient monitoring, treatment design and follow-up. Noninvasive techniques for this purpose have been the focus of many research endeavors. We previously proposed a method to extract pure parasympathetic and pure sympathetic indices based on the impulse response between instantaneous lung volume and heart rate. Identification of this impulse response involves a dual-input, single-output system in which one input interacts with the output in closed-loop. To identify this relatively complicated system, we propose here a new system identification technique based on a weighted-principal component regression method. Asymptotically, this technique implements model selection in the frequency domain. Therefore, in contrast to the conventional methods, it allows the data to play a significant role in determining candidate models. Moreover, the estimated model parameters reflect a trade-off between bias and variance to reach a relatively small mean squared prediction error. We employ experimental data to demonstrate that this technique is superior to a more traditional technique in terms of measuring cardiac autonomic indices.